The Community Policy Analysis System (COMPAS): A Proposed National Network of Econometric Community Impact Models

Abstract

Devolution of authority and responsibility from the Federal Government to state and local governments is, and will continue to be, one of the most dominant public policy issue for communities for the next decade. Block grants, deregulation, welfare reform, health care reform, education reform, agricultural policy reform, various state waivers, and other terms fill the national policy dialogue and all are symptomatic of devolution. To communities, especially rural communities, devolution spells the end of many of the safety nets that protected local governments, school districts and other public entities from some economic and social hardships. At the same time devolution enhances opportunities for local leadership and increases the returns to aggressive and innovative public decision making. In this environment, the value of economic and social information, accurate projections and analyses of policy alternatives is particularly great. This in turn is creating an opportunity for those involved in the decision support sciences. The Community Policy Analysis System (COMPAS) initiative is a response to this opportunity. It addresses the information needs of policy makers at the Federal, state and local levels. At the Federal level, there is a growing need for a better understanding of the local consequences of federal policy, especially policy that devolves responsibility to local governments. Similarly, state governments require information on the consequences of their policies on local governments as both state and local responsibilities change.

Full text

1
THE COMMUNITY POLICY ANALYSIS SYSTEM (COMPAS):
A PROPOSED NATIONAL NETWORK
OF ECONOMETRIC COMMUNITY IMPACT MODELS
Thomas G. Johnson
James K. Scott
September 1997
Community Policy Analysis Center
University of Missouri-Columbia
Devolution of authority and responsibility from the
Federal Government to state and local governments is,
and will continue to be, one of the most dominant
public policy issue for communities for the next decade.
Block grants, deregulation, welfare reform, health care
reform, education reform, agricultural policy reform,
various state waivers, and other terms fill the national
policy dialogue and all are symptomatic of devolution.
To communities, especially rural communities,
devolution spells the end of many of the safety nets that
protected local governments, school districts and other
public entities from some economic and social
hardships. At the same time devolution enhances
opportunities for local leadership and increases the
returns to aggressive and innovative public decision
making. In this environment, the value of economic and
social information, accurate projections and analyses of
policy alternatives is particularly great. This in turn is
creating an opportunity for those involved in the
decision support sciences.
The Community Policy Analysis System (COMPAS)
initiative is a response to this opportunity. It addresses
the information needs of policy makers at the Federal,
state and local levels. At the Federal level, there is a
growing need for a better understanding of the local
consequences of federal policy, especially policy that
devolves responsibility to local governments.
Similarly, state governments require information on the
consequences of their policies on local governments as
both state and local responsibilities change.
The need, under these emerging circumstances, for
better decision support at the local level is obvious.
The diversity of conditions in rural communities means
that generic, or aggregated decision support tools
probably conceal more than they reveal. Broad
generalizations about policy impacts are usually
uninformative at best, misleading at worst. It is clear,
for example, that to conclude that trade liberalization
will lead to overall increases in income and
employment is an important aggregate projection but it
tells us little about the changes that will be experienced
by individual communities or what their optimum
responses to these changes might be.
In response to these policy trends, a group of regional
economists and rural social scientists have identified a
set of modeling tools which can be used to provide
policy decision support for state and local government
officials, including input-output modeling, cost/benefit
analysis, and industrial targeting. In addition, the group
has developed a plan to build a collaborative
community policy analysis network that will eventually
extend to selected rural communities in twenty-five
states. With initial support from the Rural Policy
Research Institute (RUPRI), the four regional rural
development centers and a variety of other sources, the
group has also outlined the structure of econometric
community models for each state that will compare the
economic, demographic, and fiscal impacts of a variety
of economic or policy scenarios. The models are
intended to be used in conjunction with other decision
tools to provide maximal flexibility and a capacity for
rapid response to queries by local and state policy
decision makers. This paper will focus on the
specification and development of the COMPAS
econometric community models. It will describe the
conceptual framework of the proposed models, report
on applications of the models in two states, and briefly
discuss plans for future development and support of the
COMPAS network. The plan takes into account the
realities of secondary data availability at the community
level and it attempts to build on current conceptual
foundations from the social sciences and regional
science. It is evolutionary in that it will be designed to
be flexible and continually improved upon; and it
addresses the institutional and constitutional differences
among states and communities.
The COMPAS model discussed below is based
primarily on the authors’ experiences with the Virginia
Impact Project (VIP) model, and Missouri’s Show Me

