Michigan State University Extension
Tourism Educational Materials - 33519758
06/06/02
MEASURING TOURISM IMPACTS AT THE COMMUNITY LEVEL
List of files and visuals associated with this text.
Source: Maine Agricultural Experiment Station
Author: Reiling, Stephen (Editor)
ID: Miscellaneous Report 374
Year: 1992
Measuring Tourism Impacts at the Community Level
edited by
Stephen D. Reiling, Professor
Department of Agricultural and Resource Economics
College of Applied Sciences and Agriculture
University of Maine
Orono, Maine 04469
CONTENTS
Introduction
Assessing Changes in Tourism in the Northeast
T. L. Brown and N.A. Connelly
Classification of Recreation and Tourism Communities Using
Cluster Analysis and Data Management Techniques
M.I. Bevins and R.R. Zwick
Tourism Development Simulation Model: The Game
T.J. Tyrrell
A System for Estimating Local Economic Impacts of
Recreation and Tourism
D.J. Stynes and D.B. Propst
The Impact of Tourism on Local Government Public Service
Expenditures
M.F. Teisl and S.D. Reiling
INTRODUCTION
Many rural communities in the U.S. have experienced
significant change in the last two decades. Some have
prospered from the urban-to-rural migration patterns of
the 1970's and expanded employment opportunities; others
have declined because of the farm problems of the 1980's
and the lack of employment opportunities for residents. In
either case, the changes have made the governance of rural
America more complex. An influx of people from urban areas
with different values and expectations creates new demands
for local government services. However, the revenue base
of rural communities may not be sufficient to fulfill the
expectations of new residents. In those communities with
constant or declining populations, community leaders are
seeking new economic opportunities to stabilize or reverse
the economic problems they face.
In the Northeast, one of the viable economic opportunities
available to rural communities continues to be tourism and
related outdoor recreational activities. Tourism currently
contributes to the economic base of many rural communities
in the region, and other communities are interested in
attracting tourism development to increase job
opportunities. The impacts of tourism development are
usually local in nature, and planning decisions regarding
such developments are made by local officials. However,
community leaders often lack the knowledge to effectively
evaluate the positive and negative impacts that tourism
may have on the local area. Because of this lack of
information, rural community leaders are seeking
assistance from extension and research personnel at the
land-grant universities.
In 1986, a six-year Northeast regional research project
entitled "Tourism Impacts and Development Alternatives
from the Local Rural Government Perspective"(commonly
referred to as NE-163) was initiated to address some of
the issues facing local government officials regarding
tourism development. Researchers from the land-grant
universities in the following states have participated in
the project: Maine, Vermont, Massachusetts, Rhode Island,
Connecticut, New York, New Jersey, Delaware, and Michigan.
Representatives from New Jersey and Connecticut retired
early in the life of the project and were not replaced.
One of the objectives of the project is to develop, test,
and refine procedures that would enable local officials to
evaluate the relative benefits and costs of future tourism
development alternatives. The purpose of this publication
is to present the results of several studies conducted by
members of the NE-163 research committee that contributed
to this objective.
The first two papers address the question of identifying
the importance of tourism in the states, counties, or
communities in the Northeast. The study by Tom Brown at
Cornell University uses shift-share analysis and location
quotients for the lodging sector to assess changes in the
tourism industry in the region. The results show that the
tourism sector, as approximated by lodging receipts, grew
at a rate greater than the overall retail sector in the
Northeast. However, the rate of growth varied
substantially among states and counties, thereby causing
significant changes in the geographical distribution of
tourism activity within the region. The results are useful
in that they identify the counties where tourism activity
is concentrated in the region. This allows local
communities to determine where they stand in terms of
current tourism activities, relative to other counties in
the Northeast.
The paper by Mal Bevins of the University of Vermont and
Rod Zwick of Lyndon State College in Vermont uses cluster
analysis to classify Vermont communities. Two of the seven
community clusters identified are related to tourism. The
results of the cluster analysis are then used to determine
whether a simpler procedure based on secondary data could
be used to classify communities. A spreadsheet approach
based on key community characteristics is developed that
approximately reproduces the results of the cluster
analysis. Finally, the community clusters are analyzed in
terms of the property tax burden on local residents. The
results indicate that residents of the two tourist-related
community clusters have lower property tax burdens than
residents in the non-tourist-related community clusters,
thus suggesting that the tourism development in those
communities reduces the property tax burden on local
landowners.
The third paper by Tim Tyrrell of the University of Rhode
Island presents the Tourism Development Simulation Model
(TDSM). Originally, this model was envisioned to provide a
way for community planners to make "ball park" estimates
of the impacts associated with different types of tourism
development. In the early phases of construction, however,
it became clear that impacts vary substantially with the
characteristics of the community and its responses to
development, as well as with the characteristics of the
development itself and the types of tourists it attracts.
A complex set of interactions make it difficult to specify
"average impacts" for different types of tourism
development in different types of communities.
Consequently, the model was transformed into a simulation
game in which the player is forced to consider the
consequences of development and the complex interactions
within the community that, together, determine the
ultimate impact of the development. The value of the game
is that the player begins to recognize a much broader
range of potential impacts than just jobs or income. For
example, the model forces the user to consider the need to
balance local government budgets and the potential
environmental and social impacts associated with the
tourism activity. The model is based on results of several
tourism research projects; hence the model offers a
realistic view of the issues that a community must address
in assessing the ultimate impacts associated with tourism
development.
One of the more difficult and expensive aspects of tourism
studies designed to measure the economic impact of tourism
is the need to determine the spending profile of tourists
who are attracted to a community by a new development.
This spending profile depends on several factors including
whether visitors are local or non-local residents, day
users or overnight visitors, and the type of lodging used
by overnight visitors.
Complex visitor surveys often have to be conducted to
construct the visitor spending profiles. The fourth paper
by Dan Stynes and Dennis Propst at Michigan State
University addresses this issue by constructing visitor
spending profiles for different types of visitors or
segments of tourists, such as day visitors, overnight
visitors who camp, and overnight visitors who use motels
for lodging. By constructing separate spending profiles
for the different segments of homogeneous visitors, the
spending profiles are more stable across space. Hence,
local planners can determine the types and numbers of
visitors that a tourism development will attract and then
use the spending profiles constructed from data collected
at other locations to determine the level of tourist
spending associated with the new development. This
approach significantly reduces the time and cost of
estimating the total tourist spending associated with a
new development.
Stynes and Propst also discuss how input-output models, in
conjunction with the total expenditures of tourists, can
be used to measure the indirect and induced effect
associated with the tourists' expenditures in the local
area. The last paper by Mario Teisl and Steve Reiling at
the University of Maine examines the impact that tourism
has on local governments' cost of providing local
government services. Like residents, tourists consume
local government services during the time they stay in the
community. These services include police and fire
protection, water and sewer services, and solid waste
disposal. Again, the amount and types of services demanded
vary with the type of tourist. For example, day visitors
demand a different mix and quantity of services than do
overnight visitors or seasonal residents. Their paper
presents the results of a study that estimated how tourism
affects the costs of providing nine different local
government services using data for 379 towns in Maine.
They conclude that tourism does have an impact of the
costs of delivering several types of the public services,
but the additional costs incurred to provide services to
tourists is quite small relative to the total costs of
providing local services.
All of these papers contribute to the stated objective to
develop, test, and refine procedures that enable local
officials to evaluate the potential benefits and costs of
tourism development. Although additional work is needed to
further assist local officials, the contributions
presented in this report should simplify the task of
evaluating the potential impacts that communities may
incur from tourism development.
ASSESSING CHANGES IN TOURISM IN THE NORTHEAST
Tommy L. Brown and Nancy A. Connelly
Department of Natural Resources
Cornell University
INTRODUCTION
Tourism is acknowledged to be among the leading industries
in each state in the Northeast. Yet tourism is so diverse
that it is often difficult to measure. Only at the site
level, such as at a park, resort, or for a specific
fishery, can researchers use survey methods to estimate
the economic impacts of tourism in a straightforward and
relatively uncomplicated way. As examples of such research
see Gitelson and Grafe (1990) and Connelly et al. (1986).
These methods also have been applied to special events
such as community festivals (Mitchell and Wall 1989).
For larger geographic entities, it becomes much more
difficult to accurately measure the economic impacts of
tourism. The basic measurement problem stems from the fact
that most tourist expenditures occur at retail
establishments such as restaurants, sporting goods stores,
marinas, golf courses, and amusement areas whose clientele
also include local residents. The owners of such
businesses usually have only a vague estimate of the
proportion of annual receipts that is derived from
tourists versus local residents. Indeed, this proportion
often changes from weekday to weekend and from season to
season.
In cases where secondary data are available on a monthly
basis, such as state employment statistics broken down by
counties and municipalities, one can derive reasonable
estimates of the proportion of employment and business
receipts that are from tourists versus local residents
(e.g., Connelly and Brown 1988; Brown and Connelly 1986).
Such estimates are derived intuitively from graphics and
assume a base level of trade from local residents. Sales
or employment above this base level are then attributed to
tourism. This type of analysis is useful at the state or
sub-state level, but the cooperation of the state commerce
or economic development agency is needed before the
monthly employment and sales data can be obtained.
Estimates and comparisons of tourism over time for a
broader region require not only the use of secondary data,
but of regional analysis methods to develop indices of
change. In an earlier analysis Brown (1980) used
shift-share analysis of lodging receipts to estimate
changes in the share of Northeastern tourism that each
state in the Northeast received from 1972 to 1977. He then
used location quotients to examine the relative dependence
of both Northeastern states and counties on tourism in
1977 and changes from 1972. This study presents an update
of the Brown (1980) study and compares data from 1977 to
1987.
METHODS
The comparative analysis of tourism uses lodging receipts
as the index of tourism. The rationale for this is first
that of all the Standard Industrial Codes (SIC) of
economic sectors pertaining to tourism, only the lodging
sector has an insignificant amount of trade from local
residents. Second, most tourism studies find lodging
receipts to be a significant portion of total trip
expenditures, and it is unusual to find localities where
tourism originating from day-use visits is an important
economic force.
The degree to which states gained or lost in their "share"
of tourism is analyzed by the use of shift-share analysis.
This traditional method has been described in detail by
Dunn (1960) and Ashby (1964). It was applied to tourism in
the Northeast by Brown (1980) and also by Cole (1980).
Brown (1982) also used shift-share analysis to examine
changes in tourism among major regions of the U.S.
Various measures using location quotients are used to
examine the relative dependence of county and state
economies on tourism. Described in detail by Izard (1976),
location quotients (LQs) can be developed for employment,
sales, income, or other economic measures. LQs compare the
proportion of a smaller area's economy (i.e., states and
counties, in this study) devoted to a particular sector
(i.e., lodging) with the analogous portion for a larger
geographic area (i.e., the Northeast) of which the smaller
area is a part. Thus, if a given county in the Northeast
had a lodging sales LQ of 1.0 with respect to the
Northeast, the proportion of the county's retail sales
derived from lodging would be the same as for the
Northeast as a whole. An analogous LQ of 2.0 would signify
that the proportion of the county's receipts derived from
lodging was twice that of the Northeast. Brown (1981)
found that heavily tourist-oriented counties in New York
had 1977 LQs that exceeded 11.0.
The database used for this study is the Census of Selected
Services, a publication of the Bureau of the Census, U.S.
Department of Commerce. Data from this quinquennial
publication for 1972, 1977, and 1987 (the most recent year
available) were used. Major problems occurred with the
1982 Census of Selected Services in the categorizing of
individual firms into standard industrial codes (SIC). As
a result, 1982 data are not comparable to other years and
were not used.
Lodging is classified under SIC codes as a service, but
nevertheless is a part of the broader retail economy. To
operationally define the "broader retail economy" of which
lodging is a component, we selected the total of retail
trade plus the service industries as reported by Bureau of
the Census. The Bureau of the Census definition of
services changed somewhat from 1977 to 1987, although this
change did not effect the lodging industry. For
comparability in showing broader economic changes from
1977 to 1987, we retained the 1977 definition of services.
We then converted 1977 data to 1987 constant dollars.
One major factor occurred in the Northeast between 1977
and 1987 that had a profound effect on some businesses
classified as lodging establishments:'the advent of casino
gambling in Atlantic City (Atlantic County), New Jersey.
As classified, lodging receipts in New Jersey increased
four-fold from $1.0 billion in 1977 to $4.0 billion in
1987 (1987 constant dollars). While much of this real
spending increase is attributable to tourism, most of the
receipts are not lodging receipts per se, and are not
comparable with lodging data elsewhere in the Northeast.
As a result, an approximation was made to eliminate the
portion of receipts attributable to casino gambling. This
was done by assigning the 68 lodging establishments in
Atlantic County with more than 25 rooms total lodging
receipts of the average hotel/motel in the rest of New
Jersey. Doing so reduced the estimate of 1987 lodging
receipts for the state from $4.0 billion to $1.15 billion.
All dollar values expressed in the Results section are in
constant 1987 dollars to facilitate comparison of real
growth. All data are derived from the Census of Selected
Services or Census of Retail Trade.
RESULTS
Receipts in the retail trade and services sector of the
economy grew 22.1% over the decade from 1977 to 1987, from
$428.6 billion in 1977 to $523.5 billion in 1987. During
the same period, lodging receipts increased by over twice
that rate (48.5%), from $7.1 billion to $10.5 billion
(Table 1). Lodging receipts grew in absolute terms in
every state in the Northeast (Table 1, column 3).
Figure 1 shows categories of net percentage change in
lodging receipts adjusted for inflation, at the county
level, comparing 1987 and 1977 data. Some of the counties
with the largest net percentage growth in lodging receipts
were urban and suburban counties. These include counties
containing the cities of Boston and Springfield, NU;
Stamford and Bridgeport, CT; Rochester and Nassau County
(western Long Island), NY; and some Philadelphia-area
counties. A number of rural counties also posted gains of
at least 50% in lodging receipts between 1977 and 1987.
Many of these counties, though rural in character, are
within a short drive of major metropolitan areas. Notable
exceptions occur in northern and central Vermont, in
northeastern and central Pennsylvania, and in southern
West Virginia.
Many other counties in the Northeast experienced gains in
lodging receipts that fell within the second category in
Figure 1 of 11% to 49%. Such counties were found in each
state and were generally well spread throughout the
region.
Counties for which data are available that declined
notably in lodging receipts between 1977 and 1987 were
generally those that either are not well developed with
tourism attractions, and/or are at a locational
disadvantage with respect to distance from population
centers. Examples of such declines include much of
noncoastal Maine, parts of the Adirondacks and far
northern New York, and the Southern Tier of New York and
the Northern Tier of Pennsylvania.
Shift-Share Analysis
Traditional shift-share analysis breaks the total economic
change over a given period into three components: regional
growth, industrial mix, and state shares. These are shown
in columns 2-4 of Table 2; the sum of these columns equals
the total change (column 1). The first component, regional
growth, indicates the amount of growth that would have
occurred had lodging grown at only the rate of the retail
service economy in the Northeast. A comparison of columns
1 and 2 shows that West Virginia was the only state for
which lodging grew less rapidly (only slightly so) than
the retail-service economy in the Northeast. The second
component of total change, the industrial mix (Table 2,
column 3), reflects the performance of the lodging sector
in comparison to the retail-service economy. It is
calculated by multiplying the 1977 lodging
Figure 1 Net percentage change in lodging receipts for
counties in the Northeast, 1977-1987 (Vis. 1)
Table 1. Lodging receipts for the Northeastern states,
1977 to 1987
(Vis. 2)
receipts for each state by the difference between the
percentage growth in the lodging sector in the Northeast
(48.5%) and the percentage growth in lodging receipts in
the Northeast (22. 1%). Because lodging growth
outperformed the overall retail-service sectors in the
Northeast by 26.4%, the industrial mix for each state is
positive and equal to 26.4% of its 1977 lodging receipts.
The third component, the state shares, measures the degree
to which each state increased or decreased its share of
regional lodging receipts. It is calculated by multiplying
each state's 1977 lodging receipts by the difference in
percentage change of the state's increase and the region's
increase in lodging receipts from 1977 to 1987. From Table
2, column 4, we see that some major changes occurred in
the state shares of lodging receipts over this period.
Massachusetts increased its share by nearly $350 million,
while Pennsylvania's share decreased by a similar amount,
$334 million. Other states with sizable increases (over
$50 million) in their share include Connecticut, Maryland,
New Jersey, and the District of Columbia. States with
comparable decreases in their share of lodging receipts
include New York and West Virginia; Maine fell just short
of a $50 million loss in its share.
The net relative change, column 5 of Table 2, is the
amount of change over and above the regional growth
(Column 1 minus Column 2). (It is also the sum of changes
related to the industrial mix and state shares, columns 2
and 3). As pointed out above, West Virginia is the only
state with less lodging growth than in the overall
retail-service economy, and therefore a negative net
relative change.
Columns 6 through 8 provide some of the parameters at the
top of Table 2 in percentage format. Column 6 shows that
lodging receipts in Massachusetts and Connecticut roughly
doubled from 1977 to 1987, and that all states increased
their lodging receipts by at least 20%. Column 7
percentages the total industrial shift among the states.
Of all the growth in lodging receipts over and above the
retail-service economy growth rate of 22.1%, 34.1%
occurred in New York, and over half occurred in the two
economically large states of New York and Pennsylvania.
Finally, Column 8 percentages the positive and negative
shifts in the state shares. Half of the positive shifts in
the state shares occurred in Massachusetts, while
Pennsylvania and New York combined for over 75% of the
losses in state shares.
Location Quotient Analysis
A comparison of 1987 and 1977 location quotients (LQs) for
lodging receipts for the Northeastern states is shown in
Table 3. In 1987, the retail-service economy of Vermont
was over three times as dependent on lodging as the
Northeast as a whole. West Virginia and Maine had LQs of
2.55 and 2.00, respectively. Connecticut, Rhode Island,
Maryland, and New Jersey were relatively less dependent on
tourism, with LQs of approximately 0.8.
Table 2. Shift-share analysis for lodging receipts for the
Northeastern states, 1977 to 1987.
(Vis. 3)
The 1987/1977 lodging sales LQ provides an index of the
change in the relative dependence of the economy on
tourism over this ten-year period. It should be pointed
out that for large entities such as states, we would not
expect retail or service LQs to deviate substantially over
a ten-year period except for cases of emerging industries
or unusual situations. West Virginia had the highest
1987/1977 lodging LQ (1.20).
Categories of 1987 lodging sales LQs for counties for
which census data were not suppressed due to a small
number of firms are shown in Figure 2. Approximately 25
counties in the northeast had lodging sales LQs of greater
than 5.0. For the most part, these include the most
popular tourism destination areas in the Northeast. Among
them are areas of coastal Maine that include Acadia
National Park; parts of the White Mountains of New
Hampshire; parts of the Green Mountains and some of the
renowned ski areas of northern Vermont; Cape Cod,
Nantucket, and Martha's Vineyard; the Adirondacks, some
parts of the Catskills, and the Thousand Islands regions
of New York; the Maryland Eastern Shore, including Ocean
City and the Assateague Island National Seashore; the
Pocono Mountains of Pennsylvania; and portions of the Blue
Ridge Mountains in West Virginia.
Lodging location quotients with respect to the retail-
service economy in the Northeast, 1987(Vis. 4)
Table 3. Location quotients of states for lodging receipts
with respect to the retail service economy of the
Northeast (as defined in the 1977 Census of Service
Industries).
(Vis. 5)
Approximately 45 additional counties fell within the
second category of having LQs of between 2.00 and 5.00.
Some of these counties represent excellent tourist
destination areas that are located where the economy is
more diversified than counties in the first category.
Examples include all of southern and central Vermont;
Niagara Falls, Chautauqua Lake, and the Lake Champlain
Valley in New York; the Berkshires in Massachusetts; the
Amish country in Pennsylvania; Newport and coastal Rhode
Island; the Delaware shore; and Atlantic City, New Jersey.
Counties with low to average LQs fall into two general
categories. The first is large urban areas. These areas
have very diversified economies, and although they may
receive very large numbers of tourists (e.g., New York
City, Boston, Philadelphia), their retail and service
economies as a whole are not highly dependent on tourism.
