V V ‘

 

tw's

r ltBRARY "
1 Michigan State
,____ University

 

.J

This is to certify that the
thesis entitled

IDENTIFYING THE DETERMINANTS OF SUCCESS FOR
PURCHASE OF DEVELOPMENT RIGHTS BALLOT PROPOSALS
IN MICHIGAN

presented by

Jason Keith Evans Ball

has been accepted towards fulfillment
of the requirements for the

MURP degree in Urban and Regional Planning__

 

 

V/AL W

Major Professor’s Signature

4971—94 2! ZOO‘I
Date

MSU is an Afflnnative Action/Equal Opportunity Employer

 

PLACE IN REI‘URN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 KzlProleocsPresIClRCIDateDueindd

 

IDENTIFYING THE DETERMINANTS OF SUCCESS FOR PURCHASE OF
DEVELOPMENT RIGHTS BALLOT PROPOSALS IN MICHIGAN

By

Jason Keith Evans Ball

A THESIS

Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of

MASTER IN URBAN AND REGIONAL PLANNING
Urban and Regional Flaming

2009

ABSTRACT

IDENTIFYING THE DETERMINANTS OF SUCCESS FOR PURCHASE OF
DEVELOPMENT RIGHTS BALLOT PROPOSALS IN MICHIGAN

BY

Jason Keith Evans Ball

The purpose of this research is to test the hypothesis that higher socioeconomic
status and increasing development pressure have a positive relationship with success of
Purchase of Development Right ballot proposals in Michigan. In order to test this
hypothesis a dataset has been created consisting of all minor civil divisions that have
considered a PDR millage in Michigan and a range of variables relating to socioeconomic
status, development pressure, and political influences. A bivariate correlation is used to
evaluate the effect each independent variable has on the dependent variable, a stepwise
regression based on select variables and a linear regression using all variables determine
the degree to which the most significant variables as well as the entire model explain the

variance in Michigan PDR millage votes.

The sum of this research is that PDR millages in Michigan have been more likely
to pass in urban and suburban communities with a high level of educational attainment
and low rates of owner occupancy that are located in counties where farmland value did
not increase rapidly between 1987 and 2007. Other variables that were found to have a
significant influence on the percent of “yes” votes for a PDR millage are total population,
change in home value, home value, rural population, mobility of population, population
density, percent population white, and support for George W. Bush in the 2004

presidential election.

This thesis is dedicated to my wife Heidi. Without her I would have never graduated

with any remaining sanity or sense of humor.

iii

ACKNOWLEDGEMENTS

I would like to acknowledge the contributions of my primary professor, Dr. Patricia
Machamer and committee members, Mark Wyckoff, and John Melcher. Your
tremendous insight and willingness to help me see the forest through the trees in this
project has been invaluable. In addition, a great deal of assistance on statistical methods
was provided by Dr. John Schwietzer and Paul Curran, whose patience is much

appreciated.

iv

TABLE OF CONTENTS

LIST OF TABLES

CHAPTER 1
INTRODUCTION

The unique situation of Michigan ______________________________________

Rationale for research

CHAPTER 2
LITERATURE REVIEW

Introduction to chapter 2 ______

 

National trends and case studies

Theoretical basis of PDR and open space preservationm

ooooooooooooooooooooooooooooooo

 

 

Community characteristics and open space preservation

Socioeconomic and political data

 

Land use change and development pressure _________

CHAPTER 3
METHODOLOGY

 

Datasets

Analysis _______________________________________________________________________________

Variables

CHAPTER 4
RESULTS
MCD dataset results

 

 

 

County dataset results _________________________________________________________
Community characteristics _________________________________________________
Development pressure .........................................................

Political variables

CHAPTER 5
DISCUSSION

----------------------------------------------------------------------------------

 

------------------------------

 

vi

viii

\0-13

14
14
15
2l
25
26
30

35
35
37
41

43

46
49
50
52

53
53
6O
63
65

67

76

LIST OF TABLES

 

 

 

 

 

 

 

Table 1.1: Selected Michigan 2002 Census of Agriculture rankings __________________________ 5

Table 1.2: Selected Michigan commodity values _________________________________________________________ 6