2
Community Impact Model which have evolved over the
last decade. However, these models, are themselves
just a recent chapter in a long tradition of community
modeling by rural development researchers (see
Halstead, Leistritz, and Johnson for a history of just
some of these models). The novel aspect of this project
is the attempt to create models for communities
throughout the nation.
KEY PRINCIPLES
There are many considerations involved in modeling a
community for policy analysis. The following
assumptions are based on conceptual logic and/or
empirical studies of communities. Each are reflected in
the proposed COMPAS framework.
1. While economic and social relationships know
no geopolitical boundaries, policy provisions,
public services, taxing authority, and data, do.
Therefore, county, municipal, and public
service boundaries should be at the basis of
any policy model.
2. Communities within states share common
constitutional limitations and responsibilities,
and have developed comparable institutions.
3. Communities with similar economic bases
have similar economic structures. Because of
the importance of climatic, geographic, social
and political influences, economic bases are
frequently quite homogeneous across
geographic regions.
4. Communities of similar size and with similar
geographic relationships to nearby larger and
smaller communities, perform similar central
place roles and are likely to exhibit similar
responses to economic (and policy) stimuli.
5. The fundamental engine for economic growth,
decline, and change at the local level is
employment. Community impacts are effected
through the labor market which allocates jobs
between the currently unemployed, residents
of nearby communities (incommuters), current
residents who work outside the community
(outcommuters), and new entrants to the local
labor market.
6. Changes in employment, unemployment,
commuting, labor force, population, school
enrollment and income, lead to changes in
housing needs, property tax base, public
service demands, and transfers to households
and local governments.
These principles guided the estimation and
development of the Virginia Impact Projection (VIP)
model and the Show Me Model for Missouri
communities. Both models are systems of
econometrically estimated equations for rural towns,
counties and cities in the respective states, using both
cross-sectional and time series data. Experience with
the estimation of these models indicates that with
careful selection of variables and functional form,
stable coefficients can be estimated for communities
with a wide variety of sizes and economic bases. Basic
institutional differences cannot be captured with a
single set of parameter estimates, however.
Furthermore, attempts to apply the model to other states
have underscored the importance of differences in the
structure of public service provision. Therefore, only
states with very similar local government structures will
be candidate for grouping together.
MODEL STRUCTURE
While many different model structures could generate
comparable policy analyses, the COMPAS models will
share a basic structure. The COMPAS models will
based on the assumptions above as well as others about
the way in which rural and small city economies work,
about the way in which local governments make
decisions, and about the conditions under which local
public services are provided. In the following pages,
the first and most simple of the COMPAS models will
be described.
Labor Market Equations
The labor market concept plays a central role in the
COMPAS models. The models are built on the
assumption that economic growth is caused largely by
exogenous increases in employment. This is not to say
that employment at the community level is not
responsive to local conditions but rather, that these
responses will be dealt with as direct changes or shocks
to be introduced to the models. In this simple model,
demand can be viewed as perfectly inelastic at the
exogenous level of employment. Total labor supply is
perfectly elastic at the prevailing regional or national
wage level (adjusted for local cost of living, amenities,
etc.). Labor supply is composed of two components:
locally employed residents and locally employed
non-residents or incommuters. Locally employed
residents equals the resident labor force less
unemployed outcommuters. These relationships are

3
described below in Figure 1.
Figure 1: The Conceptual Labor Market
In- and out-commuters are separated here, rather than
combined into net commuters, because they exhibit
different in preferences for public services, spatial
amenities, occupational characteristics of households,
and because sub-markets for different labor skills
persist. Labor force and incommuters are positive
components of supply and outcommuting is a negative
component. Unemployment is a residual negative
component of supply. Eliminating wages from the
component supply curves by substituting the inverse
demand curve, as amended, derives the expressions.
This introduces employment (demand) to the supply
components. More formally, the model is developed as
follows:
(1) XD = XS ,
equates demand and supply (local employment and
employed labor force from all locations). The demand
curve is
(2) XD = f(w),
(where w is the wage rate) which when inverted
becomes
(3) w = g(XD).
Decomposing labor supply into its components gives
(4) XS = XLF - XU - XO + XI.
Each component of supply is a function of employment
and a vector of supply shifters,
(5) XLF = fL(w,ZLF) = fL(g(XD),ZLF),
(6) XO = fO(w,ZO) = fO(g(XD),ZO), and
(7) XI = fI(w,ZI) = fI(g(XD),ZI),
where, XD is labor demand (local employment), XS is
labor supply, made up of its components, XLF (resident
labor force), XO (outcommuters), XI (incommuters), and
XU (unemployed), w is the wage rate, and the Zs are
Community
Labor Force
Employment
External
Employment
External
Labor Force
Commuting
Shed Outcommuting
Incommuting