This is also true of many suburban counties that are
proximate to large urban centers. The second category
would include counties that are heavily agricultural or
forested, often located great distances from urban centers
or in areas that are not connected by interstate highways,
and where major tourism attractions have not been
developed. Particular examples include several counties in
western New York and southern West Virginia.
SUMMARY
The combination of shift-share analysis and location
quotients (LQs) provide an excellent means of assessing
the importance of tourism in the various states and
counties in the Northeast, and whether the dependence of
these areas on tourism is increasing or decreasing. These
methods are most useful for making preliminary assessments
and comparisons. More in-depth analyses would be necessary
to develop an understanding of the factors that are
related to increased or decreased tourism dependence.
It should be pointed out that having a high lodging LQ,
which implies a strong dependence of an area on tourism,
is of itself not necessarily a positive statistic. A high
lodging LQ also implies that some other areas of the
economy have low LQs and are thus underdeveloped. LQs are
valuable in demonstrating the relative dependence of an
area on tourism, and in combination with actual data on
receipts from lodging and other important tourism sectors
(e.g., eating and drinking places, miscellaneous amusement
and recreation firms, sporting goods firms) can provide
information on whether tourism is growing or declining.
Shift-share analysis can be applied to any level of
geographic entity, but is probably most useful at the
state or large regional level. With several hundred
counties in the Northeast, each would have only a tiny
share of the region's tourism receipts; hence a
shift-share analysis would be relatively meaningless. An
examination of net percentage change over time (i.e.,
Figure 1) would be more meaningful. At the state level,
where there are a limited number of units, it is much more
reasonable to think of shares and whether each state has
kept its "share" of Northeast tourism dollars, or whether
that share has increased or decreased over time.
These methods of regional analysis are useful to
practitioners as well as to researchers. Because the
methods are rather intuitive, they can be used
successfully with planners and local tourism advisory
committees to help them evaluate the stance of their
county or area with respect to tourism.
REFERENCES
Ashby, L.D. 1964. The geographical redistribution of
employment: an examination of the elements of change.
Survey of Current Business. U.S. Department of Commerce,
Office of Business Economics, October. pp. 13-20.
Brown, T.L. 1982. The competitive position of tourism
in the Northeast, 1972-1977. Journal of the Northeastern
Agricultural Economics Council 11(l): 21-24.
-----. 1981. Assessing changes in tourism in New York
State. Cornell University Agricultural Experiment Station
SEARCH - AGRICULTURE publication. 10 pp.
-----. 1980. Assessing changes in the importance of
tourism in the Northeast. Proceedings, 1980 National
Outdoor Recreation Trends Symposium. U.S. Forest Service
General Technical Report NE-57, Volume 2, pp. 151-160.
Brown, T.L., and N.A. Connelly. 1984. Tourism in the
Adirondack region of New York. Cornell University, Dept.
of Natural Resources Research and Extension Series No. 21.
68 pp.
-----. 1986. Tourism and employment in the Adirondack
Park. Annals of Tourism Research 13:481-489.
Cole, G.L. 1980. Changes in recreation oriented travel
in the Northeast between 1972 and 1977.
Proceedings,1980 National Outdoor Recreation Trends
Symposium. U.S. Forest Service General Technical Report
NE-57, Volume 2, pp. 139-146.
Connelly, N.A., and T.L. Brown. 1988. The impact of
tourism on employment in New York's coastal areas. Dept.
of Nat. Res. Research and Extension Series No. 32. 58 pp.
Connelly, N.A., T.L. Brown, and D.J. Allee. 1986.
Assessing the economic impact of state parks located near
urban areas in New York and the effect of these impacts on
the budget allocation process. Human Dimensions Research
Unit Public. 86-6. Cornell Univ. 62 pp.
Dunn Jr., E.S. 1960. A statistical and analytical
technique for regional analysis. Papers and Proceedings of
the Regional Science Association 6:97-112.
Gitelson, R.J., and A. Grafe. 1990. Economic impacts
associated with whitewater boating on the Upper
Youghiogheny River. Proceedings, 1990 Northeastern
Recreation Research Symposium. U.S. Forest Service,
Northeastern Forest Experiment Station, General Technical
Report NE-145, pp. 65-70.
Izard, W. 1976. Methods of regional analysis. Cambridge:
The M.I.T. Press. 784 pp.
Mitchell, C.J.A., and G. Wall. 1989. The arts and
employment: a case study of the Stratford festival. Growth
and Change 20(4): 31-40.
CLASSIFICATION OF RECREATION AND TOURISM COMMUNITIES
USING CLUSTER ANALYSIS AND DATA MANAGEMENT TECHNIQUES
Malcolm L. Bevins, University of Vermont
Rodney R. Zwick, Lyndon State College
INTRODUCTION
Policy makers, planners, and the public commonly
characterize communities in broad categories when
initiating investment, planning, and regulatory decisions.
Often categorization schemes (e.g., rural vs. urban) are
based on planners' or policy makers' subjective
perceptions of communities (Kinds of Communities Working
Group 1986). In some states, policy formulators have
dichotomized communities as"growing or decaying,"
"agricultural or industrial," and "old or new" as a
consequence of the community's traditional economic base.
Because the traditional economic base is declining, many
rural communities have embraced tourism and recreation
development as strategies to improve their economic
well-being (Liu and Var 1987; Allen et al. 1988). These
changes in rural development strategies and economic base
of communities suggest that previous categorization
schemes may not be viable. New classifications, therefore,
should be explored that differentiate rural communities
based on tourism and recreation-related attributes, and
attributes related to economic growth. Such attributes
(variables) may be useful in describing and comparing
communities and for determining rural policy initiatives.
Classification schemes (typologies) should be empirically
derived if they are to be useful for analyzing and
comparing communities, distributing resources (e.g., state
and federal revenues), and establishing public policy
(e.g., location of a state park). Empirically derived
classifications (typologies) provide a means by which we
can better predict performance and behavior in the "real
world" (Bealer 1987). Some rural community typologies have
been developed from socioeconomic data obtained from
secondary sources, such as census and state generated data
(Humphrey et al. 1987; KOC 1986). Many, however, continue
to be based on subjective judgements. Few classification
schemes of rural communities have focused on physical
attractiveness, growth, or degree of tourism development,
except as variables related to resident attitudes or
reactions to tourism expansion (Ahmed 1986; Liu and Var
1987; Pizam 1980; Allen et. al. 1988).
Once classification schemes have been constructed, their
usefulness in describing "real world" aspects of rural
areas should be tested. Too often classifications are
developed that are not linked to public policy. Such
classifications do not help the practitioner comprehend
and work with the socio-political aspects of communities
(Feldman 1987). The test of usefulness of the
classification scheme then is the linkage between the
scheme and "real world" events.
The purpose of this study was to develop an empirical
classification (typology) of rural communities based on
attributes related to tourism, growth, and recreation
amenities, and to determine if the developed typology can
distinguish between communities on a variable of
socio-political concern, (i.e., property taxes). Two
questions guided the research:
1. Can a typology of communities be constructed from
secondary data source attributes related to recreation,
tourism development, and growth?
2. Does the constructed typology distinguish between
community tax burdens (i.e., index of median property tax
related to an index of median income)?
The changing economic base of rural areas requires a
rethinking of our classification of rural communities. Our
communities have become too complex to use a single type
descriptor for analytical purposes. One tool, cluster
analysis, lets us simultaneously analyze several variables
to generate a more useful, mutually exclusive community
label or "typology. Such classifications may then help
organize our thinking about communities by systematizing
and reducing data into discrete categories. The
development of meaningful public policy to wisely address
community needs requires the use of such techniques.
CLUSTER ANALYSIS APPROACH
Our approach in developing a constructed typology
(classification scheme) involved using only secondary data
sources. Data for this study were obtained from the Center
for Rural Studies at the University of Vermont, data
collected by the authors in previous studies, secondary
source data from the Agency of Development and Community
Affairs, and from other Vermont state agencies. The
attributes selected consisted of factors that would
differentiate among communities in terms of growth,
tourism development, and recreation amenities, while
satisfying criteria for inclusion in a clustering
solution. Humphrey et al. (1987:52) suggest that variables
used as input to clustering solutions should
* be limited in number,
* be relatively uncorrelated with each other (i.e.,
unique),
* have comparable means and standard deviations.
Ten attributes were selected to differentiate among
communities. The selection of attributes was based on (1)
the attribute's relation to community growth, recreation
attractiveness, and development of tourism; (2)
uniqueness; and (3) availability of secondary data source
materials for all 246 Vermont communities (See Figure 1).
Ten independent variables describing Vermont community
conditions in 1985 and 1986 were assessed:
1. Population change between 1980 and 1986.
2. Total value of all property per capita.
3. Total value of all property per acre.
4. Value of all commercial property.
5. Value of all agricultural property.
6. Value of all properties used as second homes.
7. Population density.
8. Water acreage (lake or major river).
9. Degree to which residents find work within the
hometown.
10. Commercial lodging capacity.
These attributes were chosen as surrogates for (1) local
growth communities, (2) area "class" (two attributes), (3)
commercial services, (4) open space land use, (5) area
desirability, (6) area congestion, (7) water recreation
potential, (8) overall business activity, and (9) traveler
attraction potential. In the aggregate they reflect the
city and town attributes that are in the forefront of
policy formulation facing Vermont today.
Because city and town data are highly skewed on many of
the attributes considered in this study, the data had to
be standardized to meet the criteria of comparable means
and standard deviations. For example towns differed
considerably on the value of their commercial property,
thus a process had to be found to transform data to lessen
the natural variation and highly skewed differences that
exist. Likewise, towns vary on one attribute more than
another (e.g., residential population density vs.
percentage of population working outside the town), thus
population density would be more highly weighted in
cluster analysis. Using a process outlined by the Kinds of
Communities (KOC) Working Group (1986) for examining
clusters of Massachusetts cities and towns, the attributes
for towns in this study were converted to rank orders and
the ranks were subjected to cluster analysis. This
transformation simultaneously solves the problems of
"outlier" data and natural variability while allowing for
simplicity of understanding (KOC 1986). Attributes were
ranked so that the highest value had a rank of "1"; towns
or cities having a higher median rank on an attribute,
therefore, have more of that attribute.
Community attributes were then submitted to cluster
analysis. Cluster analysis is a descriptive data reduction
procedure that combines what appear to be relatively
unrelated cases (e.g., towns and cities) into homogeneous
subsets or categories (i.e., towns and/or cities having
commonality). Used in a wide variety of disciplines,
cluster analysis is essentially a mapping procedure(KOC
1986).The output from the mapping allows one to identify
and label the cluster, and understand how cases group
together. Various types of cluster analytic procedures are
available, but generally they can be divided between
hierarchical and nonhierarchical techniques. A
nonhierarchical procedure was selected for this study
because of the relatively large number of cases. The
nonhierarchical technique uses multivariate profiles to
sort the cases
Reduction in the number of variables.(Vis. 6)
into k-clusters based on "seed" points (Goldsmith 1987).
The initial seed points are automatically defined by the
cluster program, and each case is assigned to an initial
seed. A centroid for the clusters is computed and used as
a seed point for the next iteration on which the cases are
sorted (Goldsmith 1987). Programs such as SAS's FASTCLUS
program allow the user to specify the number of clusters
they wish to explore. While the selection of the number of
clusters in many cases is arbitrary in nature, the
researcher can compare different numbers of clusters from
successive cluster runs to determine which number of
clusters most clearly and simply describes the
similarities between towns. Similarity in nonhierarchical
cluster analysis is defined as the similarity of a town or
city to the cluster center; the average Euclidean distance
within a cluster, then, is an indication of how well the
cluster accounts for the towns arrayed in the cluster (KOC
1986). The smaller the average distance, the tighter the
cluster and the more alike are the towns within that
cluster, and the easier it becomes to interpret the
cluster solution.
Clusters that emerge from the cluster analytic procedure
can be thought of as constructed types possessing some
common characteristics. This commonality (or homogeneity)
can then be used for comparison or for distinguishing
between types on other variables in the "real" world. For
example if community types are similar with respect to
growth attributes they may be equally homogeneous in
characteristics such as tax revenue per capita.
Table 1. Criteria used to classify towns (Boolean
Classification Method).
(Vis. 7)
Two dependent variables subsequently were analyzed within
the seven cluster frame: (1) tax burden (i.e.,
relationship between income earned and property taxes
paid), and (2) amount of local property owned by community
residents. The graphic representation of these two
relationships is shown in Figures 2 and 3.
Clearly, cluster analysis worked well, as a research tool,
in objectively classifying communities. It worked equally
as well in applying such a classification scheme to
differentiate the impact of development.
Town Type Determined Using Cluster Method(Vis. 8)
Classified by Town Type 1986(Vis. 9)
Table 2. Property values in Vermont by town types 1970 and
1989 (Boolean Classification Method)
(Vis. 10)
DATA MANAGEMENT APPROACH
We are living in a state of rapid change. Communities may
shift from one type to another type as development occurs.
Therefore, a model, easier and faster to use than the
cluster analysis model, is needed to monitor change at a
relatively low cost. The cluster analysis model described
above is an effective tool to monitor change, but to
reevaluate ten variables on an annual basis is extremely
time consuming and expensive.
It was apparent from our cluster analysis model that
property values (disaggregated) was the most important
variable explaining community classification. The second
most important variable was population density. Armed with
this knowledge, we set forth to develop a simple system
that could be used to classify communities over time. Our
goal was to develop a system that could be handled with a
desktop personal computer, thereby opening use of this
analytical tool to a much larger group.
After investigating many spreadsheet programs, we decided
to use Computer Associates' SuperCalc. We found the data
management program (within the SuperCalc spreadsheet) to
be very effective and relatively easy to use. Nearly all
of our basic data were already in a spreadsheet format and
could be used directly or imported into SuperCalc.
SuperCalc's data management function was used to perform a
boolean search of the full data set, to extract a list of
communities that met a given set of criteria values. The
criteria were different for each community type, so that a
community would fall into one class only (mutually
exclusive classification).
Table 1 shows the criteria employed to classify Vermont
communities using boolean search techniques. A data set
for 1989 was tested using boolean procedures. Test results
indicated remarkably similar findings to the 1985 test
using cluster analysis. However, it became apparent that
more accurate type names should be used. All communities
that were primarily residential should carry the name
residential--thus the old name "commercial high density"
was re-named "residential commercial center." The old
"rural bedroom community" was re-named "residential
rural." The old "waterfront bedroom" community became
"residential limited commercial."
It was decided that all communities that were primarily
recreational should carry that name in its title. Thus,
"major resort" communities were re-named "recreational
commercial centers." The old "recreational" communities
were re-named "recreational non-commercial." Lastly,
"rural woodland" communities were renamed "lowest
population density" communities (several that fell into
this class had very little forested acreage).
Town Type Determined Using Boolean Search(Vis. 11)
Residential property tax burden was calculated for 1989
using the clusters identified through the boolean search
procedure (Figure 4). A comparison of Figures 2 and 4
clearly shows that the simpler boolean search procedure
yielded essentially the same results as the more complex
cluster analysis.
The principal advantage that the boolean technique has
over cluster analysis is the ability to analyze community
change over a period of years. Property values in Vermont
towns in 1970 were compared with property values in 1989,
for each community type (Table 2).
This analysis shows that residential property values in
recreational commercial centers increased from 12 percent
of the state residential total value in 1970 to 23 percent
of the state total in 1989. The reverse situation was
found in residential commercial centers where the portion
of the total state residential value dropped from 43
percent to 33 percent in this 19-year period. This was an
important finding, as it clearly showed movement of
permanent residences from the higher density cities to the
lower density rural recreation centers.
A similar gain and loss situation was found in analyzing
changes in commercial property values. While the urban
centers had 58 percent of the state total commercial value
in 1970, they had only 51 percent in 1989. In 1970, the
rural recreation centers had 20 percent of the state total
commercial value. In 1989, this increased to 39 percent.
The impact of such shifts on local government is
clear--there was a loss of tax base in urban residential
areas and gain in tax base in rural commercial recreation
areas. Another favorable impact on the rural commercial
recreation communities was the increased diversity in the
tax base, important in times of economic uncertainty. The
non-commercial rural recreation centers failed to share in
the good fortune of the commercial recreation centers.
There was very little growth in these communities between
1970 and 1989. Their share of second homes values dropped
from 35 to 21 percent of the state total.
IMPLICATIONS
The two approaches tested in this research resulted in the
creation of a scientifically useful method of classifying
communities for further analysis. The two types of
recreation/tourism communities that the models identified
are very different and call for very different policies to
guide growth and development.
This research does not provide the answers to complex
questions relating to growth and development, but it does
provide a reasonable frame for balanced dialogue on such
issues. Such dialogue is necessary if we are to develop
sound long-range policies governing community growth in
Vermont and the nation.
REFERENCES
Ahmed, S. 1986. Understanding residents' reaction to
tourism marketing strategies. Journal of Travel Research
25:13-18.
Allen, L.R., P.T. Long, R.R. Perdue, and S. Kieselbach.
1988. The impact of tourism development on residents'
perceptions of community life. Journal of Travel Research
27:16-21.
Bealer, R.C. 1987. Metatheory: Unlocking the mysterious
door to topologies In Rural People and Places: A Symposium
on Typologies, ed. A.E. Luloff, pp. 2-14. University Park,
PA: Northeast Regional Center for Rural Development.
Feldman, R.J. 1987. Condensation for extension community
development staff: Uses of and concerns about community
typology construction. In Rural People and Places: A
Symposium on Typologies, ed. A.E. Luloff. pp. 164-169.
University Park, PA: Northeast Regional Center for Rural
Development.
Goldsmith, H. 1987. Developing a strategy for cataloging
the residential environments of nonmetropolitan rural
areas. In Rural People and Places: A Symposium on
Typologies, ed. A.E. Luloff, pp. 103-128. University Park,
PA: Northeast Regional Center for Rural Development.
Humphrey, A.B., J.S. Buechner, and W.F. Velicer. 1987.
Differentiating geographic areas by socioeconomic
characteristics. The Northeast Journal of Business and
Economics 13:47-63.
Kinds of Communities Working Group. 1986. Technical manual
for clustering Massachusetts cities and towns. Boston:
Office of Executive Planning, Massachusetts Department of
Education.
Liu, J., and T. Var. 1987. Resident perceptions of the
environmental impact of tourism. Annals of Tourism
Research 14:17-37.
Pizam, A. 1980. Evaluating social impacts of tourism:
Case of Cape Cod, Massachusetts. Tourism Recreation
Research (December): 3-7.
TOURISM DEVELOPMENT SIMULATION MODEL: THE GAME
Timothy J. Tyrrell
Department of Resource Economics
University of Rhode Island
INTRODUCTION
Tourism industry development has its greatest impacts at
the community level. These impacts reach far beyond jobs,
wages, and revenues. Social and environmental consequences
such as crime, congestion, and water pollution also can
have serious impacts on the quality of life of residents
as well as on the quality of vacations for tourists.
Community leaders and planners have indicated a need for
information about these impacts.
Since 1980, researchers at a number of universities in the
Northeast and elsewhere have worked toward developing a
framework and a database with which any locality could
make "ball-park" estimates of the impacts of typical types
of tourism development. A major research finding has been
that it is not easy to define typical communities, typical
developments, or average impacts. Another finding has been
that the complex interactions between residents, tourists,
industry, and government make it difficult to justify a
simple framework for analysis. On the other hand it is
difficult to present the correct framework (assuming one
exists) in a simple manner. In order to test our
preliminary research hypotheses, a tourism-planning
computer game has been developed which pits the player
against the development pressures of a hypothetical small
New England community.
In addition to meeting our objective of illustrating our
long-term research goals in a palatable format the game
has been used successfully in the classroom and, in fact,
has taken on a project status of its own.
Project Background: A Game?
The Tourism Development Simulation Model (TDSM) is a
computer program developed in conjunction with a series of
projects on the impacts of tourism development. The
emphasis in the earliest project was on the development of
a methodology that could be used by individual
municipalities and state agencies to estimate and project
impacts caused by alternative tourism product mixes at the
local level. A follow-up project focused on the fiscal
impact portion of the model.