Table 1.3: PDR millage votes in Michigan __________________________________ __ __ ______ 11
Table 2.1: National land conservation ballot measures ________________________________________________ 22
Table 3.1: Jurisdictions and rate of success __________________________________________________________________ 36
Table 3.2: Independent variable description _________________________________________________________________ 39
Table 3.3: Development pressure county-level data .................................................... 40
Table 4.1: Independent variable abbreviations _______________________________ 43
Table 4.2: MCD bivariate correlation results ______________________ 45
Table 4.3: MCD stepwise regression model____ ............................... 45
Table 4.4: MCD stepwise regression coefficients___ _________________________ 46
Table 4.5: MCD linear regression model ______________________________________________________________________ 46
Table 4.6: County bivariate correlation results _____________________________________________________________ 47
Table 4.7: County stepwise regression model ______________________________________________________________ 47
Table 4.8: County stepwise regression coefficients _____________________________________________________ 48
Table 4.9: Model significance _____________________________________________ 48
Table 4.10: Community characteristics variables _________________________________________________________ 49
Table 4.11: Development pressure MCD regression variables ___________________________________ 51
Table 4.12: County-level development pressure variables __________________________________________ 51
Table 4.13: Political variables ________________________________________________________________________________________ 52
Table 5.1: Farmland value change _________________________________________________________________________________ 56
Table 5.2: MCD datasets comparison ___________________________________________________________________________ 57

vi

Table 5.3: MCD datasets regression comparison ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Table 5.4: Double-counted jurisdictions ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Table A]: County MCDS ______________________________________________________________________________________________

Table B.1: MCD dataset independent variable correlations",

 

Table C. 1: Rural MCD descriptive statistics ________________________________________________________________
Table C.2: Urban MCD descriptive statistics _______________________________________________________________
Table C.3: MCD descriptive statistics ___________________________

 

vii

59
64
67
69
73
74
75

LIST OF FIGURES

Figure 1.1: Millage vote location __________________________________________________________________________________ 12

viii

CHAPTER 1: INTRODUCTION

“NEITHER NATURE NOR PEOPLE ALONE CAN PRODUCE HUMAN
SUSTENANCE, BUT ONLY THE TWO TOGETHER, CULTURALLY WEDDED.”

-WENDELL BERRY, THE UNSETTLING OF AMERICA, 1977

The importance of preserving land in an undeveloped state is difficult to deny.
Faith in the ability of “the market” to effectively direct grth has created a nation of
sprawling development and with it environmental degradation, a loss of social capital,
and a generation of Americans largely disconnected from the land that sustains them.
The permanent preservation of land in an undeveloped state provides myriad benefits. In
addition, it may provide a solution to urban sprawl. However, purchasing large amounts
of land for use simply as open space to prevent the built environment from bleeding into
the country side is not fiscally possible for government, nor is it a responsible use of land
as a resource. A balance must be struck between preserving land, taking advantage of the

resources provided by land, and creating prosperity for people.

Land is permanently preserved in four ways, property can be purchased out-right
by a government or conservation organization, a conservation easement can be donated
by a land owner to a conservation organization, a unit of government or organization can
enter into an agreement to purchase a landowner’s right to develop her land (purchase of
development rights (PDR)), or a land owner in a designated “sending zone” can sell her
development rights to a developer in a “receiving zone” to allow higher-density

development (transfer of development rights (TDR)). Zoning, the most common tool

used to regulate the development of land is cheap and relatively effective, but it is
imperrnanent and can be unpopular with landowners, especially if used to severely limit
its use. Zoning districts are subject to political pressures and population shifts, and
effective open space zoning requires twenty or more acres per dwelling, reducing the
value of land without compensating owners. PDR (also known as purchase of
conservation easements) and TDR both provide solutions to the impermanence and

decrease of land value inherent to zoning without requiring the outright purchase of land.

The purpose of this research is to identify the factors that determine the success of
Michigan ballot proposals to purchase the development rights to land. Nationwide and
regional studies in the Northeast and West have consistently found that higher
socioeconomic status is a strong predictor of the eventual success of land preservation
ballot measures (Chapter 2 discusses this in detail). This research demonstrates that PDR
millages in Michigan have been more likely to pass in communities with a high level of
educational attainment and a low rate of owner occupancy that are located in counties
where farmland value increased relatively slowly between 1987 and 2007. Although
these findings do not necessarily coincide with previous research, significant differences

exist between the characteristics of the datasets used, requiring further investigation.