4
supply shifters for the various components of supply.
Given the discussion and the conceptual model above,
equations 4 through 7 can be expressed as follows in
equations 8 through 11.
1
(8) Unemployed = Labor Force + Incommuters -
Employment - Outcommuters
All three components of labor supply will be primarily
determined by employment in the location in question.
In addition, they will depend on relative housing
conditions, costs of living, quality of public services,
tax levels, the mix of jobs, and similar variables in the
location of employment, versus alternative locations. A
very important variable in the supply components is
area of the data unit. Smaller units will include fewer
resident laborers, and define more as outcommunters
and incommuters because the cross the borders of the
unit. Larger units will incorporate more destinations
and residences of workers and, therefore, define more
workers as being locally employed, and thus fewer
outcommuters and incommuters. In addition,
commuting will depend on the distance between place
of residence and place of work.
(9) Labor force = f(employment, housing
conditions, cost of living, public services,
taxes, industry mix, area).
(10) Outcommuting = f(employment, external
employment, external labor force, housing
conditions, cost of living, public services,
taxes, industry mix, area, distance to jobs).
(11) Incommuting = f(employment, external
employment, external labor force, housing
conditions, cost of living, public services,
taxes, industry mix, area, distance to
residence).
Population is hypothesized to be a function of labor
force and variables that affect the labor force
participation rate and the dependency ratio.
(12) Population = f(labor force, participation rate,
dependency rate),
Where the dependency rate is the ratio of the non-
working population to the working population.
Fiscal Impact Equations
Changes in the tax base and changes in the need for
expenditures usually accompany changes in
employment. New employers, employees and
population require expenditures for services and
investments in infrastructure. The demands for public
services by residents depend on such factors as income,
wealth, unemployment, age, and education. As growth
changes these characteristics, the demand per resident
will rise or fall. Furthermore, as a community grows
the average cost of producing public services often
decreases, until all economies of size are captured, and
then increases, when inefficiencies creep in to the
process. Together, the changing demand and efficiency
determinants mean that each economic change will
have a unique effect on needed expenditures.
It is assumed that local governments consider the
demands of their constituents, and provide the desired
level of services at the lowest possible cost. When tax
bases and the demand for expenditures are known, local
governments are assumed to adjust tax rate to balance
their budget.
Following Hirsch (1970 and 1977); Beaton; Stinson;
and Stinson and Lubov; unit cost of public services are
hypothesized to be a function of the level, and quality
of services, important local characteristics (input factors
and demand factors), input prices, and the rate of
population growth. Furthermore, theory suggests that
public services may be subject to increasing, and/or
decreasing returns to size. Based on these theoretical
relationships local government service expenditures per
capita are hypothesized to be determined as follows:
(13) Expenditures = f(quality, quantity, input
conditions, demand conditions).
For each type of expenditures (public works, police
protection, administration, parks and recreation,
welfare, education, fire protection, etc.) the independent
variables are defined differently. For education
enrollment is the quantity variable, teachers per
thousand students is a quality variable, federal aid and
change in enrollment are input conditions, and income,
real property, and employment are demand conditions.
For police protection, population is the quantity
variable, solved crimes is the quality variable, percent
population in towns, incommuters, and miles to the
nearest metropolitan area are input conditions, and
income and personal property are demand conditions.
Many non-local revenues (from state and federal
agencies) are at least partially formula driven. Even
when this is not the case, certain local characteristics
may indicate the expected level of these revenues. In
addition, non-local revenues are frequently an inverse
function of the locality's ability to pay and a direct
function of its degree of political influence. Ability to