In the early stages of design and development of this
model, it became readily apparent that the scope of the
work was very broad-too broad to produce the quick and
comprehensive results users wanted. It was, in fact,
difficult to describe the goals of the project or its
potential value to targeted users. It was decided,
therefore, that a simplified version of the model could
provide the basis of a computer game, and that this game
concerning tourism development in a hypothetical community
could provide an alternative means of describing the
usefulness of the major project objectives.
The production and distribution of the game has been so
successful that it has taken on a project status of its
own. Priced at cost, it was initially distributed as an
amusing "think piece" for researchers and planners, and as
an palatable introduction to related research projects.
The game is loosely based on the impact assessment
framework developed in a study of the impacts of tourism
on thirteen coastal communities in the Northeast (Tyrrell
et al. 1985). A typical small New England community was
characterized using some empirical results and some "best
guess" approximations to relationships between major
variables. The relationships were then refined to reflect
the "reality" exhibited in recent community case studies
(NE-137 Block 1985).
In addition to providing the intended introduction to
long-term project goals, the processes of designing and
constructing the game and using it as a community behavior
simulator have taught us much about the implications of
"reality" and "reasonableness" on mathematical
representations of community behavior. These lessons have
had a considerable impact on our primary goal of
developing a tourism development simulation model.
Designed to illustrate the principles of community
budgeting and planning in the context of tourism industry
development, the game has been used as an educational tool
for students of tourism planning at the graduate and
undergraduate level, and non-degree participants in
leadership development programs in both the U.S. and
Canada. For all audience types, the educational vehicle
provided by the computer game is accepted
enthusiastically. As with others (Loukissas 1983), we have
found this approach to have definite advantages over
traditional seminar presentations.
This paper describes the details of the game's design and
describes the importance of the "reasonableness" criterion
used in the development of the underlying model. It also
attempts to persuade the reader that a computer game,
based on a "best-guess" model, can provide a useful
research and educational tool.
The Tourism Industry Focus
The tourism industry has many dimensions of concern to
community researchers. It is recognized to include a wide
range of business types whose activities impact a broad
cross-section of the host community (Murphy 1985).
Tourists are of many types as are the residents of the
affected community. In conducting community-based tourism
research it is therefore important to identify and measure
impact generators and recipients by their types so that
policies can address specific problems and needs. In the
model underlying this game, however, all tourist-related
businesses are represented by an industry average, and all
non-tourist-related businesses are represented by another
average. In a similar way, residents are grouped and
tourists are grouped. In contrast to the research
approach, the emphasis here is on the general dynamics of
the tourism industry as it relates to other societal
components in a typical community setting. A flow chart of
the major components is provided in Figure 1.
The community is viewed as a circle whose outer ring is
the natural resource base. Within that is the social
system which provides the context for all community
activities. Finally, in the center are the four major
components of the model: the tourism industry, "other"
industry, local government, and the residents. Outside the
community are four factors that influence and are
influenced by community activities: exports, non-resident
capital, non-resident laborers, and tourists. The arrows
that connect the components and outside factors illustrate
the major linkages assumed by the model that underlies the
game.
Specific adjustments in behavior of tourism businesses,
"other" businesses, residents, and tourists are determined
by equations relating causal factors to specific actions.
This follows the basic econometric approach of modeling
market systems (Pindyck and Rubinfeld 1976). By
sacrificing the behavioral differences of different types
of producers and consumers, the simulation-game goes
beyond equations that are typically estimated in research
projects and speculates about a very complex set of
important interactions within the community. Of particular
interest were the social and environmental impacts of
development that can change the "quality of life" for
residents, the "bottom line' for businesses, and the
"quality of the vacation" for tourists. Such impacts are
well documented in the literature (Mathieson and
Wall 1982). Finally, the game recognizes the political
nature of the decision-making process, the budget-
balancing skills required by individual community
officers, and the unpredictable disturbances to the
planning process caused by unforeseen events.
A Community Setting
In order to simulate tourism development in the most
realistic way possible, the model centers on the decisions
made by an administrative officer of a typical small town
in New England in the late 1970's. At the beginning of a
game there are 20,000 residents in 10,000 households, and
the population is growing at 1% per year. Half the
population is considered to be in the labor force. 7500 of
these are employed by the tourism industry, 2250 are
employed in the "other" industry of the local economy, and
the unemployment rate is initially 2.5%. Wage rates are $
10,000 per year in the tourism industry and $20,000 per
year in the other industry. Although average firms in both
tourism and "other" industries begin the simulation
breaking even, profits earned in the subsequent years are
distributed to the residents of the community as
additional income. Consequently disposable household
income averages $11,255 per year in year 1.
The natural resource base of the community is assumed to
be a major recreation area, thus the draw for tourists and
a major factor in the quality of life in the community.
This resource might be a park or beach that is valued by
the community and attractive enough to draw tourists to
town, but fragile in the sense that increased use, in the
absence of preventative investment, will result in a
reduction in its aesthetic qualities. A number of works
report on such impacts on recreational resources (Young
1973).
In the initial year of the simulation 200,000 tourists
visit the community, staying an average of 7 days each,
and visiting the natural resource recreation site on every
day of their stay. Residents each make 10 visits per year
to the recreation site. Residents and tourists pay the
same price per visit to the recreational site, which is
managed by the industry, however the tourists pay a
visitor's fee to the town.
Figure 1 Major components of the game community(Vis. 12)
The town administrator has the authority to invest these
and other town revenues in the protection/enhancement of
the natural resource base.
Year-to-year variation in weather is one example of the
unpredictable nature of the community environment. The
model assumes that changes in the average seasonal
temperature from its norm of 75 degrees F will influence
use of the recreational resource by residents as well as
tourists. Each year of the simulation, average seasonal
temperature is assigned a random value from 70 to 80
degrees. When temperatures drop below 75 degrees, visits
to the site decrease, and when temperatures exceed 75,
they increase.
As residential and tourist populations increase they are
assumed to impact the social and environmental
characteristics of the community as well as its economy.
These are manifested in crime and congestion rates and
water quality measures known to the community, reported as
social and environmental indexes and as headlines in the
local press. Thus, the quality of life in the community is
assumed to be reflected in indexes for three basic
community qualities: economic, social, and environmental.
The environmental index is assumed to reflect the quality
of the recreational resource (Wennergren and Fullerton
1972). The geometric mean of the three indexes, weighted
by the population of the community, is used as an overall
measure of community well-being. Such quality of life
indexes have been pursued for a number of years, with
varying degrees of success. (Booz, Allen, Inc. 1973). This
well-being measure might also be thought of an indicator
of the economist's social welfare function (Henderson and
Quandt 1958).
In addition to the raw indexes, the game player is
presented news headlines that correspond to the severity
of the index above or below their initial, equilibrium
position. For example, when the social index drops 30%
below its initial value the player is informed "Crime and
Congestion at Record Levels. Police Frustrated." And when
the environmental index rises 30% above its initial value
the player is informed "National Task Force Adopts Local
Environmental Plan as Model."
The Game Player
The game player assumes the role of an unusually potent
town administrator who must guide development using
tourism promotion and three other types of expenditure
programs: "other" industrial development programs, social
programs, and environmental projects. These expenditure
categories are assumed to include all major expenditure
items in a town's budget except for administration, which
is ignored by the model. In order to balance the town's
budget the player must adjust tax rates for residential
and business property and decide upon daily tourist fees.
The player's objective is to achieve balanced economic,
social, and environmental development as measured by their
respective indexes. At the end of each year of office the
player is presented with quantitative and qualitative
assessments of performance, and given the opportunity to
change any of the seven fiscal controls. The game ends
after 10 years of service or when the player quits or is
fired, which ever comes first.
THE TOURISM DEVELOPMENT MODEL
Model Structure
Some 40 equations containing more than fifty parameters
characterize the components of community activity and
development in the small hypothetical town. Most of these
(19 equations) characterize the tourism industry and
"other" industry behavior. These include basic supply and
demand relationships for the tourism product and the
"other" product produced in the town, as well as derived
demands for labor and supplies and business relocation.
The multi-market model framework is common to macro
modeling efforts (Pindyck and Rubinfeld 1976).
Supply and demand equations are logarithmic and contain a
lagged dependent variable. This formulation characterizes
behavior as "partially" adjusting toward some distant
equilibrium value. The amount of the adjustment declines
each year following the initial change in the independent
variable. The rate of adjustment is determined by the size
of the lag coefficient. Such models are frequent in
econometric research studies since their coefficients are
directly interpretable, and the implied dynamics is
intuitively appealing.
The logarithmic specification implies that the coefficient
of each explanatory variable is a short-run elasticity
(the percentage change in the dependent variable in the
first year resulting from a 1% change in the independent
variable). Long-run elasticities (the ultimate percentage
change in the dependent variable after a 1% change in the
independent variable) are computed as the ratio of the
short-run elasticity to (1 - lag coefficient). The
complete set of behavioral equations and their
coefficients are listed in Appendix A.
Random Events
As described above, weather is an important cause of
random variations in visits to the recreation site and has
important implications for the tourism industry. Part of
the change in export sales by the "other" industry in the
community is also generated as a random event. The model
assumes the percentage deviation of current from previous
export sales is randomly selected in the range from -5% to
+5%.
Early in the administration of the town it is unlikely
that any unforeseen event will alter the outcome of
control decisions. By the fifth year of the
administration, however, there is a 60% probability that
some such unforeseen event has occurred. These events are
announced as "Important News Flash(es)."
Five types of random events are built into the model:
F: FIRE DESTROYS ___ % OF THE INDUSTRIAL DISTRICT,
G: GAS SHORTAGE THREATENS TO REDUCE TOURIST VISITS
BY ___%,
0: OIL TANKER SPILLS - REDUCES WATER QUALITY INDEX
BY ___%,
U: UNEXPECTED MEDIA ATTENTION PROMISES TO INCREASE
TOURIST VISITS BY ___% and
B: BAD DEBTS LEFT BY PREVIOUS PLANNER UNCOVERED
AMOUNTING TO $________
The percentages and the dollar amount indicated by the
News Flashes are also determined randomly. The percentages
range from 25 to 100%, and the dollar amount of the
deficit can range from 0 to $250,000.
The choice of the five events was meant to correspond to
five of the major uncertainties facing tourism community
administrators: unexpected catastrophes in tourism and the
"other" industry, a direct threat to tourism
attractiveness because of pollution, unforeseen (and
unplanned) rapid growth in tourism, and a personal threat
to the job security of the administrator.
The player is given the opportunity to respond to these
events before the impact is levied on the social,
environmental, and economic indexes. The successful
planner will take corrective actions given this
opportunity.
The probability of a subsequent event is low for the
immediate future after an event, but increases as the
years pass. This may be an unrealistic representation of
the planner who spends considerable time responding to
crises. The game events may be repeated or not and the
severity of the events, as measured by the percentage or
dollar amount, will vary from occurrence to occurrence and
game to game. It is very unlikely that any two games will
be similar.
COMMUNITY IMPACTS
Population Change
The 20,000 or so residents of the game community work
either in the tourism industry or the "other" industry.
There is a natural growth in the population of 1% per year
as well as migration into and out of the community
depending on the difference between the percentage of
unemployed workers and an assumed "frictional" employment
level of 3%. If the unemployment rate is above the
frictional level there will be population migration out of
the community; if the unemployment rate is less than 3%
there will be some migration of population into the
community.
Income Generation and Wealth
Residential income is the sum of wages paid to workers and
profits earned by owners in the tourism and "other"
industries. It is assumed that the major portion of
current income is expended on necessities purchased from
the "other" industry and luxuries including recreation-
days purchased from the natural resource managers, the
tourism industry.
The wealth of residents is characterized by the value of
residential property (broadly defined). Residential
property values are influenced by the number of residents;
with more residents, demand for property rises and
property values increase. Property taxes are the only
taxes paid by residents, and they are controlled by the
rate set by the game player.
Town Finances
The town derives revenues from five sources: daily tourist
fees, residential property taxes, tourism industry
property taxes, "other" industry property taxes, and
interest income on any accumulated budget surplus from
previous years (at a 10% annual rate). The town allocates
its budget to five categories of expenditures: tourism
promotion, "other" economic programs, social programs,
environmental programs, and an automatic interest expense
deducted for any accumulated deficit from previous years
(also at a 10% annual rate).
The Quality of Life
Economic, Social and Environmental Impacts
In addition to the income generated for residents by the
tourism and "other" industries in the communities, there
are a host of other economic, social, and environmental
characteristics that constitute the quality of community
life. All these characteristics are assumed be represented
by a set of four indexes. The economic index rises with
the average wage rate and falls with unemployment. The
social and environmental indexes rise with expenditures on
social and environmental programs and fall with increases
in the numbers of tourist-days and residential
recreational visits.
A social-environmental index, which is also used as a
measure of tourism attractiveness in the demand equation
for tourist visits, is provided by the geometric mean of
the social and environmental indexes. A quality of
community life is provided by the geometric mean of the
social-environmental index and the economic index. This
index is assumed to be the objective function for the
community and determines the annual game score.
GAME PLAY AND USE OF CONTROLS
Active Play
The first active screen of the game requests that the
player provides three initials for identification. These
are later used to record the player's final score for
comparison with other scores. The player is asked at once
whether the seven economic controls should be changed from
their initial settings. Figure 2 shows the actual screen
presented by the game.
As soon as the player is "READY FOR THE NEXT YEAR TO
BEGIN," a community overview screen is shown such as the
one in Figure 3. This screen provides current and past
values for tourism industry revenue, "other" industry
revenue, real disposable household income, the
unemployment rate, and the accumulated town surplus or
deficit. This screen also presents up to five news
headlines determined by the indexes described above and
the player's current and cumulative score.
The following sequence of screens afford the player the
opportunity to view five detailed informational
screens including:
"LOCAL GOVERNMENT FINANCIAL STATEMENTS,"
"SOCIAL ENVIRONMENTAL STATISTICS,"
"TOURISM BUSINESS STATISTICS," and
"OTHER INDUSTRY STATISTICS."
When the player is finally satisfied that enough
information has been viewed and that the appropriate
adjustments to the seven controls have been made, the next
year begins, and the model calculates changes to the
community and its industries. Overview results for the
next year are presented summarizing these calculations.
When a player chooses to change one or more of the
economic controls, a screen is presented that indicates
the current level of each of the controls and how controls
can be changed (Figure 4.) When all controls have been
changed to the desired values, the player begins the next
year of the administration.
Player Performance
The first informational screen of each year provides the
annual score for the player based on the overall well-
being index. However, the player (administrator) also must
attempt to balance the budget in order to keep from being
fired. If the deficit (or surplus) exceeds more than a few
percent of the total budget the planner is defamed in the
press. If the deficit (or surplus) is greater than about
10% the player will be fired despite the overall well-
being of the community. Thus, the relative size of the
town's budget deficit or surplus
Figure 2. Introductory screens
THE GAME:
You have been hired as the town planner for a community
with an important tourism industry, it generates 55%
of local income. The town budget is currently $11.9
million. Town revenues last year were generated from a $1
daily tourist tax ($1.4 Million last year), and property
taxes from residents ($8 M), tourism businesses ($2 M),and
other businesses ($0.5 M). The last administration
allocated $5.5 M to social programs, $1.5 M to tourism
promotion, $4.5 M to environmental programs, and $0.4 M to
other economic development programs. Your task is to
adjust tax rates and expenditures to achieve:
YOUR GOAL: GET THE HIGHEST POSSIBLE SCORE IN 10 YEARS
By maximizing social welfare in the town and keeping
yourjob as long as possible. Social welfare is defined as
a product of three indexes: economic, social and
environmental. Each is affected by the residents and
tourists.
Strike a key when ready...
PLEASE ENTER 3 INITIALS TO IDENTIFY YOURSELF
? TJT
THANK YOU, TJT
ECONOMIC CONTROLS LEFT IN PLACE BY THE PREVIOUS
ADMINISTRATION:
C(1) = $ 1.00 = FEE CHARGED EACH TOURIST
C(2) = $ 20.00 = EFFECTIVE RESIDENTIAL PROPERTY TAX
RATE (PER $1000)
C(3) = $20.00 = EFFECTIVE COMM./IND. PROPERTY TAX
RATE (FOR $1000)
C(4) = $1,500,000.00 = BUDGET FOR TOURIST PROMOTION
C(5) = $400,000.00 = BUDGET FOR OTHER ECONOMIC PROGRAMS
C(6) = $5,500,000.00 = BUDGET FOR SOCIAL PROGRAMS
C(7) = $4,500,000.00 = BUDGET FOR ENVIRONMENTAL
PROGRAMS
DO YOU WANT TO LEAVE CONTROLS UNCHANGED (Y OR N) ? Y
Figure 3. Community Overview
YEAR OF OFFICE = 1
TOURISM IND.REVNU. $82,181,280 This yr.; $79,999,880 Last
yr.; 2.73 percent change
OTHER IND.REVENUE $65,444,108 This yr.; $66,115,800
Last yr.; -1.02 percent change
REAL DISP HHD INC. $11,419 This yr.; $11,255 Last yr.;
1.45 percent change
UNEMPLOYMENT RATE 3.3 This yr.; 2.5 Last yr.;
32.83 percent change
ACCUMULATED TOWN
SURPLUS (DEFICIT-) $39,089 This yr.; $3- Last yr.
TOWN BUDGET:BALANCING ACT APPLAUDED BY TOWN'S PEOPLE.
PLANNER NEWS:MAYOR APPLAUDS PLANNING STAFF - PROJECTS GOOD
TIMES AHEAD
SCORE FOR TJT, THIS YEAR: 102, THIS GAME: 102
DO YOU WISH TO SEE LOCAL GOVERNMENT FINANCIAL STATEMENTS
(Y OR N) ? Y
Figure 4. Changing Controls
DO YOU WANT TO LEAVE CONTROLS UNCHANGED (Y OR N) ?
MOVE CURSOR TO CONTROL TO BE CHANGED
(USING UP AND DOWN ARROWS)
C(l) = $1.00 = FEE CHARGED EACH TOURIST
C(2) = $20.00 = EFFECTIVE RESIDENTIAL PROPERTY TAX
RATE (PER $1000)
C(3) = $20.00 = EFFECTIVE COMM./IND. PROPERTY TAX
RATE (PER $1000)
C(4) = $1,500,000.00 = BUDGET FOR TOURIST PROMOTION
C(5) = $400,000.00 = BUDGET FOR OTHER ECONOMIC PROGRAMS
C(6) = $5,500,000.00 = BUDGET-FOR SOCIAL PROGRAMS
C(7) = $4,500,000.00 = BUDGET FOR ENVIRONMENTAL
PROGRAMS
TYPE F1 TO EXIT
TYPE F2 TO CHANGE VALUE F2
ENTER NEW VALUE FOR CONTROL, THEN PRESS RETURN. C(1) =
1.50
F1
plus the overall well-being index are used to evaluate the
player's job performance. It is in fact the
budget-balancing problem that forces most game players
into early retirement.
The End of The Game and Scores
The game ends when the planner quits, is fired, or
survives 10 years in office. The score consists of the
average of the community well-being index weighted by the
residential population and summed over the
number of years in office normalized so that 100 points
per year reflects maintaining the initial levels of all
indexes. The distribution of scores generated from 100
plays where the only criterion was to balance the budget
is given in figure 5. The scores underlying the
illustration ranged from 402 to 1072 with only thirteen
scores above the norm of 100 points per year for the
entire ten-year period.
The distribution of scores obtained when following a
specific decision rule has been a major aide in assessing
the performance of the model. For example, the scores in
Figure 5 were determined by the rule that each year the
tourist fee would be changed by exactly the ratio of the
current deficit (or surplus) to the current number of
tourist days. We have adopted this budget-balancing,
status-quo type decision rule as our base line
distribution of scores. In this distribution, the major
cause of variation was games in which the player was fired
prior to the ten-year maximum. The secondary cause of
variation was the influence of random events on community
well-being. As the bimodal distribution of the figure
suggests, a typical game involved two events. When the
first event was serious, the budget-balancing, status-quo
player could not compensate and was fired between the
fourth and seventh year as reflected in the cluster of
scores between 400 and 700. The second event occasionally
resulted in a firing, but because of its timing late in
the game most players reached the tenth year.
Every game ends with the opportunity to view a listing of
the ranked scores from the previous 100 games played.