The focus here is on PDR because local governments have not been explicitly
enabled to implement TDR in Michigan. There are a variety of differences between the
two tools, but most simply, since TDR is market-based it is less-expensive and is more
complicated to implement. PDR relies on the fundamental principles of private property
ownership to permanently preserve land while keeping it in private ownership. Just as a

landowner has the right to farm her land or extract minerals, she also has a right to

develop it to its fullest permitted potential. Any one of these rights in land can be
separated from the others, and sold or donated to another party. In the case of PDR, the
right to develop land is purchased from a landowner by a private organization or unit of
government, thereby maintaining the land’s undeveloped state, while allowing the owner

to retain other landowner rights and responsibilities. (Daniels & Bowers, 1.997)

By no means is the purpose of this research to advocate for PDR as a silver bullet
to the land use problems facing Michigan. However, the ability of PDR to preserve land
in perpetuity while simultaneously keeping land in economic production provides a
positive-sum solution in a land use game that is all too often viewed in zero-sum terms.
Additionally, it must be emphasized that truly successful land preservation programs
never occur in a vacuum. Successfirl land use decisions require long-range strategic
planning, public involvement, and excellent leadership. Preserving a farm surrounded by
l-acre lots is a negative-sum result, the farmer views the business of farming as doomed,
the residents are burdened with noise, dust, and odors, and the local government is faced
with conflicting land uses that are difficult to reconcile. This hypothetical situation is
entirely possible if land preservation efforts are not based on a comprehensive land use

plan that is effectively implemented over a large area.

To-date, nineteen open space/farmland preservation millage votes to purchase
development rights have taken place in eighteen Michigan jurisdictions, eight of which
have passed.I Millages to purchase land out-right for public parks or open space are not

considered. The nature of fee-simple purchases (all rights in land) is significantly

 

lPeninsula Township, in Grand Traverse County, the first Michigan jurisdiction to pass a PDR millage
voted on, and passed a PDR millage in 1994 and 2002, both votes are included in the dataset.

different from PDR, and the prevalence of such programs is far too great for

consideration in this context.

In order to frame the situation and the purpose of this research, the unique

character of Michigan’s physical and political attributes needs to be discussed briefly.

The Unique Situation of Michigan

Michigan is facing challenges unlike any it has seen in the past. The core of its
economy and the value-system of its people are in question as quality manufacturing jobs
have fled the state at a feverish pace since the late 1990’s. The deeply ingrained belief
that a high school education and a strong back entitle a Michigan resident to a well-
paying job on an assembly line is a thing of the past. In order to rectify this situation
Michigan must turn its focus away from attracting the next big automotive factory and
instead examine the assets that set the state apart from the rest of the country, and the

world.

Michigan’s most obvious asset, and arguably one of the most under-utilized is its
natural resource base. In 1980, Michigan was home to 11 million acres of farmland and
18.2 million acres of forest land. Less than 3 million acres of land were developed.
(Michigan Land Use Leadership Council, 2003) According to the 2007 Census of
Agriculture, Michigan is still home to over 10 million acres of farmland and over 50,000
farms. (USDA National Agricultural Statistics Service, 2009) Furthermore, Michigan’s
farmland is especially valuable. The state ranks 31St nationally in number of acres
devoted to farms, ranks 22nd in Market Value of Agriculture, and is 12th in Agricultural

Market Value per acre (see Table 1.1).

Table 1.1: Selected Michigan 2002 Census of Agriculture rankings (American Farmland Trust, 2007)

 

 

 

 

Number of Farms 16
Land in Farms 31
Market Value of Agriculture 22
Agricultural Market Value Per Acre l2

 

 

 

 

More important than Michigan’s sheer amount of farmland and open space is its
ability to produce specialty crops and provide unique landscapes that do not exist
elsewhere. Michigan produces crops that are vital to the national and global market and
is blessed with the nation’s second-most diverse agriculture. (Daniels & Bowers, 1997)
As an example, in 2007 Michigan produced 230 million pounds of tart cherries, almost
eighty percent of the tart cherries produced in the United States. (National Agricultural
Statistics Service, 2007) Table 1.2 displays other Michigan agricultural products in the
top ten in sales nationally. Additionally, Michigan is home to over 3,000 miles of Great
Lakes shoreline, over 11,000 inland lakes, and 3.9 million acres of state forests.