5
pay is usually related to per capita income, personal
property per capita, and real property per capita.
(14) Non-local aid = f(expenditures, income,
personal property, real property).
Another important source of local revenues is sales tax
revenues. The level of retail sales is primarily a
function of income. This relationship is expected to
change with the size of the locality since larger
localities are usually higher order service centers. The
number of incommuters is also hypothesized to
influence sales because they increase the daytime
population of the community. Sales tax revenues are
hypothesized, therefore, to be:
(15) Sales tax Revenues = f(income, employment,
incommuters).
Other local revenues, other than property taxes, include
licenses, fees, fines, forfeitures, and special
assessments. These revenues are hypothesized to be
related to the level of commercial activity (retail
activity) in the community and the income level. Thus:
(16) Other Tax Revenues = f(Sales tax revenues,
income).
Real property includes both residential and business
property and, therefore, will be influenced by the level
of personal income as well as the size of the economic
base. Both personal and real property are hypothesized
to be positively related to the number of outcommuters
since these families represent a source of wealth that is
not supported by the local economic base.
(17) Real Property = f(income, employment,
outcommuters),
(18) Personal Property = f(income, outcommuters).
There are a number of ways to close this type of mode.
In the case of the VIP model it is assumed that local
government expenditures are determined first, and real
and personal property tax rates are set to cover those
expenditures not met by non-local aid and sales tax
revenues and other tax revenues. This implicitly
assumes that budgets are balanced each year. An
alternate assumption (the one used in the Show Me) is
that the tax rate remains constant and that economic
changes lead to fiscal deficits or surpluses.
THE MODELS APPLIED
To date, the VIP and Show Me models have been
developed for forty to fifty communities. Similar
models have been developed and applied in the several
communities in Iowa (Swenson, 1996), Idaho (Fox and
Cooke, 1996) and Wisconsin (Deller and Shields,
1996). Local advisory committees are usually
appointed to review the baseline projections, help form
the scenarios, review the model’s projection, and to
help interpret the results. The models have been used
for a variety of purposes including analyses of
annexations, jurisdictional mergers, new industries,
existing industries, industry closures, university
research parks, shopping centers, residential
developments, location of industrial sites and, and
general development strategies. They have also been
used for goal planning for several communities. Goal
planning with the models is achieved by estimating the
conditions necessary to bring about a desired set of
terminal conditions.
The models have generally been popular with local and
state governments. Policy makers are generally
somewhat skeptical until they come to appreciate the
information generated and become more confident in
the projections. Repeat users of the model’s projections
especially like the comparability of the results from
case to case, and across communities.
DISCUSSION
The devolution of policy decisions from central to local
control will bring communities many new opportunities
and many significant new challenges. Especially those
that are small or otherwise disadvantaged may now
need the capacity to assess the future impacts of a
variety of expected or proposed changes. The
Community Policy Analysis System is one approach for
rural development researchers to assist in developing
that capacity.
COMPAS models now exist in at least five states.
Preliminary plans are now in place to extend that to
seven, fifteen, and ultimately twenty-five states over the
next three years. In the next six months, researchers
involved in this initiative will review, refine and test the
conceptual framework of the COMPAS models and
specify data and research standards that will make
results from these models comparable and compatible.
If resources are available, these researchers will form a
network designed to provide analysis of the community
impacts of local, state and federal policy alternatives.

6
REFERENCES
Beaton W. Patrick. The Determinants of Police Protection Expenditures. National Tax Journal 24 (1974): 335-349.
Deller, Steve and Martin Shields. The Wisconsin Extension Fiscal Impact Model: Preliminary Results. Presented at
the Community Economic Modeling Conference. June, 1996. Madison, Wisconsin.
Fox, Linnette and Stephen Cooke. Data for the Idaho Fiscal Impact Model. Presented at the Community Economic
Modeling Conference. June, 1996. Madison, Wisconsin.
Halstead, John M., F. Larry Leistritz, and Thomas G. Johnson. 1991. The Role of Fiscal Impact Models in Impact
Assessments. Impact Assessment Bulletin 9 (Fall): 43-54.
Hirsch, Werner Z. Economics of State and Local Government. McGraw-Hill, New York, 1970.
Hirsch, Werner Z. Output and Costs of Local Government Services, Paper for the National Conference on
Non-Metropolitan Community Services Research, Ohio State University, Columbus, Ohio, January 21, 1977.
Johnson Thomas G. A Description of the VIP Model, Unpublished manuscript, Department of Agricultural
Economics, Virginia Tech, Blacksburg, Virginia, April 1991.
Johnson, Thomas G. Representative Community Analysis, paper presented in a symposium entitled, "Rural Impacts
of Public Policies: Improved Analytic Frameworks," at the annual meetings of the American Agricultural
Economics Association, Orlando, Florida, August 2, 1993.
Stinson, Thomas F. The Dynamics of the Adjustment Period in Rapid Growth Communities, Prepared for
Presentation at the WAEA Annual Meetings, Bozeman, Montana, July 24, 1978.
Stinson, Thomas F. and Andrea Lubov. Segmented Regression, Threshold Effects, and Police Expenditures in Small
Cities. American Journal of Agricultural Economics 64 (November) 1982: 738-746.
Swenson, David. Spatial/Institutional Interrelationships. Presented at the Community Economic Modeling
Conference. June, 1996. Madison, Wisconsin.
1
If ones estimate of employment is defined as jobs, rather than the number of persons employed, then it will include second jobs. In
this case, employment as defined here equals jobs less second jobs. Alternatively, one must augment the supply of labor by the number of
individuals holding second and third jobs.