These results identify the players by initials, and the
individual games played by the score earned, the number of
years the office was held, and the nature of the "NEWS
FLASH" events that occurred while in office. These are
indicated by the initial letter of the news flash (F, G,
0, U, or B) and provide a basis to evaluate the specific
events that have most dramatically affected the well-being
of the game community. This list of scores has been useful
to teachers for evaluating uses of the game and relative
success of individual students in dealing with
hypothetical community problems.
APPLICATIONS
In Education
Beyond individual play as described above, the game has
proved to be of value in classroom exercises to simulate
town politics and decision making. In this application
each member of the class is assigned a role
and a role card which indicates a specific attitude about
development issues. Game years are marked off every 30
minutes. At the beginning of each new year individuals
meet in groups to develop policy and strategies for
dealing with other groups. As the year progresses groups
interact. At the end of the year the mayor and the rest of
the administration is advised by various public and
private groups about actions to be taken on behalf of the
community. These actions are input as changes to the seven
controls on the local economy and the game provides a
report of the influence of the changes and news on any
special occurrences.
It is of particular interest to note that a major problem
with the model was uncovered by students who figured out
how to "beat the game." Property taxes, in the model, are
treated as reductions in income prior to its allocation to
leisure and non-leisure goods. As such, the model predicts
no social outrage following rapid property tax increases
that are put back into social and environmental programs.
Accordingly, students have found that extremely high
scores can be earned by raising property taxes and
spreading the proceeds across social programs. This result
has lead to revisions in the game as well as the
identification of new topics for tourism research.
In Research
As suggested by the distribution of scores shown in Figure
5, the game acts as a stochastic simulation model when a
specific decision rule is adopted. A research version of
the game offers the player the opportunity to change
behavioral parameters and automatically restart with
specific decision rules. This enables the study of the
influence of individual parameters and identification of
behavioral relationships that lead to extreme scores. This
is the basis for an ongoing study of the model, its
parameters, and future refinements of the model.
Figure 5. Distribution of 100 Game Scores
(Vis. 13)
SUMMARY AND CONCLUSIONS
It has been the purpose of this paper to present the
details of a project that started toward a research
objective, later became an educational project, and is
currently on-going as a combined research and educational
program. Based in research findings from a variety of
studies, a "straw-man" computer game was constructed to
help in the presentation of a difficult concept in a new
research effort: the dynamic and stochastic nature of
tourism community behavior. This game was "borrowed" for
use in the classroom and subsequently revised for
educational purposes. We are convinced that game play is a
very exciting way of teaching tourism planning. We have
also found that students and others playing the game can
provide us with an unexpectedly rich source of research
assistance'. In our quest for a "reasonable"
characterization of tourism community behavior. We have
therefore altered our research procedures to explicitly
account for the research-education interaction. It has
been to our great delight that, on more than one occasion,
students have identified inadequacies in the model by
their ability to "beat the game."
The future for this type of project is bright. The model
has been used by over 50 researchers and 500 students, and
feedback from its use has suggested a wide range of
revisions. These include adding more sectors of the
economy toward the goal of a Leontief model of
transactions, adding more types of tourists and residents,
and adding more specific impacts thus far only generalized
as social and environmental impacts. It has also been
suggested that more of the complexities of community
government be included such as the role of zoning on
development. Many of these suggestions could result in
major improvements to the model; however, the game's
simplicity would need to be sacrificed. In the process of
deciding upon such improvements, it will be our overriding
concern to maintain a balance between the model's
usefulness as an educational tool and as a research
vehicle.
REFERENCES
Booz, Allen Public Administration Services, Inc. 1973.
The quality of life concept: A potential new tool for
decision makers. Prepared for the Environmental Protection
Agency. Washington, DC.
Henderson, J.E., and R.E. Quandt. 1958. Microeconomic
theory. New York: McGraw-Hill.
Loukissas, P.J. 1983. Public participation in community
tourism planning: A gaming simulation approach. Journal of
Travel Research 22(l):18-23.
Mathieson, A., and G. Wall. 1982. Tourism: Economic,
physical and social impacts. London: Longman.
Murphy, P.E. 1985. Tourism: A community approach. New
York: Methuen.
NE-137 Block. 1985. Six individual studies funded by
Agricultural Experiment Stations in the Northeast as part
of the Regional Hatch Project NE-137: "Economic Impacts of
Tourism on Rural Development in the Northeast. Proceedings
of the 1985 National Outdoor Recreation Trends Symposium
II. Myrtle Beach, SC, February 24-27, 1985. Department of
Parks, Recreation and Tourism Management, Clemson
University, Clemson, SC. Vol. II, pp. 230-298.
Pindyck, R.S., and D.L. Rubinfeld. 1976. Econometric
models and economic forecasts. New York: McGraw-Hill.
Tyrrell, T., P. Manheim, M. Altobello, T. Brown, M.
Hall-Arber, and M. Okrant. 1985. A Framework for
Evaluating the Impacts of Coastal Tourism Product Mixes at
the @cal Level: A Northeastern U.S. Regional Project.
Proceedings of the Northeast Economic Development
Symposium, Amherst, Massachusetts, May 28-30, 1985.
Northeast Regional Center for Rural Development, Pub. No.
40, Pennsylvania State University, University Park, PA.
pp. 74-88.
Wennergren, E.B., and H.H. Fullerton. 1972. Estimating
quality and location values of recreation resources.
Journal of Leisure Research 4 (Summer): 170-183.
Young, G. 1973. Tourism: Blessing or blight? Middlesex,
UK: Penguin Books.
APPENDIX A. THE BEHAVIORAL EQUATIONS
Equations of the model underlying the game are given in
Tables A1-A6. Initial coefficients are given in Table A7.
Table A1. Recreation and tourism behavioral equations.
Number of Tourists Visiting (TOUR):
LTOUR = CONST + B(l)LTOUR-1 +B(2)LTEXP-1 + B(3)LC(4) +
B(4)LSQEQ-1
Average Length of Visit in Days (DPTOUR):
LDPTOUR = CONST + B(5)LDPTOUR-1 +B(6)LWEATHER +
B(7)LSQEQ-1
Total Number of Tourist-Days (QTOUR):
QTOUR = TOUR * DPTOUR
Total Number of Visits made to the Recreation Site by
Residents per Year (QRES):
LQRES = CONST + B(12)LQRES-1 + B(13)LP + B(14)LDDHHDI-1 +
B(15)LWEATHER
Total Visits to Recreational Site Per Year (Q):
Q = QTOUR + QRES
Where:
L = a prefix indicating the logarithmic transformation,
-1 = a suffix indicating a one year lag.
B(l), B(2),... the predetermined coefficients.
C(4) = community expenditures on tourism promotion, one
of the seven controls described in the text.
TEXP-1 = average expenditures per day by tourists
= C(l) + P from the previous year, where C(l) is the
daily fee charged tourists.
SQEQ-1 = combined social - environmental index.
WEATHER = difference in degrees F from 75.
P = price per visit to the recreation site.
DDHHDI = deflated disposable household income
Table A2. Recreation/tourism industry behavioral
equations.
Number of Firms in the Tourism Industry (FIRMS):
LFIRMS = CONST + B(16)LFIRMS-1 + B(17)LPROFRAT-1
Average Price Charged per Day by Local Tourism Industry
(P):
LP = CONST + B(8)LP-1 + B(9)LQF-1
Average Wage Rate per Year per Employee of Tourism
Industry (W):
LW = CONST + B(10)LW-1 + B(11)LQF-1 + B(50)LEMPT-1
Number of Tourism Industry Employees per Firm (EMP):
LEMP = CONST + B(45)LEMP-1 + B(46)LQF
Average Assessed Value of Taxable Property per Firm in the
Tourism Industry (PVPFRM):
LPVPFRM = CONST + B(18)LPVPFRM-1 + B(19)LPROFRAT
Property taxes paid per tourism firm (TAXE):
TAXE = C(3) * (PVPFRM/V1000)
Where:
L = a prefix indicating the logarithmic transformation,
-1 = a suffix indicating a one year lag.
B(8), B(9),...the predetermined coefficients.
C(3) = industry property tax rate per $1000, one of the
seven controls.
PROFRAT = The average ratio of Revenues to Operating
costs in the Tourism Industry
QF = Average visits per tourism firm.
EMPT = Ratio of the number of local jobs to the number
of persons in the labor force.
Table A3. "Other" industry behavioral equations.
Number of Units of Product Exported by the "Other" Local
Industry (OEXPORT):
LOEXPORT = CONST + B(23)LOEXPORT-l(+-5%) + B(22)LC(5)
Number of Units of Production Sold Domestically by the
"Other" Local Industry (ODOMEST):
LODOMEST = CONST + B(23)LODOMEST-1 + B(24)LPOTHER +
B(25)LDDIPH
Number of Firms in the "Other" Local Industry (OFIRMS):
LOFIRMS = CONST + B(28)LOFIRMS-1 + B(29)LOPROFRAT
Average Price Charged per Unit of Product by the "Other"
Industry (POTHER):
LPOTHER =CONST + B(20)LPOTHER-1 + B(21)LQOTHERF
Average Wage Rate per Year per Employee of "Other"
Industry (WOTHER):
LWOTHER = CONST + B(26)LWOTHER-1 + B(27)LQOTHERF+
B(50)LEMPT
Number of Employees per Firm in the "Other" Industry
(OEMP):
LOEMP = CONST + B(47)LOEMP-1 + B(48)LQOTHERF
Average Assessed Value of Taxable Property per Firm in the
"Other" Industry:
LOPVPFRM = CONST + B(30)LOPVPFRM-1 + B(31)LOPROFRAT
Property taxes paid per "Other" firm (OTAXE):
OTAXE = C(3) * (OPVPFRM/1000)
Where:
L = a prefix indicating the logarithm.
-1 = a suffix indicating a one year lag.
B(20), B(21), ... the predetermined coefficients.
C(3) = industry property tax rate per $1000.
C(5) = community expenditures on "other" economic
development programs.
DDHHDI = deflated disposable income per household
OPROFRAT = The ratio of Revenues to Operating costs in
the "Other" Industry
QOTHERF = Average output per "other" firm.
EMPT = Ratio of the number of local jobs to the number
of persons in the labor force.
Table A4. Residential population and income equations.
Number of Residents in the Community (RES):
RES = RES-l*(1.01) + B(32)(3% - UNEMP)
Residential Income (INCOME):
INCOME = (WAGE+PPFRM)*FIRMS + (OWAGE+OPPFRM)*OFIRMS
Residential Property Value per resident (PVPRES):
LPVPRES = CONST + B(43)LPVPRES-1 + B(44)LRES
Property taxes paid per Resident (RTAXE):
RTAXE = C(2) * (PVPRES/1000)
Where:
L = a prefix indicating the logarithmic transformation,
-1 = a suffix indicating a one year lag.
B(32),... = the predetermined coefficients.
C(2) = residential property tax rate per $1000.
UNEMP = the percent of the labor force unemployed
PPFRM and OPPFRM = profits per firm in the tourism and
"other" industries.
Table A5. Town budget and expenditure equations.
Tourist Fees (FEES):
FEES = C(l) QTOUR
Residential Property Taxes (RTAXES):
RTAXES = RTAXE * RESBASE
Industrial Property Taxes (TAXES + OTAXES):
TAXES + OTAXES = TAXE * FIRMS + OTAXE * OFIRMS
Interest Income/Expense (INTINC and INTEXP):
INTINC =.10 * ACCSURP-1 if ACCSURP-1 > 0.
INTEXP =.10 * ACCSURP-1 if ACCSURP-1 < 0.
Town Budget (BUDGET):
BUDGET = FEES + RTAXES + TAXES + OTAXES + INTINC
Town Expenditures (EXPS):
EXPS = C(4) + C(5) + C(6) + C(7) + INTEXP
Town Surplus or Deficit (SURP):
SURP = BUDGET - EXPS
Accummulated town Surplus or Deficit (ACCSURP):
ACCSURP = ACCSURP-1 + SURP
Where:
-1 = a suffix indicating a one year lag.
C(l) = residential property tax rate per $1000.
C(4) = tourism promotion expenditures
C(5) = "other" economic development expenditures
C(6) = social program expenditures
C(7) = environmental program expenditures
Table A6. Economic, social and environmental quality
indexes.
Econonomic Quality Index (YQ):
LYQX = CONST + B(49)LSW-1 + B(41)L(RPCI/AVGPCI) +
B(42)LEMPT
YQ = YQ-1 + YQX
Social Quality Index (SQ):
LSQX = CONST + B(33)LSW-1 + B(34)LSQEX + B(35)LQRESMA +
B(36)LQTOURMA
SQ = SQ-1 + SQX
Environmental Quality Index (EQ):
LEQX = CONST + B(37)LSW-1 + B(38)LEQEX + B(39)LQRESMA +
B(40)LQTOURMA
EQ = EQ-1 + EQX
Combined Social - Environmental Index (SQEQ):
LSQEQ = (LSQ + LEQ)/2
Overall Community Quality (Social Welfare) Index (SW):
LSW= (LYQ + SQEQ)/2
Where:
L... = a prefix indicating the logarithm.
-1 = a suffix indicating a one year lag.
B(33), B(34),... = predetermined coefficients.
RPCI/AVGPCI = the ratio of current real per capita
income to its initial value.
EMPT = Ratio of the number of local jobs to the number
of persons in the labor force.
QRESMA and QTOURMA = two-year moving averages of the
number of resident recreation visits and tourist
days, respectively.
Table A7. Coefficient definitions and initial values.
(Vis. 14)
LAG: Lag coefficient, usually one year on explained
variable.
SRE: Short-run elasticity, due to nature of double log
model.
SREMA: Short-run elasticity on 2-period moving average of
explanatory variable.
A SYSTEM FOR ESTIMATING LOCAL ECONOMIC IMPACTS OF
RECREATION AND TOURISM
Daniel J. Stynes and Dennis B. Propst
Michigan State University
INTRODUCTION
Estimates of the local and regional economic effects of
recreation and tourism activity have been in high demand
over the past decade (Fleming and Toepper 1990; Getz 1991;
Johnson et al. 1989). The vast majority of research on
economic impacts of recreation and tourism has been single
case studies. Although hundreds of such studies have been
conducted, our understanding of how impacts vary across
types of projects and communities remains somewhat
inadequate for making predictions. The lack of progress in
general knowledge is due in part to the wide variation in
projects and communities/regions that have been studied,
as well as considerable variation in the methods and
quality of the tourism economic impact studies that have
been conducted. The lack of consistency in procedures and
measures make it difficult to compare results across
studies in search of more general knowledge.
Recreation and tourism economic impact studies tend to
fall into three groups. In the first group are spending
studies, which estimate tourist spending as an indicator
of the direct effects or importance of tourism to a
region. Regional economic structure and secondary effects,
if included at all in these studies, are generally
captured by the application of published multipliers.
Abuses of multipliers in these studies are quite
widespread (Archer 1984).
A second group of studies focuses on local regional
economic structure and the development of input output
models for tourism-oriented economies. These studies,
carried out largely by regional economists, are not as
prevalent as spending studies. Improvements in the
analysis of the regional economic structure in these
studies are often offset by less attention to the
estimation of tourist spending.
In the third category are studies that both estimate
spending and apply these estimates to a regional economic
model. Combining the regional economist's understanding of
regional economic models with the recreation and tourism
researcher's understanding of visitor spending patterns
and tourism industry structure yields a more complete
assessment of tourism's economic effects.
A SYSTEM
In this paper we summarize a system for estimating the
economic impacts of recreation and tourism. The purpose of
the system is to reduce the costs of gathering spending
data, estimating local input-output models, and
interfacing the two. By making spending data more
available to regional economists, and input-output models
more accessible to recreation and tourism researchers, we
hope to encourage more complete and comparable studies of
recreation and tourism's local economic effects. The
system has evolved over a number of applications and
continues to be refined, as further studies are conducted.
Our focus has been on developing procedures for estimating
impacts that
1. Do not require extensive primary data collection.
2. Are applicable to a wide range of situations and
questions.
3. Can be carried out at different levels of detail and
accuracy, as the situation may dictate.
4. Can generate estimates of direct spending as well as
estimates of indirect and induced effects in terms of
output, income, or employment.
5. Automates many routine and technical matters so that
users can focus on application and interpretation.
6. Are tailored to the particular problems and
characteristics of recreation and tourism applications.
The elements of such a system have been known for some
time to researchers and regional economists studying
recreation and tourism impacts. The basic problem is to
obtain reliable estimates of the changes in recreation and
tourism spending associated with an action, usually for
well-defined visitor market segments, and then to bridge
these final demand estimates to regional input-output
models in order to estimate local economic impacts (Propst
1985, 1988; Alward and Lofting 1985; Cordell et al. 1987;
Bergstrom et al. 1990; Johnson et al. 1989).
Recent developments in microcomputer-based input-output
modeling software (Brucker et al. 1987) have made regional
economic models much more accessible and therefore most
recent efforts have been closely tied to such models.
Building on the studies cited above, our system relies
extensively on the IMPLAN input-output modeling system.
IMPLAN has been developed by the Land Management Planning
Systems Group of the USDA Forest Service in Fort Collins,
CO (Alward and Palmer 1983). The U.S. Army Corps of
Engineers Waterways Experiment Station has supported the
development of spending profiles for various visitor
segments and the general development of the system
described here. The framework for the system and the
research agenda to develop it were established in a
national workshop on assessing the economic impacts of
recreation and tourism held at Michigan State University
in 1984 (Propst 1985).
THE BASIC APPROACH
The basic approach for estimating the economic impacts of
recreation and tourism on a local area involves five
steps:
1. Defining the problem. Problem definition requires
the specification of the actions or activities of interest
and the delineation of the impacted region.
2. Estimating the change in final demand (spending)
associated with the action. For recreation and tourism
applications this usually involves estimating spending or
changes in spending of affected visitors.
3. Estimating a regional economic model for the local
area. The model captures the economic structure of the
region, usually via an input-output model or one of
several simplifications thereof (e.g., economic base
models, location quotients and regional multipliers).
4. Converting from estimates of visitor spending within
broad categories to a final demand vector within specific
economic sectors. This step involves what is known as
"bridging" and "margining."
5. Applying the final demand vector to the regional
economic model and interpreting the results.
The definition of the problem in step one identifies the
relevant market segments for which spending estimates are
needed and defines the region for which a model must be
estimated. Given a set of m recreation or tourism market
segments, spending profiles for each segment, and
estimates of the number of visitors by segment, spending
can be estimated as a simple weighted average, as follows:
(1)
m
Sj = N * S Mi * sij
i=1
Where
Sj = total spending in category j, j=1,...J
N = total number of visitors
m = number of segments
Mi = segment i's share of total visits, i=1,...m, and
sij = average spending of a member of segment i on
category j. These are what we term "spending profiles."
Total spending can be calculated by summing across
spending categories. The economic impact of this spending
is obtained by applying the vector of spending(Sj,j=1,..J)
to a set of category/sector-specific multipliers or an
input-output model. An intermediate step is required to
transform the spending vector (S) to a final demand vector
(FD), consistent with the sectors and assumptions of the
model. The spending vector generally must be bridged and
margined to produce a suitable final demand vector.
Bridging converts the categories in which visitor spending
is measured to the sectors in the regional economic model
receiving this spending. As part of the bridging process,
the retail, wholesale, and transportation margins
associated with retail purchases must be separated out and
the remaining retail dollars allocated to appropriate
local production sectors or imports.
The final demand vector can then be applied to an
input-output model for the region. If detailed estimates
of effects by sector are not needed, a set of Keynesian
type multipliers (Archer 1984), derived from an
input-output model or secondary sources, can be used to
estimate impacts. Impacts can be reported in terms of
sales, income, value added, or employment in total or for
individual sectors. Impacts can also be separated into
direct, indirect, and induced effects, and ratio
multipliers may be calculated. Impacts can also be
estimated separately for each segment if desired (Strang
1971). In symbols,
FDk = Sj * Bjk (2)
I = R * FDk (3)
Where
FD = a final demand vector of spending changes
B = a bridge table to convert from spending categories
j to sectors k
R = represents an input-output model or a set of
sector specific Keynesian multipliers. Formally,
R would be the Leontief inverse (R= (I-A)-1) when
using an input-output model.