(Michigan Department of Natural Resources, 2008)

Despite its resources, the ever-increasing loss of farmland in Michigan since the
middle of the 20‘h century has been well publicized. In the 1970’s it was estimated that
Michigan farmland was being converted to other uses at a rate of 300 acres per day
(Governor's Task Force on the Future of Agriculture, 1970) and in 1994, a study
commissioned by governor Engler estimated farmland loss at 240 acres per day.

(Michigan Farmland Agriculture Development Task Force, 1994) In 1997 Norris et al.

Table 1.2: Selected Michigan commodity values (National Agricultural Statistics Service, 2002)

Commodity National Rank in

Value nl‘Salcs

 

Fruit, Tree Nuts, and Berries 5

 

Nursery, F loriculture, Greenhouse, and Sod

 

Cut Christmas Trees and short rotation Woody Crops

 

OOAQ

Milk and other Dairy products from Cows

 

 

 

 

conducted a land cover change analysis and totaled farmland loss fiom 1982 to 1997 at
1.5 million acres, or more than 13% of all the farmland in the state (more than 270
acres/day during the time period) (Norris, Soule, Weissert, Stuart, & Skole). Finally, in
2003 Governor Granholm signed an executive order creating the Michigan Land Use

Leadership Council.

“The Executive Order charged the council with studying and identifying trends,
causes, and consequences of urban sprawl and providing recommendations to the
governor and the legislature designed to minimize the negative effects of current
and projected land use patterns on Michigan’s environment and economy.”

(Michigan Land Use Leadership Council, 2003, p. 1)

Among other things, the final report of the Council found, “Funding available for the
state’s current PDR program is inadequate. Interest from farmers far outstrips available

funding.” (Michigan Land Use Leadership Council, 2003, p. 48)

The sum of these statistics is preserving farmland and open space in Michigan is
vital to its future. In 1974, the Michigan Legislature adopted PA 116, the state’s first

farmland preservation program, largely as a result of reports estimating that Michigan

was losing farmland at a rate of 300 acres per day. (Norgaard, 1999) The program does
not preserve farmland outright; instead it provides a significant tax break to encourage

farmers to keep their land in agricultural use.

Owners of farmland participating in the PA 116 program enter into a development
rights agreement with the state of Michigan; the agreement transfers the rights to develop
their land to the state for at least 10 years. In return, the landowner receives a state
income tax credit for the amount property taxes exceed 3.5% of household income,
thereby effectively reducing farmland owners’ property taxes without depleting the local
tax base. Upon the expiration of the development rights agreement a lien is placed on the
property equal to the tax credits from the previous seven years. The lien is due when the

land is sold or converted to a non-agricultural use.

Income collected fiom the PA 116 program (liens paid) provides funding for the
state Agricultural Preservation Fund. The primary purpose of the Agricultural
Preservation Fund, created in 2000, is to provide money for state grants to local units of
government to purchase development rights. Once administrative costs have been
budgeted for, and grants to local units of government have been made, any money
remaining in the fimd in excess of $5 million may be used by the state to purchase

development rights.

As of 2004, the state of Michigan PDR program had purchased 57 easements and
protected over 13,000 acres of farmland at a price just over $2,000 per acre. (Adelaja, et
al., 2006, p. 20) This is significantly less land than required for Michigan to maintain

agriculture as a viable industry. In order to protect half the farmland vulnerable to

development in Michigan, over 800,000 acres would need to be preserved, at an
estimated cost of around $2 billion. (Adelaja, et al., 2006) A possible solution to the
short-coming in the state’s ability to preserve the necessary land base is a concerted effort
at the local level, “The amount of money allocated for open space protection by local and
state governments exceeds the budget of the largest federal land protection program in
the US, the Conservation Reserve Program.” (Nelson, Uwasu, & Polasky, 2007, p. 380)
Local purchase of development rights programs are enabled in Michigan by the
Michigan Zoning Enabling Act (MZEA) (PA 110 of 2006) and article I part 21 subpart
ll of the Natural Resources and Environmental Protection Act (NREPA) (MCL
324.2140-324.2144). The MZEA permits local units of government to adopt a PDR
ordinance, fund a PDR program and provides specific requirements for operation of a
PDR program. While NREPA permits governmental or private entities to acquire
conservation easements, but provides no specific requirements for the operation of a
preservation program. NREPA also provides guidelines for the state PDR program (PA

116) and creates a basis for granting money to local units of govemment..