I = impacts, expressed as changes in output, income,
or employment resulting from the change in final
demand.
INFORMATION/DATA REQUIREMENTS
The three equations identify the five basic information
requirements for a system to estimate the economic impacts
of recreation and tourism.
1. An estimate of total visits or visitors (N) affected
by an action.
2. A set of market segments and the percentages (Mi =
market shares) of visitors within each segment.
3. Spending profiles (sij) giving the average spending
within an appropriate set of spending categories j for
each segment i.
4. A bridge table (Bjk) to convert spending from
categories j to a final demand vector by economic sectors
k.
5. An input-output model for the region or a set of
sector-specific multipliers.
The first three items are needed to estimate spending, the
last two to estimate secondary effects of this spending
and identify the impacted sectors. Of these five
information requirements, obtaining spending estimates and
a suitable regional economic model has accounted for the
majority of time and cost in conducting tourism impact
assessments. Improvements in micro-computer based
input-output modeling systems, like IMPLAN, make the
estimation of local input-output models relatively quick
and easy. Our system is therefore grounded in the IMPLAN
system. IMPLAN permits the estimation of customized
input-output models for any county or grouping of counties
in the United States. IMPLAN also includes a basic
framework and software for carrying out impact analyses
with the model (Taylor et al. 1992).
We build on the IMPLAN system primarily by reducing the
costs of assembling spending estimates and bridging these
to models estimated within IMPLAN. In the simplest of
applications of our system, the analyst identifies a set
of visitor segments affected by a given action. Existing
spending profiles for these segments are examined on a
spreadsheet and adapted or adjusted as necessary. The
analyst estimates the number of visitors affected by the
action within each segment. Segment shares are multiplied
by the spending profiles for each segment and summed to
obtain total spending by sector, as in equation (1). A
bridge table converts this spending to a final demand
vector. The analyst estimates an input-output model
for the local region using IMPLAN, imports the final
demand vector, and uses IMPLAN's impact analysis
module to estimate impacts.
The key user input is an estimate of the number of
different types of visitors that are affected. Based on
the situation or project to be evaluated, the user must
estimate how many visitors are affected and must be able
to estimate the proportions of these visitors represented
by the principal market segments. Some judgement is also
required in selecting and/or adjusting the spending
profiles, defining the study region, and estimating the
input-output model. The number and types of technical
adjustments and application of the system will vary with
the nature of the problem and the skills of the analyst.
THE BASIC SYSTEM: FIVE ELEMENTS
1. Market segments. We identify various sets of recreation
and tourism market segments and provide a framework for
estimating spending and impacts of particular segments. To
be appropriate for economic impact analysis, the segments
should be easily identified, reasonably homogeneous in
their spending patterns, and relevant to the kinds of
management, planning, and marketing decisions to be
evaluated. While suitable segments may vary from one
situation to another, most situations recommend that local
visitors be distinguished from non-locals and that
overnight visitors be distinguished from day users.
Separating visitors based on lodging type (campers vs
motel users) and transportation mode (auto vs air) is also
recommended, as these usually imply distinct spending
patterns. Many applications suggest various recreation
activity segments, such as boaters, anglers, sightseers,
downhill skiers, and the like.
2. Spending profiles. We have empirically estimated
spending patterns for particular segments and evaluated
the generalizability of segment spending profiles across
sites. A spending profile is a vector of the amount spent
by an average visitor across a set of spending categories.
Supplementing an initial set of spending profiles are a
set of standardized spending categories and measurement
protocols for generating spending estimates for additional
segments or testing the generalizability of these profiles
over time and space. The system facilitates the
accumulation of comparable spending data for a variety of
recreation and tourism market segments over different
regions and time. Price indices are used to update
spending profiles over time, and models will be proposed
for generalizing spending patterns over space. A
spreadsheet combines spending profiles by segment with
user entered market shares to estimate total spending.
3. Margining and bridge tables. A set of recommended
margining and bridge tables is provided to assist in
matching recreation and tourism spending data with
standard economic sectors, in this case IMPLAN's 528
sectors. The formal procedures in IMPLAN involve choosing
spending categories from an IMPLAN database file and
entering the amount of total spending in each category
from step (2) above. By simplifying and automating the
bridging and margining process, we hope to reduce errors
that are commonly made at this step. Many recreation and
tourism impact assessments have neglected the necessary
margining of tourist retail spending and applied
multipliers directly to total spending. As many goods that
are purchased by tourists are not produced in the local
area, impact estimates can be significantly inflated if
multipliers are applied to total retail spending rather
than appropriate local margins.
4. Local input-output models. The IMPLAN system has been
adopted as the system for generating an input-output model
for a local region. IMPLAN is a flexible system for
estimating customized models for any county or grouping of
counties in the United States. While the system is broadly
applicable, it has been developed specifically to evaluate
the impacts of resource management and policy questions
(Alward and Palmer 1983).
5. Impact estimation procedures. The IMPLAN system also
includes software for estimating impacts of alternative
management activities. This basically involves applying a
vector representing the final demand change associated
with an action to the regional input-output model. In a
typical tourism application, the final demand vector is
the change in visitor spending in the local region.
Recommended sectoral aggregation templates for use with
IMPLAN and reporting formats for recreation and tourism
are also being proposed as part of the system.
VISITOR SEGMENTS
Segments play three important roles in the system. First,
segments are the means by which an analyst identifies who
is affected by a particular action. The more precisely a
user can identify impacted market segments, the more the
economic impact estimates can be tailored to the
particular situation. Second, by forming segments that are
relatively homogeneous in their spending patterns, we
reduce variation and sampling errors in survey-based
estimates of spending. It is not uncommon in recreation
and travel spending studies for sampling errors to
approach and even exceed one hundred percent of the mean.
Given extremely wide variation in spending patterns across
distinct types of tourists, there are significant
efficiencies in estimating spending within more
homogeneous subgroups of visitors. Finally, segments are
the primary vehicle for explaining why one site or region
may generate a different pattern of spending and impacts
than another. Total spending will depend on both the
number and types of visitors that are affected. The
differences in the spending impacts of a day use area,
campground, or resort development can largely be explained
by the kinds of visitors that will be attracted.
One of our assumptions is that the segment spending
profiles will be more generalizable across regions, sites,
and applications than the numbers and types of visitors
that are affected by an action. The latter must be
estimated in each specific instance by someone
knowledgeable with the study area and the proposed action.
Our system helps to redirect some attention from the
technical issues surrounding estimates of spending and
input output models, to the more basic questions of how
many people are affected and who they are.
The degree to which spending profiles are generalizable
across applications hinges a great deal on how the
segments are defined. An engineering approach to
estimating visitor spending suggests some of the key
variables for defining segments. To produce a trip,
visitors generally require transportation, food, and
lodging (if overnight). Expenses in these categories will
depend on party size, length of stay, and the mode or type
of transportation, food service, and lodging.
Expenses are also incurred for recreation, entertainment,
and souvenirs. These expenses will vary across recreation
activity groups. Parties making trips for particular
activities like boating, fishing, and downhill skiing will
incur expenses that are specific to these activities. In
short, much of the systematic variation in spending can be
explained by length of stay in the area, party size,
lodging type, transportation mode, distance traveled, and
primary activities. Variation in spending due to party
size and length of stay can be handled in the choice of
the units of analysis (e.g., visitor day, visit, or party
trip). The remaining variables form the basis for defining
segments with similar within segment spending patterns.
No single set of segments will be applicable everywhere.
The applications that follow illustrate a range of
segmentations that were developed for specific
applications. As more spending profiles are assembled in a
consistent manner for distinct visitor segments and
regions, users will be able to selectively choose and
adjust a set of profiles to fit a particular application.
Where more reliable local spending estimates are required,
spending surveys may be conducted. It is, however, our
belief that adjusting or adapting existing profiles that
have been carefully constructed will often be preferred to
attempting to survey visitors with inadequate time or
resources for the job. This is particularly the case when
the applications of an impact analysis do not require much
more than rough estimates of total spending or jobs.
SPENDING CATEGORIES
Spending categories have been developed to permit
flexibility in the level of aggregation, while also
facilitating the process of bridging tourist spending data
to IMPLAN sectors. For example, food purchased for
off-site consumption is measured separately from on-site
restaurant spending.
The basic survey instrument we have developed for the U.S.
Army Corps of Engineers measures trip spending in 33
detailed categories and durable goods spending within 32
categories. These are grouped into 8 major trip categories
and 5 durable goods categories. The categories are shown
in Table 1 along with estimates of average spending for
visitors to the Upper Mississippi River System in 1990-91.
Both trip and durable goods spending are estimated on a
party-trip basis, so that total spending for a given year
can be estimated by multiplying these averages by the
number of party trips to an area. Calculations are
normally carried out using the average spending of
specific visitor segments, as illustrated in subsequent
tables.
Bridge tables have been developed for converting these
trip and durable spending categories to IMPLAN's 528
sectors. As part of the bridging process durable goods
purchases are divided between items purchased new, items
bought used from a dealer, and items bought used from
individuals. In assessing economic impacts of new
purchases, margins are taken, and the remainder of costs
are bridged to manufacturing sectors. Only the margins are
included for used goods bought from dealers, and used
goods bought from households are allocated to household
income. Other complexities of handling durable goods
purchases are discussed elsewhere (Stynes and Propst
1992).
A further complication in defining spending categories is
capturing where spending takes place in order to identify
the impacted region. Recreation and tourism spending is
normally divided between (1) spending at home, (2)
spending en route, and (3) spending at the destination.
Our trip spending profiles
Table 1. Trip and durable goods spending categories,
USACE and UMRS studies spending profiles for visitors to
Upper Mississippi River System, 1989-90 ($ per party per
trip).
(Vis. 15)
divide spending by category between spending within 30
miles of the recreation/tourism site or destination
and spending elsewhere on the trip. A 30 mile radius of a
site was selected to capture the spending at the
destination. In many parts of the country, this roughly
corresponds to a county or perhaps two counties if the
recreation site is on the county border. Where this is the
case, the measure of where spending occurs can be made to
correspond with the region for which impacts are being
estimated. In estimating impacts for larger regions, some
portion of the "beyond 30 mile" spending can be included.
Trip spending can be aggregated into as few as 5
categories if desired. For example, a typical tourist
spending profile from a simple travel spending survey
might only include lodging, food and beverages,
transportation, recreation, and miscellaneous expenses.
Some studies separate auto from other travel expenses,
restaurant meals from groceries, and motel expenses from
camping. We do not recommend measuring spending in every
study in as detailed categories as we report here. It will
generally be more efficient to measure spending in broader
categories and, if desired, use more detailed profiles,
like ours, to apportion amounts within subcategories
according to the percentages we have reported. For
example, the portions of food and beverage expenses that
we report going for groceries vs restaurant meals, can be
used to further disaggregate a survey estimate that only
included the category "food and beverages." These
allocations, of course, rest on assumptions that the
percentages we report generalize, at least within
subcategories, to the intended applications. The more
precise spending categories reduce aggregation errors in
applying the input output model and are necessary to
estimate impacts within more narrowly defined local
economic sectors. Buying meals in restaurants will have
somewhat different local impacts than purchasing
groceries.
COMPARABLE ESTIMATES OF VISITOR SPENDING AND ECONOMIC
IMPACTS
An important motivation for the system we have developed
is to facilitate and encourage more standardized and
comparable estimates of spending and economic impacts. A
standard survey instrument, data gathering method, and
analysis package has been developed for generating visitor
spending profiles and estimates within designated market
segments.
Our spending data collection system relies on a short
on-site interview followed by a mailback questionnaire to
measure trip spending. The instrument and approach are a
refinement to procedures developed initially for the PARVS
study (Cordell et al. 1987). Although other units of
analysis are possible, we measure and estimate spending on
a party-trip basis. Spending on durable goods, if desired,
is gathered in the on-site part of the survey, to avoid
confusion with trip spending. The on-site survey also
provides the basic information to form segments and serves
to check for and adjust for potential non-response biases
in the mailback instruments.
The procedures have been used at 12 USACE reservoirs
across the country (Propst et al. 1991) and at 150 sites
along the Upper Mississippi River System (Propst et al.
1992). Other applications of the measurement procedures
are in progress. Across the two major applications to
date, more than 4,500 visitors have been sampled and
comparable detailed trip spending data have been gathered
for more than 3,000 visitor parties. Trip spending
profiles can be estimated within less than 10% sampling
error for primary visitor segments. Segments with higher
sampling errors can be targeted for future surveys, and by
combining data sets across sites, sufficient samples can
be built over time to estimate spending for more narrowly
defined user groups.
The comparability of data across different sites permits
the testing of models to explain variations in spending
profiles of particular segments across sites and regions.
For example, it is clear that visitor spending depends on
the availability of places to spend money in the local
area. The regional economic structure will therefore
influence direct spending as well as the secondary
(indirect and induced) effects of this spending. Simple
spending shifters based upon rough indices of places to
spend money and local price indices is one approach we are
pursuing. Having reasonably comparable data across a range
of sites and regions is a necessary prerequisite to these
efforts.
APPLICATIONS
We briefly summarize four applications of the general
system which illustrate the stages in its development and
various ways the system may be used.
Spending Impacts of a Marina
We first implemented the idea of building spending
profiles for narrowly defined visitor segments in a study
of Michigan boaters (Stynes et al. 1983). The segments
allowed us to apply the results of a statewide spending
survey to a wide range of applications and study areas.
Building segment spending profiles into an electronic
spreadsheet proved successful in bringing statewide survey
results down to a local level, where most decisions were
being made.
Spreadsheets were developed for general boating
applications and for assessing spending impacts of marinas
on local communities. Users of the spreadsheets simply
estimated the numbers and types of boats attracted by a
program or facility and entered these figures on a
spreadsheet to obtain estimates of boater spending (See
Table 2). Price indices were later added to allow users to
update the spending profiles over time. Given suitable
local information, users could also adjust the spending
profiles to account for difference in local areas. Lacking
better local data, the statewide averages could be used.
In the application to marinas, boaters were classified
into three segments based on the characteristics of their
boat, propulsion type, and size. The segments captured
differences in spending patterns of power boaters vs
sailors, and the higher spending associated with larger
craft. Spending categories were divided between trip
expenses and annual costs for storage, insurance, and
maintenance. In this early application we did not go
beyond estimates of direct spending. Table 2 depicts the
basic spreadsheet. A user enters the number of slips and
segment shares at the top. These are multiplied by segment
spending profiles in the middle of the table to yield
total spending by segment and sector at the bottom. One
can readily simulate spending effects of alternative
scales of marina development and mix of boaters attracted.
The spreadsheet also facilitates the production of
histograms and piecharts, as desired.
Table 2. Michigan marina boater spending spreadsheet,
1989 spending estimates for a 75 boat marina.
(Vis. 16)
a. Total craft-related spending calculated as number of
craft times per boat averages.
b. Total trip spending calculated as per day average times
number of boat days.
Economic Impacts of USACE Reservoirs on Local Economies
The U.S. Army Corps of Engineers (USACE) Waterways
Experiment Station provided support to develop the primary
components of a system to estimate the impacts of USACE
activities on local economies. The components of this
system included reasonably standardized methods for
measuring and reporting spending information, empirical
estimates of spending for a suitable set of market
segments, and procedures to link the spending data to
regional economic models. The IMPLAN system fit our needs
for input-output models very well, so we concentrated on
measuring spending in categories consistent with IMPLAN
and developing interfaces to take advantage of IMPLAN's
model estimation and impact estimation routines. USACE
visitors were divided into twelve segments defined by four
dichotomous variables: local resident or not, boater or
non-boater, day user or overnight visitor, and staying in
a campground or motel. Trip and durable goods spending
profiles were estimated for each segment based on surveys
at twelve USACE projects across the country. Spending
profiles of individual segments were compared across the
12 USACE projects, and data were also pooled to develop
"national average" spendingprofiles for each segment.
These trip spending profiles are summarized in Table 3 for
the 12 segments. The spending data can be used to estimate
impacts at the local, state, or national level for the
particular. projects (estimates for Lake Shelbyville, IL,
are reported in Jackson et al. 1991). More importantly,
the profiles provide a basis for making estimates for
similar projects around the country, without incurring the
costs of additional visitor spending surveys. Users of the
system may choose spending profiles for a comparable
project, a weighted average from different projects, or
the "national averages." The latter, for example, have
been applied to total USACE use by segment to obtain
estimates of total spending nationwide by all USACE
visitors in 1990 (Propst et al. 1991).
Tables to bridge visitor spending to IMPLAN's 528 sectors
were developed. By applying the estimated spending to an
input-output model for the local region (estimated using
IMPLAN), direct, indirect, and induced effects of Corps
visitors can be estimated in terms of sales, income, and
employment. IMPLAN's analysis, reporting, and aggregation
schemes provide considerable flexibility in the types of
reports and economic measures that can be produced.
Aggregation schemes have been developed to focus
specifically on the sectors most affected by recreation
and tourism activity.
Economic Impacts of Recreation Sites Along the Upper
Mississippi River System (UMRS)
A study of the economic impacts of recreation along the
Upper Mississippi River System provided the opportunity to
repeat the same spending measurement procedures as in the
previous study at a more diverse set of sites providing a
broader array of recreation activities. The segments
defined in the previous study were reduced from 12 to 6 in
order to maintain adequate sample sizes within the
overnight segments. Profiles for particular segments of
UMRS visitors can be applied to subregions and individual
sites. As in the previous study, analysts must be able to
identify the number and types (segments) of visitors
affected by an action. As spending was measured in the
same categories as in the previous study, the bridge
tables and other procedures for adapting the spending data
for use with IMPLAN are the same. The study provides an
additional set of spending profiles that are more
representative of sites that attract large numbers of day
users and sightseers. Table 4 illustrates the durable
goods spending profiles for the six UMRS visitor segments.
We will be combining the data sets from the surveys at the
UMRS and 12 USACE reservoirs in order to develop models
for explaining variations in spending patterns across
sites.
Economic Impacts of National Parks
A final example illustrates how the system may be readily
adapted to existing spending data sets and other
recreation and tourism resource management organizations.
Great Smoky Mountains National Park (GRSM) had conducted a
visitor spending survey in 1985 (Peine and Renfro 1988).
The survey measured spending of visitors over the previous
24-hour period within 10 broad categories. The original
analysis did not use a segmented approach; however, key
segmentation variables were included in the survey. The
survey data were therefore further analyzed to generate
spending profiles for eight distinct segments of park
users: (1) LOCAL = local visitors, (2) NL-DAY = day users
from outside the local area, (3) CAMP-IN = campers staying
in the park, (4) MOT-IN = visitors staying in park hotels,
(5) OVN-IN = backcountry users, (6) CAMP = campers staying
outside the park, (7) MOTEL= visitors staying in motels
outside the park, and (8) OVN = other overnight visitors
(See Table 5).
By estimating the numbers and shares of visitors within
each segment, spending and economic impacts can be
estimated for alternative mixes of visitors. The
spreadsheet underlying Table 5 was used to estimate
Table 3. Trip spending profiles by segment and major
categories, pooled averages 12 USACE reservoirs, 1989-90.
(Vis. 17)
Table 4. Spending on durable goods by visitor segment and
category ($ per party-trip) Upper Mississippi River
System, 1990-91.
(Vis. 18)
spending of summer-fall visitors, spring-winter visitors,
and visitors staying inside the park. Spending and
economic impacts were also estimated for particular
communities and counties around the park and projected to
subsequent years based on park visitation data. The GRSM
visitor spending estimates were bridged to IMPLAN sectors
by matching the spending categories used by Peine and
Renfro with those in our existing bridge tables. IMPLAN
software was used to estimate a model for the local area,
import the final demand vector, and estimate impacts
(Stynes 1992).
Table 6 reports the employment effects of $188 million in
visitor spending in 1985 within a three-county region on
the Tennessee side of the park. IMPLAN sectors have been
aggregated in the table to 33 broader categories that high
light the impacted sectors. Visitor spending contributed
to 6,773jobs in the region, about Table 5. Visitor
spending profiles by segment, Great Smoky Mountain NP,
1985.
Table 5 (Vis. 19)
14% of all jobs in the three-county area in 1982. Similar
tables maybe produced to express impacts in terms of
sales, income, or value added. Notice that the multipliers
are much lower than those commonly used in recreation and
tourism studies, and the principal secondary effects are
from household spending of wages and salaries earned from
tourism (induced effects). The use of an input-output
model permits a quite detailed analysis of the impacted
sectors. By estimating impacts separately for individual
segments, visitor market segments can be directly linked
to the economic sectors that they benefit. More generally,
the example illustrates how reasonably accurate and fairly
detailed local economic impact estimates can be generated
without the expense of gathering any new primary data.
This will likely be the most common use of the general
system.
CONCLUSIONS
This basic system for estimating economic impacts can be
used for a wide range of recreation and tourism
applications. The principal difference across applications
or clients is the set of spending profiles that must be
assembled and correspondingly, the definition of suitable
segments. The segments for the USACE are oriented toward
water-based recreation activities and defined to be
consistent with the Corps' use estimation and reporting
system. Between the national study at 12 reservoirs and
the UMRS study, spending profiles have been estimated for
the Corps'primary user groups. Further research will be
needed to model spatial variations in spending and refine
spending estimates for some subgroups, but the existing
Profiles provide a firm basis to build upon.
Table 6. Employment effects, Great Smoky Mountains NP 1985
visitor spending, IMPLAN Estimates for Blount, Cocke, and
Sevier Counties, TN.
(Vis.20)
The USACE segments will be useful for application to,
other water-based recreation sites. In order to apply this
system to other types of sites, spending profiles must be
assembled for appropriate segments. For example, spending
profiles for well-defined segments of tourists, forest
recreationists, particular recreation activity groups, and
state or national park visitors are needed. Many such data
sets exist, but they must be individually evaluated and
assembled into a more accessible database. To the extent
that consistent spending categories, units of analysis,
and measurement procedures are used, spending estimates
may be compared across studies and sites to assess
generalizability of these profiles.
Our system is not intended to preclude primary collection
of spending data, when it is justified and affordable, or
to advance one input-output modeling system over another.
We encourage the careful estimation of spending patterns
of visitors to different sites/regions, and especially the
estimation of spending profiles for homogeneous subgroups
of visitors. Our system should help to encourage greater
comparability of estimates across studies, and therefore
facilitate the building of a recreation/tourism spending
database. This, in turn, will permit research into the
generalizability of spending patterns across applications
and greatly reduce the costs of making spending estimates.
Adapting our general procedures to other input-output
modeling systems is straightfoward. The use of different
input-output modeling systems, however, does introduce
other comparability issues. Distinct input-output models
employ a host of different assumptions, may begin with
different databases and sectoral definitions, and often
define and calculate multipliers in slightly different
ways. These issues are of more interest to regional
economists than most recreation and tourism researchers.
We have therefore restricted our attention to a single
system at this time. IMPLAN has the advantage of being in
the public domain and being the system most familiar to
analysts studying impacts of natural resource-related
activities.
By selecting a particular input-output modeling system and
providing some standardization in spending measurement and
reporting, we have been able to simplify the process for
estimating economic impacts. The standardization should
also improve comparability of results across distinct
applications. Greater comparability is a key to developing
more generalizable knowledge about the impacts of
recreation and tourism on local economies. The system
should also greatly reduce the costs of conducting
economic impact assessments by reducing the costs of
acquiring spending data and applying it to a regional
economic model. In this way, we hope to shift some
attention from visitor spending surveys to improved
estimates of the numbers and types of visitors affected by
various actions. The system should also facilitate more
complete analyses of the local economic impacts of
recreation and tourism activity and the structural
characteristics of tourism-oriented economies.
REFERENCES
Alward, G.S., and Palmer, C.J. 1983. IMPLAN--An
input-output analysis system for forest service planning.
In Proceedings of the First North American Conference on
Forest Sector Modeling, ed. R. Sepalla, C. Row, and A.
Morgan. Oxford: AB Academic Publ.
Alward, G.S., and E.M. Lofting. 1985. Opportunities for
analyzing the economic impacts of recreation and tourism
expenditures using IMPLAN. Paper presented at 30th Annual
Regional Science Association Meeting, Philadelphia, PA,
Nov 14-16, 1985.
Archer, B.H. 1984. Economic impact: Misleading multiplier.
Annals of Tourism Research. 11: 517-518.
Bergstrom, J.C., H.K. Cordell, A.E. Watson, and G.A.
Ashley. 1990. Economic impacts of state parks on state
economies in the south. Southern J. of agricultural
Economics 22:69-77.
Brucker, S.M., S.E. Hastings, and W.R. Latham III. 1987.
Regional input-output analysis: A comparison of five
"ready-made" model systems. Review of Regional Studies
17(2): 1-25.
Cordell, H.K., J.C. Bergstrom, G. Ashley, and A. Watson.
1987. Report on estimates of economic impacts of proposed
recreational development at Land Between the Lakes. Final
report prepared for Tennessee Valley Authority. U.S.
Forest Services, Southeast Forest Experiment Station,
Athens, GA.
Fleming, W.R., and L. Toepper. 1990. Economic impact
studies: Relating the positive and negative impacts to
tourism development. Journal of Travel Research 28:35-42.
Getz, D. 1991. Assessing the economic impacts of festivals
and events: Research issues. Journal of Applied
Recreation Research 16(l): 61-77.
Jackson, R.S., D.J. Stynes, D.B. Propst, and L.E. Siverts.
1991. Economic impact analysis as a tool in recreation
program evaluation. Instruction Report R-91-_, Department
of the Army, Waterways Experiment Station. Vicksburg, MS.
(In press).
Johnson, R.L., F. Obermiller, and H. Radtke. 1989. The
economic impact of tourism sales. Journal of Leisure
Research 21(2): 140-154.
Peine, J.D., and J.R. Renfro. 1988. Visitor use patterns
at Great Smoky Mountains National Park. Research/Resources
Management Report SER-90. Atlanta: USDI, National Park
Service, Southeast Region.
Propst, D.B (Compiler). 1985. Assessing the economic
impacts of recreation and tourism. Asheville, NC:
Southeast Forest Experiment Station.
Propst, D.B. 1988. Use of IMPLAN with public area
recreation visitor survey (PARVS) pretest data: Findings
and recommendations. Technical Report R-88-1, Department
of the Army, Waterways Experiment Station. Vicksburg, MS.
Propst, D.B., D.J. Stynes, and J. H. Lee. 1991.
Sevelopment of spending profiles for recreation visitors
to Corps of Engineers projects. Technical Report R-91- I-.
Department of the Army, Waterways Experiment Station.
Vicksburg, MS.
Propst, D.B., D.J. Stynes, and H. Jiao. 1992. Development
of visitor spending profiles for the Upper Mississippi
River System. Final Report to U.S. Army Engineers,
Waterways Experiment Station. Department of Park and
Recreation Resources, Michigan State University, East
Lansing.
Strang, W.A. 1971. The use of market segmentation in
planning for a recreation-based economy. Madison:
University of Wisconsin Sea Grant Program.
Stynes, D.J., G.L. Brothers, D.F. Holecek, and D. Verbyla.
1983, Spending patterns and economic impacts of Michigan
registered boat owners. Ann Arbor: Michigan Sea Grant
Program research report MICHU-SG-83-210.
Stynes, D.J., and D.B. Propst. 1992. Analyzing impacts of
durable goods purchases in recreation and tourism studies.
Paper presented at 4th North American Symposium on Society
and Natural Resource Management. Madison, WI. May
17-20,1992.
Stynes, D.J. 1992. Visitor Spending and the Local Economy.
Report to Great Smoky Mountains National Park. Department
of Park and Recreation Resources, Michigan State
University, East Lansing.
Taylor, C., S. Winter, G. Alward, and E. Siverts. 1992.
Micro IMPLAN User's Guide. Fort Collins, CO: USDA Forest
Service, Land Management Planning Systems Group.
THE IMPACT OF TOURISM ON LOCAL GOVERNMENT PUBLIC SERVICE
EXPENDITURES
Mario F. Teisl, Research Associate
Stephen D. Reiling, Professor
Department of Agricultural and Resource Economics
University of Maine
I.INTRODUCTION
Background
Tourism is an important component of the Maine economy.
Tourists spent about $1.6 billion in Maine during 1987
(Center for Survey and Marketing Research 1988). These
expenditures supported over 58,000 full-time equivalent
jobs that paid about $470 million in wages. In addition, a
total of $85 million was collected by the state in the
form of sales and gasoline taxes paid by tourists, and
income taxes paid by people working in the tourism
industry. These numbers reflect only the direct or primary
impacts of tourism on the Maine economy. The impacts are
even larger when the indirect and induced effects of
tourist expenditures are included in the impact measures.
Nevertheless, the direct impacts alone provide a strong
justification for those who refer to Maine as
"vacationland."
Studies measuring the economic benefits of recreation and
tourism have become quite common in the last 20 years (for
example, see Chapters 9 and 10 in Napier 1981 and
Bergstrom et al. 1990). However, a complete economic
evaluation of tourism requires measurement of the economic
costs of tourism as well as the economic benefits.
Tourism-related costs can take several forms, including
traffic congestion, crowding at recreation and tourist
facilities, environmental degradation, higher housing
costs, and a general increase in the cost of living.
Tourism can change the character of tourist communities
and reduce the quality of life for residents.
Tourism also can increase the cost of providing public
services within communities. Like residents, tourists
"consume" public services, such as water and sewer
services, solid waste disposal services, and road services
provided by local governments. The types and quantities of
services consumed by tourists depend in part on the amount
of time they spend in the community and the types of
activities in which they participate. For example,
seasonal home owners consume a larger quantity and more
diversified mix of public services than tourists who spend
only a few hours at a tourist attraction within a
community.
Several studies have investigated the relationship between
tourism and local government expenditures. Barrows and
Nilsestuen (1974) studied the fiscal impacts of reservoir
development in the Kickapoo Valley in Wisconsin and
observed increases in local government expenditures for
new roads and road maintenance, police and fire
protection, and water, sewer and waste disposal
facilities. Young (1973) also reports that tourism
development increases the demand for local public services
and stresses existing delivery systems. Schultz and
Stronge (1980) note that tourism increased public service
costs in Florida to alleviate traffic congestion, higher
crime rates, and overcrowding of hospitals and other
health-care facilities. All of these studies suggest that
tourism growth and developmentaffect some of the public
services provided by local governments.
The American Society of Planning Officials (1976) examined
the effects of recreational second home development on
local government expenditures. Some of the fiscal impacts
identified include:
-increases in solid waste disposal costs because second
home owners generate more waste per household, and because
solid waste collection costs are higher for second homes
due to their remote location and seasonal variation in
service;
-increases in road maintenance costs because of the need
to improve roads and to snowplow roads that previously
were not plowed;
-increases in police, fire protection, and health-care
costs to provide services to seasonal residents;
retirement communities and skiing developments often
require expanded local health-care services;
-local education expenditures are usually unchanged by
second home development, unless second homes are converted
to year-'round use, or the recreational development causes
secondary population growth in the year-round population.
Dwyer and Espeseth (1977) studied four rural counties
bordering a newly constructed reservoir in Illinois. Eight
years after the reservoir was completed, they found a
large increase in the number of visitors to the site, but
no significant second home development. Local governments
experienced increases in road construction and maintenance
costs and law enforcement costs.
Finally, Cole and Kuserk (1984) analyzed the fiscal
impacts of four groups: permanent residents, seasonal
residents, overnight tourists, and day tourists on five
municipalities along the Delaware shore. Local government
revenues and expenditures attributable to each group were
measured. The results indicate that seasonal residents
contribute more to local government revenues than the cost
of providing them with services. This result reflects the
fact that seasonal residents only demand services during
part of the year, but pay the same property tax rate as
year-round residents. On the other hand, overnight
tourists and day tourists consume more in local government
services than they contribute to the towns in revenues.
One of the recommendations of the study is to initiate or
expand user fees for parking and beach use so that towns
can collect a greater share of the costs incurred to
provide services to overnight and day tourists.
Purpose of Report
The overall purpose of this report is to present the
results ofa study designed to determine how tourism
affects the level of public service expenditures made by
Maine communities. Because tourists consume public
services while visiting a town, the basic hypothesis of
this study is that per capita expenditures incurred by
towns to provide public services will be higher in those
towns that accommodate high levels of tourist activity.
This study attempts to test this hypothesis using data on
public service expenditures and tourism activity for 379
towns in Maine. The specific study objective is to develop
an expenditure determinants model to determine how
community characteristics, including tourism, affect per
capita expenditures incurred to provide public services.
This study only addresses one type of cost that tourism
may impose on the community: the costs of providing public
services. The study does not attempt to assess directly
the other types of social, economic, or environmental
costs that can accompany tourism development. However,
communities may respond to these costs by providing more
or better quality public services. These types of
responses would be reflected in higher public service
delivery costs, and therefore, may be partially accounted
for in the expenditure determinants equations estimated in
this study.
In addition, no attempt is made in this study to determine
how tourism activities contribute to the revenue base of
local communities. Because Maine towns are highly
dependent on property taxes as a revenue source,
tourist-related facilities (second homes, hotels, motels
and restaurants, and retail establishments that cater to
tourists), can contribute positively to the tax revenue
base of a town. This study, however, only examines
tourism's impact on public service expenditures.
Organization of the Report
Section II contains a discussion of the methodology used
in this study. Data collection and estimation procedures
are discussed in Section III, and the results are
presented in Section IV. Finally, the implications of the
results are discussed in Section V.
II.METHODOLOGY
Estimation of cost functions is a common practice in
economics. Cost functions show the relationship between
the cost and the quantity of the good or service provided
while holding constant the other factors that influence
costs:
Total Costs = f(Quantity) Quality, Price of Inputs, Other
Supply Factors)
Total costs are directly related to the quantity of the
good provided. That is, total costs increase (decrease) as
the level of output increases (decreases), other things
being equal. Hence, the total cost curve has a positive
slope over all ranges of output. However, average total
costs may decrease (increase) as the level of output
increases, due the presence of economies (diseconomies) of
size.
The quality of the good or service, the prices of inputs,
and all other factors that influence costs are constant
along a given total cost curve. If one or more of these
factors change, the entire cost curve will shift up or
down. For example, the total cost function would shift
upward as the quality of the good or service increases,
and as the prices of inputs used to produce the good or
service increases. Other factors also influence total
costs. In the case of local government services,
population density and the physical size of the community
are examples of factors that can affect the costs of
providing some public services. Changes in these
supply-condition variables also cause the entire total
cost curve to shift up or down.
Total cost functions for public services could be
estimated to determine the influence of tourism on
public service delivery costs. However, there are two
problems with this approach. First, the construction
of cost equations for publicly provided services is
problematic because public service output levels are
difficult to measure. For example, it is difficult to
quantify the "output" of fire protection services provided
by the town. Without acceptable measures of the quantity
of public services provided, the appropriate cost
functions can not be estimated.
Another problem with the cost function approach is that it
only represents one side of the market. Factors related to
the demand side of the market also influence the level of
public services provided by a community. For example,
communities with high per capita incomes often provide
more and/or higher quality services than communities with
low per capita incomes. Furthermore, tourists may increase
the demand for some public services. As a result, the
community may have to increase the quantity of service
provided which, in turn, will increase the total cost of
providing the service. Therefore, demand conditions have
to be included in an analysis of the factors that
influence the cost of providing local government services.
Ideally, one could estimate demand and cost functions
simultaneously to determinethe level of services provided
by the community. However, this approach also requires
adequate measure of the quantity of services provided.
Because these measures are not readily available, most
researchers have adopted the expenditure determinants
method to identify factors that influence local government
expenditures for public services (Hirsch 1970).
Expenditure determinants models were introduced in the
1940s and have been refined over the years. Currently,
expenditure determinants models are estimated using
multiple regression analysis. However, these models lack a
strong theoretical base because they incorporate both
supply and demand variables in the same equation.
Nevertheless, Hirsch (1970) has noted that these models
advance our understanding of the factors that cause
expenditure levels to vary among communities.
In its general form, the cost determinants model assumes
that per capita expenditures on public services depend on
several factors that affect the cost of providing a public
service (Hirsch 1970):
EXPENDITURES PER CAPITA = F(QUANTITY, QUALITY, INPUT
PRICES, OTHER SUPPLY CONDITIONS, DEMAND CONDITIONS).
Note that the expenditure determinants equation includes
the independent variables used in the cost
function approach, including quantity and quality measures
of the service provided, the prices of inputs used to
produce the services, and other supply factors or
conditions that affect public service expenditures.
Because of the lack of adequate measures of the quantity
and quality of output provided, proxy variables are used
to represent quantity and quality measures. For example,
the population of the town is often used to approximate
the quantity of the service provided, and the crime rate,
fire insurance rating, and student-teacher ratios are
examples of the variables used to reflect the quality of
some of the public services provided.
Supply conditions, such as population density, the degree
of urbanization in the area, and the level of non-local
government aid also influence public service expenditures.
For example, sewer collection costs are usually lower in
towns where the population is concentrated in a small
geographic area. Similarly, state aid for road maintenance
has a substantial impact on road maintenance expenditures
because the town is reimbursed for all or part of the
maintenance expenses.
Unlike the cost function approach, the expenditure
determinants equation also includes variables that reflect
demand conditions. These are included because they play an
important role in community decisions regarding the
quantity or level of public service to provide. Demand
conditions include per capita income levels, real property
values per capita, and the types of industries in the
community. For example, a manufacturing plant that
generates large quantities of waste water can increase the
level of expenditures for waste water treatment facilities
in the town. Similarly, the presence of tourism can
increase the demand for some types of public services.
Commuting patterns also influence the demand for public
services, as do the demographic characteristics of the
population. For example, retirement communities often
spend more per capita for health-related public services
than other communities. Similarly, communities comprised
of a large number of families with children often spend
more per pupil for education services.
Tourism's potential impact on public service expenditures
can enter the expenditure determinants model in a number
of ways. For example, tourism can be viewed as a demand
condition that influences public service costs. However,
some forms of tourism development can have an impact
through supply condition variables. Second home
developments are often located some distance from the
central core of the community, thereby increasing solid
waste collection, road maintenance, and other public
service costs. Furthermore, second home owners often have
different demographic characteristics than year-round
residents and may, therefore, demand a different type or
level of services (American Society of Planning Officials
1976).
Finally, it is important to recognize that several
different types of "tourists" exist, including seasonal
home residents, vacationers that spend one or more nights
in a community, and day visitors that spend only a few
hours at their destination. Clearly, different types of
tourists demand different types or quantities of public
services. As noted above, seasonal residents who live in
the community part of the year utilize a wider range of
services than visitors who only spend an afternoon at the
beach. Consequently, three types of tourists are included
in the study: seasonal residents, vacationers*, and day
tourists. The three types of tourists are included
separately in an attempt to measure the impact of each
type of tourist on local public service costs.
Unfortunately, data are unavailable on the number of
tourists who visit a community in Maine. Consequently,
proxy measures have to be used to represent the level of
tourist activity in Maine communities. The number of
seasonal residents is represented by the number of
seasonal homes in the community. The number of vacationing
tourists in a given community is represented by the
commercial lodging capacity (hotel, motel rooms and
campsites at commercial campgrounds) in the town. Finally,
the seating capacity of seasonal restaurants in the
community is used to represent the number of day visitors.
Although the number of seasonal homes and the commercial
lodging capacity are considered to be good proxies for
seasonal residents
----------------------------------
* For the purposes of this study, vacationers are defined
as tourists that spend at least one night in the
community, but do not own a seasonal home. Vacationers
typically use commercial lodging establishments for
overnight accommodations, but may also stay overnight with
friends or relatives.
and vacationing tourists, respectively, the use of seating
capacity at seasonal eating establishments is not
considered to be a good measure of day-trip visitors.
However, no other measure was available.
Separate expenditure equations are specified for each of
the major public services provided by Maine towns,
including police, fire, welfare, health, recreation,
sanitation, public works, general administration and
education. The expenditure equations for each public
service and the hypothesized sign on each independent
variable are presented below.