According to the MZEA, a local legislative body must adopt a development rights
ordinance in order to implement a PDR program. PDR programs in Michigan may only
be used to protect “agricultural land and other eligible land.” (2006 PA 110 §507(2)
M.C.L. 125.3507(2)) hnportantly, the PDR program is not meant to operate in isolation
from other land use regulations, if the local unit of government has a zoning ordinance,
the PDR program must, “be consistent with the plan. . .upon which the zoning ordinance

is based.” (2006 PA 110 §508(2) M.C.L. 125.3508(2))

The MZEA also permits a variety of financing options for local units of
government to fund the purchase of development rights. The options for funding include
general appropriations, general fund revenue, special assessment districts, and the
issuance of bonds and notes. According to the Trust for Public Land’s LandVote
database, only four Michigan municipalities have voted on bond measures to preserve
land since 1988, none of which have been for the purchase of development rights to

farmland and/or open space.

Rationale for Research

By 1994, almost half of the state’s farmland was enrolled in the PA 116 program.
(N orgaard, 1999) However, development pressures were high and the primary
disadvantage of the PA 116 program, its impermanence, was apparent in many areas.
That same year Peninsula Township, in Grand Traverse County, became the first
Michigan jurisdiction, and the first jurisdiction in the Midwest, to consider taxing itself to
preserve farmland. The campaign garnered national attention; the eventual vote for the
program was covered by CBS and the New York Times, among others. (Bidwell,
Westphal, Wunsch, & Berton, 1996) The millage passed, and the campaign was viewed
as an outstanding success by local people as well as national farmland preservation
advocates. In 2002, the township again voted to increase property taxes to preserve
farmland, except this time the township voted to increase the original .75 mils tax to 2.0
mils. Both millages passed with over 55% of voters in support and a broad range of

supporters from farmers, land use experts, and local officials.

In 1998, the next Michigan jurisdiction to consider taxing itself to preserve
farmland, Washtenaw County, had a very different experience. After witnessing the
steady conversion of farmland to residential and commercial uses, Washtenaw County
attempted to adopt a PDR millage. The campaign was hotly contested, with
approximately $600,000 combined being spent during the campaign by the pro and anti-
millage campaigns. Despite the efforts of over 600 people volunteering for the pro-
millage campaign (Lonick & Garfield, 2008) the millage was soundly defeated (57%

voted against the millage).

A host of explanations were given for the millage failure; “In the post-election
reprise, supporters saw the unexpected defeat mostly as a sign of resistance to a new tax
but also to the complexity of the proposal. ‘We failed to communicate the vision that
connects the pieces,’ observes Keith Schneider of the Michigan Land use Institute.”
(Myers, 1999, p. 13) Perhaps the most telling reason given for the millage’s failure was,
“We ran a stupid campaign.” (Lonick & Garfield, 2008). The assumption made is not
that voters did not want to preserve farmland, or even that they were not willing to
increase taxes, but that they were convinced to vote against a millage based on

misinformation and a well-run anti-millage campaign.

Since the 1998 Washtenaw County vote, seventeen other jurisdictions in
Michigan have voted on a millage to purchase development rights to preserve farmland
and/or open space. (Table 1.3) Of those seventeen jurisdictions, eight have passed such a
millage, and nine rejected it. In Washtenaw County, five votes have taken place (four in
townships and one in a city) since the failed county—wide millage vote, three of which

have been successful. Thus it would seem that organizers in Washtenaw have developed

10

a recipe for success, but either that strategy is not applicable to the rest of the state, or it is

not being employed appropriately. Figure 1.1 displays the location of each millage vote

and its result.

Table 1.3: PDR millage votes in michigan

 
 

Local [hit of (im crnmcnt

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Peninsula Township 1994 55.8% 1994
Washtenaw County 42.5% 1998
Meridian Township 55.8% 2000
Peninsula Township 2002 59.0% 2002
Ann Arbor City 67.4% 2003
Ann Arbor Township 77.0% 2003
Acme Township 55.6% 2004
Bridgewater Township 45.6% 2004
Elk Rapids Township 48.1% 2004
Milton Township 40.5% 2004
Scio Township 74.6% 2004
Torch Lake Township 40.6% 2004
Whitewater Township 47.4% 2004
Agusta Township 42.5% 2005
Berlin Township 22.1% 2005
Webster Township 69.0% 2005
Lapeer County 41.4% 2006
Leelanau County 37.9% 2006
Ingham County 50.3% 2008

 

  

 

The divergence in results illustrated by the 1998 Washtenaw County and 1994
Peninsula Township PDR millage campaigns signifies the difficulty in predicting and
understanding the likelihood of success of such millages in Michigan. The Trust for
Public Land’s LandVote database, which began recording information on every
conservation ballot measure considered in the united states in 1998, shows that 83% of
conservation property tax increases proposed across the United States since 1998 have

been successful. (Kotchen & Powers, 2006) This stands in stark contrast to the success

11

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

rate 0f PDR millage votes in Michigan (47%) and the rate of passage for all conservation

property tax increases in Michigan (64%).

This project seeks to explain the variation in results of PDR millage votes in
MiChigan. Variables related to socioeconomic characteristics, development pressure, and
Political circumstances are compared with the percentage of “yes” votes in each

JuriSdiction that has voted on a PDR millage in Michigan using a linear regression
analysis. The information provided by this analysis informs local units of government
Considering adoption of a PDR ordinance and/or millage of the influences that
Community characteristics have on support for land preservation using PDR. As a
consequence, local decision makers will be more aware of the unique challenges and
Opportunities in their community and be able to increase the likelihood of a successful

millage, thereby permanently preserving more open space land.

The following two chapters discuss relevant literature and the methodology of the
1‘ eSearch in detail. The final two chapters detail the results of the research and discuss

those results.

13

CHAPTER 2: LITERATURE
REVIEW

Introduction to Chapter 2

One note on semantics is necessary prior to reviewing the literature. The terms
farmland preservation and open space preservation are closely related, but not
interchangeable. Open space preservation includes preservation of farmland, forest
lands, and wildlife habitat, generally any undeveloped land, but not city parks or sports
fields; this land may or may not remain in private ownership. Farmland preservation is
concerned solely with the preservation of land used for farming purposes, which
generally remains in private ownership. The focus of this research is on PDR millages,
which depending on the jurisdiction, may be used for the preservation of a variety of
lands. The literature reviewed uses both terms, depending on the topic. Generally,

earlier literature was focused on farmland preservation, but as the work becomes more

current, the term open space is most common.
The influence of the characteristics of local populations on the manner in which
they vote on open space preservation measures has been considered by a diverse range of
disciplines since the mid 1970’s. Researchers have hailed from Agricultural Economics,
maturing, Geography, Political Science, Environmental Studies, and Natural Resource
Management. This diversity of interest reflects the general applicability of open space
preservation. However, the result of the diversity of disciplines is a diversity of
approaches to testing similar hypotheses. This chapter begins by focusing on the

t . . . . .
heOretrcal aspects of PDR for open space preservation, briefly discusses national trends,

14

and then proceeds into a more focused review of strategies employed to understand the

relationship of community characteristics to the likelihood of successful land preservation

measures.

Theoretical Basis of PDR and Open Space Preservation

The first question to be addressed when considering PDR and open space
preservation is: What benefits does society gain from preserving open space? The
answer to this question has varied in the literature. In an influential article, which painted
agricultural preservation in a somewhat negative light, Gardner argued against the need
for non-market based farmland preservation techniques, “it is difficult to see why the
market will not allocate sufficient land to food and fiber production. The food and fiber
sector of theeconomy is as competitive as any other, and food and fiber prices adjust
rapidly to changing conditions in supply and demand.” (Gardner, 1977, p. 1030)
However, Gardner identified the inability of the market to provide for environmental
amenities and open space, “In any case, in principle at least, the market will not provide
the optimal quantity of these amenities; and there may be some justification for social
action to remedy this market failure.” (Gardner, 1977, p. 1031) In addition to framing the

inher-em difference between open space preservation and farmland preservation, Gardner

also defined the rationale for farmland conservation.

What society appears to get in the action of preserving agricultural land are at

least four joint products. . .(a) ‘sufficient’ food and fiber to meet the nutritional
requirements of a growing national and world population; (b) local economic

benefits that derive from a viable agricultural industry; (c) open space and other

15

environmental amenities that accrue chiefly to urban residents; and (d) more

efficient, orderly, and fiscally sound urban development.” (Gardner, 1977, pp.