POLICE
POLICE EXPENDITURES PER CAPITA = f(POPULATION, POPULATION
SQUARED, MILES TO NEAREST SMA, CRIME RATE, CRIME CLEARANCE
RATE, PER CAPITA INCOME, PER CAPITA REAL PROPERTY VALUES,
POPULATION PER SQUARE MILE, IN COMMUTERS PER CAPITA, TOWN
IS ON ISLAND, PRESENCE OF TOWN POLICE DEPARTMENT, SEASONAL
HOMES PER CAPITA, PER CAPITA LODGING CAPACITY, AND PER
CAPITA SEASONAL RESTAURANT SEATING)
Population and population squared are included in this and
other equations to partially capture the level or quantity
of service provided. The population squared variable
reflects the total expenditure function's classical
properties and tests the hypothesis that per capita
expenditures are subject to economies or diseconomies with
respect to population size. Miles to nearest Standard
Metropolitan Area (S.M.A.) is an urbanization measure to
test the hypothesis that per capita police expenditures
increase with increasing urbanization (Borcherding and
Deacon 1972). The crime rate is used to reflect the demand
for police services, and the crime rate is used to reflect
the quality of police services (Walzer 1972). Both of
these variables should be positively related to per capita
police expenditures. However, the crime rate can also
reflect the quality or quantity of police services (i.e.,
a high crime rate may indicate a low quantity or quality
of police services); therefore, the crime rate may be
negatively related to per capita police expenditures. Per
capita income reflects the ability to pay for services and
should be positively related to per capita police
expenditures (Weicher 1970).
Per capita real property values reflect the available tax
base of the community and the value of assets requiring
police protection; it is hypothesized to be positively
related to per capita police expenditures. Population per
square mile reflects population density and should be
positively related to per capita police expenditures
(Weicher 1970), although Sacks and Harris (1964) found
density to be insignificant in explaining per capita
police expenditures. In-commuters reflect the commercial
and industrial activity in the town and should be
positively related to per capita police expenditures
(Muller 1975). "Town is on island" is a qualitative
variable to reflect the hypothesis that island towns (with
no physical connection to the mainland) have a different
public service cost structure. These towns may be more
self-sufficient in the provision of certain services
because they can not rely on other towns when additional
services are needed.
Presence of a town police department is a qualitative
variable to indicate whether the town has its own police
department. Some towns rely on police services provided by
the county or the state. Presence of a police department
should be positively related to per capita police
expenditures. The three tourism-related variables are
hypothesized to be positively related with per capita
police expenditures (Barrows and Nilsestuen 1974; Dwyer
and Espeseth 1977).
FIRE
FIRE EXPENDITURES PER CAPITA = f(POPULATION, POPULATION
SQUARED, ARSON PER CAPITA, FIRE PROTECTION RATING, REAL
PROPERTY VALUES PER CAPITA, PER CAPITA INCOME, POPULATION
PER SQUARE MILE, PERCENT REAL PROPERTY WHICH IS
INDUSTRIAL, TOWN IS ON ISLAND, SEASONAL HOMES PER CAPITA,
PER CAPITA LODGING CAPACITY, AND PER CAPITA SEASONAL
RESTAURANT SEATING)
Arson fires per capita is included to reflect the demand
for fire services and is hypothesized to be positively
related to per capita fire expenditures. The fire
protection rating is an insurance industry rating of fire
protection quality or effectiveness. Ratings range from
one to ten, with one being a superior rating; thus, there
should be a negative relationship between the fire
protection rating and per capita fire expenditures. Real
property values per capita proxies the amount of property
to be protected and is expected to be positively related
to per capita fire expenditures. Per capita income, a
demand shifter, should be positivelyrelated to per capita
fire expenditures (Borcherding and Deacon 1972).
Population per square mile is hypothesized to be
positively related to per capita fire expenditures (Fisher
1961) because higher density areas may require special
fire fighting equipment (i.e., ladder trucks for
multi-story structures) and "a high population density
will aggravate the danger from fire" (Weicher 1970:388).
Percentage of real property valuation that is industrial
reflects the importance of industrial activities in the
community and may identify special demand factors for fire
protection services (potential of chemical spills, fires,
etc.). Therefore, percent industrial should be positively
related to per capita fire expenditures. The island town
dummy variable is hypothesized to have a positive
relationship with per capita fire expenditures because
island towns need to be more self-sufficient in providing
fire protection. The tourism variables are hypothesized to
be positively related to per capita fire expenditures
(Barrows and Nilsestuen 1974).
WELFARE
WELFARE EXPENDITURES PER CAPITA = f(POPULATION, POPULATION
SQUARED, UN-EMPLOYMENT RATE, PER CAPITA INCOME, WELFARE
AID PER CAPITA, TOWN IS ON ISLAND, SEASONAL HOMES PER
CAPITA, PER CAPITA LODGING CAPACITY, AND PER CAPITA
SEASONAL RESTAURANT SEATING)
The unemployment rate represents a demand factor and is
hypothesized to be positively related to per capita
welfare expenditures. Per capita income should be
positively related to per capita welfare expenditures
(Sacks and Harris 1964). Non-local government aid for
general assistance lowers the local cost of providing
welfare services and, therefore, should be positively
related to per capita welfare expenditures. The tourism
variables are included in the equation to test the
hypothesis that they are insignificant in explaining per
capita welfare expenditures.
HEALTH
HEALTH EXPENDITURES PER CAPITA = f(POPULATION, POPULATION
SQUARED, PER CAPITA INCOME, FIRE PROTECTION RATING,
PERCENT OF POPULATION OVER 65, MILES TO NEAREST SMA,
PERCENT OF REAL PROPERTY WHICH IS INDUSTRIAL, TOWN IS ON
ISLAND, HOMES PER CAPITA, PER CAPITA LODGING CAPACITY, AND
PER CAPITA SEASONAL RESTAURANT SEATING)
Per capita income is expected to be positively related to
per capita health expenditures (Borcherding and Deacon
1972). The fire protection rating proxies as a quality
measure because it is hypothesized that the quality of the
local ambulance/rescue squad is closely related to the
quality of the local fire department*.
------------------------
*In Maine, the ambulance/rescue squad is usually operated
by the fire department.
The percentage of the town's population over 65 is used to
reflect demand for health services and should be
positively related to per capita health expenditures
(Gabler 1971). However, people over the age of 65 are
eligible to have their ambulance services paid for by
Medicare, whereas non-medicare recipients often do not pay
for ambulance services. Therefore, towns may subsidize the
ambulance squad to cover the operating costs not paid by
Medicare recipients. If this is the case, the percentage
of population over 65 may have a negative relationship
with per capita health expenditures.
Miles to nearest SMA proxies as an urbanization measure,
and it is hypothesized towns located farther away from
larger, well-equipped, urban health facilities need to
spend more on per capita health expenditures. Percent of
real property which is industrial reflects the demand for
special health services associated with industrial
accidents, and should be positively related with per
capita health expenditures. The island town dummy variable
is expected to be positively related to per capita health
expenditures because island towns need to be more
self-sufficient in providing health services. The tourism
variables are hypothesized to be positively related to per
capita health expenditures because seasonal residents and
special recreation facilities, such as skiing areas, may
strain local health services (American Society of Planning
Officials 1976).
RECREATION
RECREATION EXPENDITURES PER CAPITA = f(POPULATION,
POPULATION SQUARED, PER CAPITA INCOME, REAL PROPERTY
VALUES PER CAPITA, INCOMMUTERS PER CAPITA, OUTCOMMUTERS
PER CAPITA, MILES TO NEAREST SMA, TOWN IS ON ISLAND,
SEASONAL HOMES PER CAPITA, PER CAPITA LODGING CAPACITY,
AND PER CAPITA SEASONAL RESTAURANT SEATING).
Per capita income and real property values per capita
reflect ability to pay for recreation services and proxy
for effective demand; wealthier communities are
hypothesized to spend more per capita for recreation
(Borcherding and Deacon 1972). Incommuters and
outcommuters may demand recreation opportunities in
the towns in which they work; thus incommuters should be
positively related, and outcommuters should be negatively
related, to per capita recreation expenditures (Keeling
1986). Miles to nearest SMA is used to proxy the presence
of substitute recreational opportunities. Cities generally
have more recreation and entertainment opportunities that
can be used by neighboring communities; therefore, towns
located farther from cities may have to provide more
recreation opportunities.
It is hypothesized that seasonal restaurant seating
capacity is positively related to per capita recreation
expenditures because day-tourists increase the demand for
recreation opportunities and thereby increase town
recreation expenditures per capita (i.e., increased beach
and park maintenance). On the other hand, seasonal
residents are expected to demand fewer recreational
opportunities because seasonal homes are often
concentrated in locations (lake front, around ski areas,
etc.) where owners can provide their own recreation.
Commercial lodging capacity is hypothesized to be
unrelated to per capita recreation expenditures because
vacationers are usually attracted to a destination for
reasons unrelated to town-related recreational
opportunities.
SANITATION
SANITATION EXPENDITURES PER CAPITA = f(POPULATION,
POPULATION SQUARED, PERCENT CHANGE IN POPULATION, PER
CAPITA INCOME, POPULATION PER SQUARE MILE, INCOMMUTERS PER
CAPITA, PERCENT OF REAL PROPERTY WHICH IS INDUSTRIAL, TOWN
IS ON ISLAND, SEASONAL HOMES PER CAPITA, PER CAPITA
LODGING CAPACITY, AND PER CAPITA SEASONAL RESTAURANT
SEATING)
Sanitation services exhibit high capital costs and
short-run capacity constraints. Consequently, one would
expect a lag in the response of sanitation expenditures to
changes in population; therefore, we hypothesize a
negative relationship between sanitation services per
capita and the percent change in population. Per capita
income, a demand shifter, should be positively related
with per capita sanitation expenditures (Borcherding and
Deacon 1972). Population per square mile is included to
test the hypothesis that less densely populated towns
provide fewer sanitation services because of high
collection costs. An alternative hypothesis is that towns
with low population densities do provide sanitation
services and incur higher per capita sanitation
expenditures (Weicher 1970). Incommuters per capita should
be positively related to per capita sanitation
expenditures. Percentage of real property which is
industrial is included to test the hypothesis that towns
with large industries incur higher per capita sanitation
expenditures. However, industries often provide their own
sanitation services; therefore, percent industrial may be
insignificant in explaining per capita sanitation
expenditures. All of the tourism variables are
hypothesized to have a positive effect on per capita
sanitation expenditures (Barrows and Nilsestuen 1974).
PUBLIC WORKS
PUBLIC WORKS EXPENDITURES PER CAPITA = f(POPULATION,
POPULATION SQUARED, PER CAPITA INCOME, REAL PROPERTY
VALUES PER CAPITA, ROAD AID PER CAPITA, INCOMMUTERS PER
CAPITA, POPULATION PER SQUARE MILE, TOWN IS ON ISLAND,
SEASONAL HOMES PER CAPITA, PER CAPITA LODGING CAPACITY,
AND PER CAPITA SEASONAL RESTAURANT SEATING)
Per capita income and per capita real property values
should be positively related to per capita public works
expenditures (Ohls and Wales 1972). Non-local road aid per
capita also should be positively related to public works
expenditures per capita. Incommuters per capita reflect
the size of the commercial sector of the town and should
be positively related to per capita public works
expenditures (Muller 1975). Population per square mile is
a density factor and should be negatively related to per
capita public works expenditures. Per capita seasonal
homes, per capita lodging capacity, and per capita
seasonal restaurant seating are expected to be positively
related to public works expenditures because of increased
demand for road maintenance (Dwyer and Espeseth 1977).
ADMINISTRATION
ADMINISTRATION EXPENDITURES PER CAPITA = f(TOTAL
EXPENDITURES, TOTAL EXPENDITURES SQUARED, TOTAL NON-LOCAL
AID PER CAPITA, PER CAPITA INCOME, REAL PROPERTY VALUES
PER CAPITA, OUTCOMMUTERS PER CAPITA, TOWN IS ON ISLAND,
SEASONAL HOMES PER CAPITA, PER CAPITA LODGING CAPACITY,
AND PER CAPITA SEASONAL RESTAURANT SEATING)
Administrative expenditures per capita are not directly
tied to the provision of any particular public service,
but are a positive function of the level of all other
public service expenditures (not including administrative
expenditures). The squaring of total expenditures tests
the hypothesis that administration expenditures per capita
are subject to economies or diseconomies with respect to
the level of town expenditures. Total non-local aid per
capita should be positively related to the level of
administration expenditures per capita (Keeling 1986). Per
capita income and real property values per capita should
be positively related to administration expenditures
because wealthier communities demand higher levels of
spending and therefore, more administrative services. The
number of outcommuters per capita is included to test the
hypothesis that communities with a high number of
outcommuters per capita provide a higher level of public
services per capita. Per capita seasonal homes, per capita
lodging, and seasonal restaurant seating are expected to
be positively related to per capita administration
expenditures because of increased demand for
administrative services.
EDUCATION
PER PUPIL EDUCATION EXPENDITURES = f(ENROLLMENT,
ENROLLMENT SQUARED, PERCENTAGE CHANGE IN ENROLLMENT,
TEACHER/PUPIL RATIO, PER CAPITA INCOME, REAL PROPERTY
VALUES PER CAPITA, EDUCATION AID PER CAPITA, EMPLOYMENT
PER CAPITA, TOWN IS ON ISLAND, SEASONAL HOMES PER CAPITA,
PER CAPITA LODGING CAPACITY, AND PER CAPITA SEASONAL
RESTAURANT SEATING)
Enrollment and enrollment squared test the hypothesis that
per pupil education expenditures are subject to economies
or diseconomies with respect to enrollment. However, the
Maine State Government attempts to equalize per student
education spending between towns. Therefore, the
enrollment variables may be insignificant in explaining
per pupil education expenditures. Percentage change in
enrollment should be negatively related to education
expenditures per pupil because teaching positions and
education administration do not vary in the short run
(Keeling 1986). The teacher/pupil ratio is a proxy for
quality of education and should be positively related to
per pupil education expenditures. Per capita income and
real property values per capita should be positively
related to per capita education expenditures because
wealthier communities often spend more per student for
educational services. However, if the state subsidy for
education works to equalize per pupil spending, then the
community's level of wealth and income may be
insignificant in explaining per pupil expenditures.
Non-local (state) education aid should be positively
related to per capita education expenditures. Employment
per capita is hypothesized to be positively related to the
level of education expenditures because businesses demand
better-educated workers (Keeling 1986). The island town
dummy variable is hypothesized to be positively related to
per capita education expenditures to reflect the increased
cost of tuitioning out students to schools on the
mainland. The tourism variables are expected to be
insignificant in the education equation.
III. DATA COLLECTION AND ESTIMATION PROCEDURES
Actual 1985 expenditures for the public services discussed
above were collected for 379 towns in Maine from the
annual reports published by the towns. Although an attempt
was made to obtain the data for all incorporated towns in
the state, the data were not available for some
communities. However, the 379 towns are representative of
all towns in Maine in terms of size distribution (Teisl
1990).
The expenditures obtained from the annual reports are
divided by the town's population in 1985 to form the
dependent variables in the expenditure determinants model.
Police expenditures represent all expenditures made for
police protection and animal control, while fire
expenditures represent all expenses for fire protection,
including fire control, hydrant rental, and
dispatch/communication costs. Welfare expenditures include
general assistance aid and town contributions to social
service groups, such as community action programs, senior
citizen programs, and family crisis services.
Health expenditures represent local ambulance/rescue squad
costs and local government contributions to alcohol and
mental health counseling services; recreation expenditures
are for community athletic programs and maintenance
oflocal parks and recreational facilities. Sanitation
expenses include solid waste collection, transfer stations
and recycling expenditures, and sewer collection and
treatment expenses. Although we would have preferred to
estimate separate equations for solid waste disposal and
sewer treatment services, many towns do not report costs
separately for these two services. Public works
expenditures primarily represent maintenance costs for
roads and other transportation facilities, but also
include streetlighting and cemetery maintenance for some
towns. General administration expenses include code
enforcement, regional planning, legal, financial and
informational services, and other general administration
tasks. Finally, education expenses include both public
school and adult education costs.
Data representing the independent variables in the
equations were collected from various secondary sources
for each of the 379 towns in Maine. Sources include Census
data, the Maine Department of Labor, Maine Department of
Public Safety, Maine State Planning Office, the Maine
Bureau of Insurance, and the Maine Department of
Education. Seasonal home data at the community level were
obtained from previous research (Reiling and Cook 1985)
and from Census data. Lodging and seasonal restaurant
seating capacity were collected from records provided by
the Maine Department of Human Services. The reader is
referred to Teisl (1990) for a more complete discussion of
the sources from which the independent variables were
obtained.
Once the data were obtained, the equations presented above
were estimated using Zellner's Seemingly Unrelated
Regression (SUR) technique. This technique is used because
the error terms in the nine expenditure equations may be
correlated with each other for two reasons. First,
expenditure levels for different public services are
interdependent because their sum is restricted by the
total revenue available to cover public service costs.
Hence, a higher expenditure level for one service may
force a decrease in spending on other services. Second,
there may be spill-over effects between different service
areas provided by local government. For example, a
decrease in police expenditures may cause an increase in
vandalism, which may require increased maintenance
expenditures by some other town departments such as public
works and recreation.
Correlation among the error terms causes inefficient
parameter estimates if the equations are estimated using
Ordinary Least Squares techniques. The SUR method gives
maximum likelihood estimators that possess the minimum
variance within the class of all unbiased estimators
(Judge et al. 1988). The reader is referred to Theil
(1971) and Judge et al.(1988) for a more complete
discussion of SUR.
IV.RESULTS
The results of the research are reported in this section.
Each of the estimated equations are presented and
discussed in terms of the hypotheses identified earlier.
Particular attention is paid to the tourism-related
variables and their effect on public service expenditures
per capita.
Before presenting the specific results, it should be noted
that the data set was examined for heteroskedasticity and
multicollinearity. Although the former does not exist, the
degree of multicollinearity is quite high among the
following two sets of independent variables: population,
population squared, and population per square mile; and
per capita real property values, seasonal homes per
capita, and seasonal seating capacity per capita. The high
correlation between two of the three variables
representing tourism is especially troublesome for the
purposes of this study and should be kept in mind when
interpreting the results presented below. The high
correlation may result in the correlated variables being
insignificant in the estimated equations that contain two
or more of the correlated variables.
The weighted coefficient of determination or R2 for the
system of equations estimated using the SUR technique is
0.63. That is, the independent variables in the set of
equations account for 63 percent of the variation in the
level of public service expenditures among the 379 towns.
The F-statistic for the equation system (101 variables and
3309 degrees of freedom) is equal to 1.36, which is
statistically significant at the one percent level. Both
of these statistics indicate that the equations explain a
significant amount of the variation in public service
expenditures among towns. The estimated equations for each
of the nine public services are presented below and
discussed briefly.
Police
The estimated expenditure determinants equation for police
services is presented in Table 1. The insignificant
coefficients for population and population squared
indicate per capita police expenditures do not vary with
respect to population size.*
Among the statistically significant variables in the
equation, per capita income, which represents a demand
shifter, is positively related to capita police
expenditures. Per capita real property values, which
reflects the available tax base of the community and the
value of assets requiring police protection, also is
positively related to per capita police expenditures.
Population density and incommuters per capita are also
positively related to per capita police expenditures, as
hypothesized. Presence of a town police department
indicates per capita police expenditures are higher for
towns that have their own police department.
Miles to nearest S.M.A. and the crime rate, included to
reflect the demand for police services, are not
statistically significant in the equation. The crime
clearance rate, which was included to reflect the quality
of police services, also is insignificant. However,
presence of a police department is significant and may
reflect quality variations. In earlier estimates, the
crime clearance rate was positive and significant when the
police department dummy variable was not included in the
police equation. Town on an island is not significant in
explaining per capita police expenditures.
-------------------------
* The population ranges over which anequation exhibits
economies or diseconomies of size with respect to
population were calculated by determing the population
level at which the partial derivative of the quation with
respect to population is equal to zero.
Table 1. S.U.R. regression results for police expenditures
per capita.
(Vis. 21)
* Denotes significance at the 10 percent level.
Per capita seasonal homes and per capita seasonal
restaurant seating capacity have negative and significant
relationships with per capita police expenditures.