1028-1029)

In later works advocating for non market-based land preservation techniques,
Gardner’s summary of the benefits of agricultural land preservation remains largely
intact. Kline and Wichelns stated, “the public’s objectives for preserving farmland can
be described generally as agricultural (preserving agricultural resources), environmental
(protecting environmental resources), and municipal (managing the growth of towns and
cities)” (Kline & chhelns, 1994, p. 224) Norris and Deaton also identified the basic
arguments for farmland preservation, “These are 1) maintaining the supply of food, 2)
local economic benefits, 3) growth management, and 4) preservation of environmental

amenities.” (Norris & Deaton, 2001)When considering farmland preservation specific to

Michigan, Skjaerlund and Sheridan wrote,

“The primary benefit of farmland preservation programs is to create a long-term
business environment for farm operations with a local community — and in the
process maintain the economic contribution of agriculture to the local economy,

ensure a local food supply, and minimize the land use conflicts between farming

operations and non-farm neighbors.” (Skjaerlund & Sheridan, 2003, p. 10)

Skj aerlund and Sheridan’s definition of public benefits reflects a common disconnect
betvveen farmland preservation and a desire for open space and environmental amenities.

F a~1‘Iznland preservation programs are commonly meant to preserve farming as an industry

16

and way of life, which can come in conflict with the public’s desire for an ideal rural

setting.

Instead of considering the preferences of individuals, Duke and Lynch (2007)
reviewed state conservation programs to identify states’ perceived benefits of open space
preservation, “states indicate five important goals (with the first three appearing most
frequently): food security, environmental services, protection of rural amenities, planned
development patterns, and a healthy local economy.” (Duke & Lynch, 2007, p. 124)
None of the common state goals identified by Duke and Lynch reference the conflict

between farm uses and non-farm uses, a chief reason farmers sell their land for

development. (Sherman, Milshaw, Wagner, & Freedgood, 1998)

Thomas Daniels, one of the foremost experts on the use of PDR, finds the primary
benefits of PDR to be its fairness to landowners and its level of permanence. He argues
for its distinct benefits over typical growth management tools, “In short, zoning is
malleable and politically vulnerable, whereas PDRs are legally sound and afford more
Permanent protection for farmland.” (Daniels T. , 1991, p. 425) However, he also points
out the primary problem with PDR as it relates to planning, is its voluntary nature. “the
Voluntary aspect may affect the equity of the program: A well-to-do landowner may be
able to afford to sell development rights whereas a poor landowner may need the higher
r eturns from selling for nonfarrn use.” (Daniels T. , 1991, p. 424) An additional issue,
also related to equity, is that PDR rewards landowners for public investments that create
a demand for their land. (Wright, 1994) As access to public services (water, sewer) is

pro\Iided, the value of farmland increases, as does the value of the development rights to

the land.

17

In addition to questioning the public benefits gained from open space

preservation, experts have also challenged the disconnected relationship between open

space preservation and growth management.

“While the advocates of farmland protection policies acknowledge that the
amenity value of agricultural open space contributes to the need for these policies,
they deny the growth management hypothesis, and respond that farmland policies

are intended to protect critical agricultural resources.” (F urseth, 1987, p. 49)

In a telephone survey of 850 residents of Mecklenberg County, North Carolina, Furseth
examined respondents’ level of support for farmland preservation and challenged the
hypothesis that growth management is “designed by affluent suburbanites to exclude less
affluent populations from moving into their communities.” (Furseth, 1987, p. 50) The
conclusion of the telephone surveys was that the majority of respondents (71 percent)
support farmland preservation. However, older, female, higher-income, better—educated

respondents were the strongest supporters of farmland protection. (Furseth, 1987)

Daniels and Lapping (2005), when considering the relationship between land
Preservation and smart growth, found, “Land markets have repeatedly failed to create
Satisfactory land use patterns and instead have fostered residential and commercial sprawl
that wastes land resources, provides too little public open space, and destroys wildlife
habitat.” (Daniels & Lapping, 2005, p. 318) As a result, they advocate for incorporating
Open space preservation into smart growth policies. The current lack of cross-over

bet\Jveen the two tools, both largely created in response to the same problem, sprawl, is

both a problem and a great opportunity.