However, lodging capacity is significant and positively
related to per capita police expenditures. The negative
signs on the seasonal home and seasonal restaurant seating
capacity variables were not expected. It was hypothesized
that all three tourism variables would be significant and
would have positive signs in the police expenditure
equation.
Fire
The estimated equation for fire services is reported in
Table 2. The statistically insignificant coefficients for
population and population squared suggest that per capita
fire expenditures do not vary across towns with different
population levels.
Statistically significant variables in the equation
include per capita real property values, the fire
insurance rating, percent industrial, and the island town
dummy variable. Per capita real property values, included
as a demand shifter to signify the amount of property to
be protected, is positively related with per capita fire
expenditures. Per capita fire expenditures also increase
significantly as the quality of fire protection, as
measured by the fire insurance rating, increases. Towns
with proportionally large amounts of industrial property
also incur higher per capita fire expenditures. Finally,
island towns, which have to be more self-reliant in fire
suppression services, also incur higher costs per capita
in providing fire protection.
Table 2. S.U.R. regression results for fire expenditures
per capita.
(Vis. 22)
* Denotes significance at the 10 percent level
Arson per capita, per capita income, and population
density, included to reflect demand factors for fire
protection, are not statistically significant in the
equation. However, other demand shifting variables,
including real property per capita and percent of property
that is industrial, are significant.
The tourism variables are not significant in explaining
per capita fire expenditures. The insignificance of per
capita seasonal homes and per capita lodging capacity was
not expected. It was hypothesized that these variables
would be significant and positively related to per capita
fire expenditures because of the increase in the number of
dwellings and businesses to be protected. The high
correlation between some of the tourism variables and per
capita real property values may contribute to the
insignificance of the tourism variables in the equation.
However, detailed discussions with town officials in Bar
Harbor suggest that tourism has very little impact on town
expenditures for fire services.
Welfare
The SUR results for the welfare equation are reported in
Table 3. The positive coefficient for population and the
insignificant coefficient for population squared indicate
per capita welfare expenditures increase linearly with
population and therefore, exhibit diseconomies of size
with respect to population.
The unemployment rate and welfare aid per capita are
significant. The unemployment rate represents a demand
factor of per capita welfare expenditures. Per capita
welfare expenditures increase by $ 0.98 for every dollar
increase in general assistance aid received from the state
or federal government. This result suggests that welfare
expenditures for general assistance are strongly
influenced by the aid received by the towns.
Per capita income, included to reflect demand, is not
significant in the welfare equation. Furthermore, there is
no statistically significant difference in per capita
welfare expenditures for island and mainland towns.
The three tourism variables of per capita seasonal homes,
per capita lodging capacity, and per capita seasonal
restaurant seating are, as hypothesized, not statistically
significant in explaining per capita welfare spending.
Table 3. S.U.R. regression results for welfare
expenditures per capita.
(Vis. 23)
* Denotes significance at the 10 percent level
Health
The estimated health expenditures equation is reported in
Table 4. The coefficients for population and population
squared indicate per capita health expenditures are
subject to economies of size for towns with populations
less than 27,250, and exhibit diseconomies of size for
larger towns.
Significant variables in the equation include the fire
insurance rating, the percentage of the population over 65
years of age, miles to S.M.A., percent industrial, and the
island town dummy variable. The fire insurance rating
indicates per capita health expenditures increase with
increasing quality.* As the percentage of the town's
population over 65 increases, there is a decrease in per
capita health expenditures. This seems to confirm that
towns subsidize the operating costs of ambulance and
rescue services provided to non-medicare recipients. Towns
located farther from urban health facilities spend more on
per capita health expenditures. This is as hypothesized
because towns located further away from well-equipped
health facilities need to spend more on equipping their
local ambulance/rescue squads and have higher
transportation costs. Rural communities also may have
---------------------------
* Recall that ambulance/rescue squad services are usually
attached to the fire department. Therefore, the fire
insurance rating is used as an indicator of quality for
health services.
higher rescue squad needs. Industrial property as a
percentage of total property, included to reflect special
demand conditions, increase per capita health
expenditures. Islands towns incur higher per capita costs
of providing health services because these towns have to
be more self-reliant in terms of providing health care
than mainland towns. Per capita income, included to
reflect demand, is not significant in explaining per
capita health expenditures.
Seasonal homes per capita and per capita lodging capacity
are not significant in explaining per capita health
expenditures. Per capita seasonal restaurant seating is
positively related to per capita health expenditures.
Analysis of the data suggests that the seasonal seating
variable reflects the demand for local health services by
skiers who use ski areas located in some Maine towns.
Table 4. S.U.R. regression results for health
expenditures per capita.
(Vis. 24)
----------------------------
* Denotes significance at the 10 percent level
Recreation
The SUR results for the recreation equation are reported
in Table 5. The coefficients for population and population
squared indicate per capita recreation expenditures are
subject to diseconomies of size for towns with a
population of less than 37,203 and exhibit economies of
size for larger towns.
Real property values per capita, reflecting demand
factors, is positively related to per capita recreation
expenditures. The signs of the coefficients for
incommuters per capita and outcommuters per capita are as
hypothesized, although the outcommuters variable is not
significant in explaining per capita recreation
expenditures. This indicates incommuters' increased
awareness of recreation opportunities near their worksite
increases the demand, thus increasing per capita
recreation expenditures. This increased demand comes from
a group that contributes little to town revenues, thus
placing the fiscal impacts on local residents. Towns
located near urban areas have lower per capita recreation
expenditures, as hypothesized. Per capita income, which
reflects demand, is not significant in explaining per
capita recreation expenditures.
Table 5. S.U.R. regression results for recreation
expenditures per capita.
(Vis. 25)
--------------------------
* Denotes significance at the 10 percent level
Seasonal homes per capita is significant and negatively
related to recreation expenditures per capita.
This seems reasonable because seasonal home-owners often
have access to recreational resources at their doorstep.
Per capita lodging capacity is insignificant, perhaps
indicating that tourists staying in commercial lodging
establishments use recreation facilities provided by the
private sector or other levels of government. Per capita
recreation expenditures increase with increasing per
capita seasonal restaurant seating. This may indicate
towns with large day-trip tourist populations incur higher
maintenance costs for parks and public beaches.
Sanitation
The regression results for the sanitation equation are
reported in Table 6. The insignificant coefficient for
population and positive coefficient for population squared
indicate per capita sanitation expenditures are subject to
diseconomies of size for towns of all sizes.
Among significant variables, per capita income, included
as a demand shifter, positively affects per capita
sanitation expenditures. Incommuters per capita,
reflecting demand from the commercial sector, indicates
per capita sanitation expenditures increase with
increasing commercial activity.
Percentage change in population, used to test the
hypothesis that there is a lag in the response of
sanitation expenditures to a change in population, has the
correct sign but is not significant. Population density,
included to test the hypothesis that less densely
populated towns provide fewer sanitation services, is not
significant. Percent industrial is not significant in
explaining per capita sanitation expenditures, indicating
industries may often provide their own sanitation
services. Island towns are not significantly different in
their per capita sanitation expenditures.
Per capita seasonal homes and per capita lodging capacity
both have positive and significant effects on per capita
sanitation expenditures. However, per capita seasonal
restaurant seating capacity is not significant in the
equation.
Table 6. S.U.R. regression results for sanitation
expenditures per capita.
(Vis. 26)
---------------------------
* Denotes significance at the 10 percent level
Public Works
The regression results for the public works equation are
reported in Table 7. The insignificant coefficients for
population and population squared indicate per capita
public works expenditures are invariant with respect to
population.
Significant variables include per capita income, per
capita real property values, road aid per capita, and the
island town dummy variable. Per capita public works
expenditures increase with increases in per capita income.
However, per capita real property values are negatively
related to per capita public works expenditures. This
result is contrary to our hypothesis, but can be explained
by analyzing the methods used by the Maine Department of
Transportation to allocate per capita road aid. One of the
most important factors considered in the allocation of
road aid is real property values per capita to reflect the
town's own ability to pay for road maintenance. Towns with
higher per capita real property values receive less road
aid. Therefore, the negative relationship between per
capita real property and per capita public works is not
surprising.
Road aid per capita increases per capita public works
expenditures by $2.02. Island towns incur higher costs of
providing public works which may be caused by high
transport costs of materials and higher labor costs.
Incommuters per capita and population density, included to
reflect special demand conditions, are not significant in
explaining per capita public works expenditures. However,
other demand variables are significant.
Per capita seasonal homes and per capita seasonal
restaurant seating capacity are significant and positively
related to public works expenditures, as hypothesized.
However, the per capita lodging capacity variable is not
significant.
Table 7. S.U.R. regression results for public works
expenditures per capita.
(Vis. 27)
----------------------
* Denotes significance at the 10 percent level
Administration
The regression results for the administration equation are
reported in Table 8. The coefficients for total
expenditures and total expenditures squared indicate per
capita administration expenditures are subject to
diseconomies of size for all expenditures levels.
Administration expenditures per capita increase with
increased levels of non-local aid per capita. Island towns
face higher costs per capita of providing administration
than mainland towns.
Per capita income, per capita real property values, and
the number of outcommuters per capita are not significant
in explaining per capita administration expenditures.
Per capita seasonal homes and per capita seasonal
restaurant seating are positively related to per capita
administration expenditures, indicating increased
expenditures for administrative services per capita as the
tourism variables increase. Per capita lodging capacity is
positive, but insignificant in explaining per capita
administration expenditures.
Table 8. S.U.R. regression results for administration
expenditures per capita.
(Vis. 28)
--------------------------------
* Denotes significance at the 10 percent level
Education
The regression results for the education equation are
reported in Table 9. Enrollment and enrollment squared are
not significant confirming the hypothesis that enrollment
does not affect per pupil education expenditures because
of state programs to equalize per pupil education
expenditures among towns.
Percentage change in enrollment indicates teaching
positions and education administration are fixed in the
short run, thus causing a lag in per pupil education
expenditures. Per pupil education expenditures increase
with increases in the teacher/pupil ratio, a common
measure of education quality. Per capita aid to education
also is significant and positively related to per capita
education expenditures.
Per capita income and real property values per capita are
not significant in explaining per pupil education
expenditures, indicating the state subsidy for education
works well to equalize per pupil spending across towns
with different per capita income and per capita real
property values. Employment per capita, included to
reflect business demand for skilled workers, is not
significant in explaining per pupil education
expenditures. The three tourism-related variables are also
insignificant, as hypothesized.
Table 9. S.U.R. regression results for education
expenditures per capita.
(Vis. 29)
----------------------------
* Denotes significance at the 10 percent level
V. SUMMARY AND DISCUSSION
The overall objective of this study is to determine,
through the use of expenditure determinants analysis, the
impact that tourism has on the costs incurred by Maine
towns to provide public services. Although the theoretical
framework for expenditure determinants analysis is rather
weak, it is used in this initial attempt to measure the
tourism's impact on local public service provision costs.
Tourism's impact is assessed through the use of three
variables in the expenditure determinants regression
equations that approximate the level of tourism in the 379
Maine towns included in the study. The number of seasonal
homes per capita is used to approximate the number of
seasonal residents, and the commercial lodging capacity
(hotel and motel rooms and campsites at campgrounds) is
used to measure overnight or vacation tourists staying in
a town. Finally, the seating capacity at seasonal
restaurants is used to represent the number of day-trip
tourists that visit a town. Other variables are included
in the regression equations to represent the quantity and
quality of the public services provided, and to account
for other demand and supply conditions that are
hypothesized to influence public service costs.
The performance of the tourism variables varies among the
nine public service equations estimated. None of the
tourism variables are statistically significant in three
of the equations: fire, welfare and education expenditures
per capita. The insignificance of the tourism variables in
the latter two equations are consistent with prior
expectations. However, in a more detailed case study of
the town of Bar Harbor, one of the leading tourist towns
in Maine, it was discovered that tourism can have an
impact on education expenditures (Reiling and Deller
1991). Seasonal homes and other tourist-related commercial
establishments increase per capita property values for the
town. Higher per capita property values reduce the level
of state funding for education received by the community,
thereby requiring an increase in the local community's
contribution to education funding. Hence, although the
results of the education equation are consistent with
expectations, tourism may have some impact on education
costs paid by the town. That is, even though total
expenditures per pupil are the same in tourist and
non-tourist towns, towns with a large tourism sector may
pay a larger share of the total costs of education than do
towns that do not have a large tourism sector. This impact
needs to be studied further.
The insignificance of the tourism variables in the fire
expenditures equation is not consistent with expectations.
We had hypothesized that the presence of seasonal home and
other tourist-related structures would result in higher
per capita fire expenditures. However, fire department
personnel in Bar Harbor indicated that tourism did not
significantly affect per capita fire expenditures.
Although there are many tourist-related structures in the
town, other types of businesses and structures would be
present if tourism did not exist, and some of these other
types of businesses or structures may be more prone to
fire than the tourism businesses/structures. Hence, this
is another issue that should be examined in more detail.
In the police expenditures equation, seasonal homes per
capita and seasonal restaurant seating capacity are
significant and negative; however, lodging capacity per
capita is significant and positive. Our hypothesis was
that all three tourism-related variables would be positive
and significant. We have no explanation for actual
results. Police personnel in Bar Harbor indicated that
tourism increases police expenditures per capita because
of the need for overtime and additional seasonal workers
during the peak tourist season. However, they could not
separate the effects caused by different types of
tourists.
Two of the three tourism variables are positive and
significant in the sanitation, public works, and
administration equations. The other tourism variable is
statistically insignificant in these equations. Hence, the
tourism variables generally conform to expectations in
these equations. Only one of the tourism-related variables
(seasonal seating capacity per capita) is significant in
the health equation. All of the tourism-related variables
are significant and have the hypothesized signs in the
recreation expenditures equation.
At least one of the three tourism-related variables is
positive and significant in six of the nine equations
estimated: police, health, recreation, sanitation, public
works, and administration. These results suggest that
tourism increases per capita expenditures for these six
public services. However, these six public services only
account for 30 percent of the average budget of Maine
towns. The three public service equations in which none of
the tourism variables are significant (fire, welfare, and
education) account for the remaining 70 percent of town
budgets, with education accounting for 65 percent of town
expenditures. Hence, even though the equations suggest
that six of nine public services are positively affected
by tourism, the magnitude of the impact may not be large
relative to the total expenditures of town for providing
public services. This conclusion is supported by the case
study of Bar Harbor (Reiling and Deller, 1991). Public
service expenditures associated with tourism in Bar Harbor
constitute only 10.7 percent of the total municipal
budget, excluding education.
Additional work is needed to refine the techniques used to
estimate the impact of tourism on local public service
expenditures. One area that deserves additional attention
is the search for a better proxy variable of day-trip
tourism. The seasonal restaurant seating capacity used in
this study is not considered to be a good representation
of day-trip tourism. Its high correlation with the number
of seasonal homes suggests that the two variables are
measuring the same effect, and that the seasonal
restaurants may, to a large degree, serve seasonal
residents.
Overall, we believe the expenditure determinants equations
provide a useful initial investigation of the impact of
tourism on pubic service delivery costs. Although
additional refinement is needed, the results are fairly
consistent with expectations and provide an initial
analysis of the impact of tourism on public service costs.
However, it is also recognized that the theoretical
justification for the expenditure determinants approach is
weak. A more acceptable approach would be to develop a
simultaneous equation system model that examines both
expenditures and revenues. This approach not only has a
better theoretical base, it also considers both the
revenue and expenditure sides of the fiscal activities of
local government. Further research should focus on the
simultaneous equation approach.
REFERENCES
American Society of Planning Officials. 1976. Subdividing
rural America: Impacts ofrecreational lot and second home
developnwnt. Washington, DC: U.S. Government Printing
Office. 139 pp.
Barrows, R., and W. Nilsestuen. 1974. Recreation and
economic development in the Kickapoo Valley. Staff Paper
No. 70, University of Wisconsin Extension Service,
Madison, Wisconsin.
Bergstrom, J.C., H.K. Cordell, G.A. Ashley and A.E.Watson.
1990. Economic impacts of recreational spending on rural
areas. Economic Development Quarterly 4(l): 29-39.
Borcherding, T.E., and R.T. Deacon. 1972. The demand for
the services of non-federal government. American Economic
Review 62(Dec): 891-901.
Center for Survey and Marketing Research. 1988. Maine
tourism study: 1986-87 economic analysis. University of
Wisconsin-Parkside, Kenosha, Wisconsin (Prepared for the
Maine State Development Office) 73 pp.
Clayton, K.C., and D. Vvrhittington. 1977. The economics
of community growth: An impact model. Southern Journal of
agricultural Economics 9(July): 63-69.
Cole, G.L., and G.J. Kuserk. 1984. Impact of tourism
on public services in selected Delaware Shore communities,
Agricultural Experiment Station Bulletin No. 456,
University of Delaware, Newark. 48 pp.
Dwyer, J.F., and R.D. Espeseth. 1977. Improved local
planning for reservoir-oriented recreation opportunities.
Research Report No. 123, Departments of Forestry and
Leisure Studies, University of Illinois at
Urbana-Champaign.
Fisher, G.W. 1961. Determinants of state and local
government expenditures: A preliminary analysis. National
Tax Journal 14(Dec.): 349-355.
Gabler, L.R. 1971. Population size as a determinant of
city expenditures and employment-Some further evidence.
Land Economics 47(May): 130-138.
Hirsch, W.Z. 1970. The economics of state and local
government. New York: McGraw Hill. 333 pp.
Judge, G.G., R.C. Hill, W.E. Griffiths, H. Lutkepohl,
and T-C. Lee. 1988. Introduction to the theory and
practice of econometrics. New York: John Willey and Sons.
1024 pp.
Keeling, J.R. 1986. A fiscal impact model for Virginia
localities, M.S. Thesis, Department of Agricultural
Economics, Virginia Polytechnic Institute, Blacksburg. 300
pp.
Muller, T. 1975. Fiscal impacts of land development: A
critique of methods and a review of issues. Washington,
DC: The Urban League. 93 pp.
Napier, T.L. 1981. Outdoor recreation planning,
perspectives, and research. Dubuque, IA: Kendall/Hunt. 208
pp.
Ohls, J.C., and T.J. Wales. 1972. Supply and demand for
state and local services. Review of economics and
Statistics 54(Nov): 424-430.
Reiling, S.D., and C.M. Cook. 1985. Selected economic
characteristics of recreational seasonal homes in Maine.
ARE Staff Paper No. 375, Department of Agricultural and
Resource Economics, University of Maine, Orono, December,
13 pp.
Reiling, S., and S. Deller. 1991. The impact of tourism on
the fiscal capacity of Bar Harbor. Report prepared for the
Town of Bar Harbor, Department of Agricultural and
Resource Economics, University of Maine, Orono, 26 pp.
Sacks, S., and R. Harris. 1964. Determinants of
state-local government expenditures and intergovernmental
flows of funds. National Tax Journal 17(March): 75-85.
Schultz, R.R., and W.B. Stronge. 1980.Tourism's impact
is not all positive. Business and Economic Dimensions
16(2): 14-21.
Teisl, M.F. 1990. An analysis of the effects of tourism on
local government expenditures. M.S.Thesis, Department of
Agricultural and Resource Economics, University of Maine,
Orono, 126 pp.
Theil, H. 1971. Principles of econometrics. New York:
John Willey and Son. 736 pp.
Walzer, N. 1972. Economies of scale and municipal police
services: The Illinois experience. Review of economics and
Statistics 54(Nov.): 431-438.
Weicher, J.C. 1970. Determinants of central city
expenditures: Some overlooked factors and problems.
National Tax Journal 23(Dec,): 379-396.
Young, G. 1973. Tourism: Blessing or blight. Baltimore:
Penguin Books. 212 pp.
This information is for educational purposes only. References
to commercial products or trade names does not imply
endorsement by MSU Extension or bias against those not
mentioned. This information becomes public property upon
publication and may be printed verbatim with credit to MSU
Extension. Reprinting cannot be used to endorse or advertise
a commercial product or company.
This file was generated from data base TD on 09/30/03.
Data base TD was last revised on 06/06/02.
For more information about this data base or its contents please contact
alexande@msue.msu.edu . Please read our
disclaimer for important
information about using our site.