18

“Striking a balance among the natural environment, working landscapes, and the
built environment is one of the biggest challenges that local governments face.
This means that planning for smart growth is at least twice as challenging as
planning for grth alone—a reality that the smart growth literature has not

recognized.” (Daniels & Lapping, 2005, p. 326)

In addition to developing the theoretical basis of open space preservation, many
efforts, like the study by Furseth, have been undertaken to define the public’s preferences

for open space preservation programs.

In 1996, Kline and Wichelns undertook research to understand the public’s desire
for farmland preservation. They conducted focus groups in Rhode Island Department of
Motor Vehicles Offices and later surveyed participants in order to understand the public’s

beliefs about the purpose of the state PDR program.

“Factor analysis. . .reveals a belief among the public that environmental objectives
such as protecting groundwater and wildlife habitat, and preserving natural places
should be important objectives of farmland preservation programs. . .agrarian
objectives, such as providing local food and preserving farming as a way of life,
should not be viewed as the most important objectives of preserving farmland.”

(Kline & Wilchens, 1996, p. 547)

Kline & Wichelns’ 1996 focus groups found the five most important objectives of
farInland preservation to be (from 1-5): Protecting groundwater, protecting wildlife
hab itat, preserving natural places, providing local food, and keeping farming as a way of

life- (Kline & Wichelns, 1996, P- 541)

19

A similar study of the public’s demand for farmland preservation was conducted
by Norris and Deaton in Kent County, Michigan, with very different results from the
Kline and Wichelns 1996 study. Phone surveys were conducted with 133 residents of the
County; the survey population was split nearly 50-50 between urban and rural residents.
The five opinions about the benefits of preserving farmland respondents agreed with most
were (in order from 1-5): Farmland provides a sense of local heritage, farmland provides

open space, farmland supports the local economy, farmland provides scenic beauty, and
farmland prevents urban sprawl. (Norris & Deaton, 2002) Obviously, some basic
differences between the geographies of Kent County and Rhode Island, in addition to
differing research methods may contribute to the differences in results. However, the

prevalence of “agrarian objectives,” as defined by Kline and Wichelns, in the Norris and
Deaton study is worth noting.

The theoretical basis for PDR and the public’s relative support for a program’s
goals are extremely important when considering the reasons for success of a particular
ballot measure. 1f the proposed program is focused on preserving farming as a way of
life, but the public’s desire for farmland is focused on preserving scenic vistas and open
Space, voters may be less inclined to increase taxes. More importantly, if the program’s
goals are not clear at the time a tax is voted on, and the tax is approved, the program runs

the risk of preserving farmland in a manner contrary to the desires of the citizens paying

f 01‘ the program.

20

National Trends and Case Studies

In order to frame the following analysis of open space preservation measures, a
basic discussion of the prevalence of open space ballot measures and their general rate of
passage on a national level is necessary. In 1998, the sudden explosion of land
preservation and growth management measures appearing on state and local ballots
across the United States gained national attention. “The local votes were quickly
described as part of a national, dramatic grassroots rebellion against sprawl, an urgent
call for preserving land, and a clear message of support for smart growth policies.”
(Myers, 1999) In the 1998 election, 240 local and state conservation ballot measures were
considered (including Washtenaw County), 72 percent of which were approved,
triggering over $7.5 billion in new spending. (Myers, 1999; Romero & Liserio, 2002;
Kotchen & Powers, 2006) It is important to note that not all of these measures were for
new taxes; many were funded using bonds or other means. Despite the high rate of
overall success, unsuccessful high-profile campaigns were not uncommon. “In general,
failed measures are perceived to result from an overly narrow focus, too little preparation,

or overconfidence in voter support for a vague green measure.” (Myers, 1999, p. 12)

Since 1998, the number of ballot measures considered nationally has varied
significantly. The variation is based on the election cycle, more measures are considered
in state and national election years than odd years, but it also tracks the relative health of
the national economy. Table 2.1 is taken from the Trust for Public Land’s LandVote
database. The distribution of these measures is not equal across the country. Since 1988,
1,204 measures have been considered in Mid-Atlantic and Northeastern states, compared

with 1,028 measures in the rest of the country. (Trust for Public Land, 2009)

21

Table 2.1: National land conservation ballot measures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1988 24 22 $1,947,133,862 $1 ,418,078,862
1989 29 22 $1 ,409,488,521 $937,676,870
1990 41 24 $1 ,386,796,066 $1,312,828,066
1991 16 10 $187,802,360 $168,157,360
1992 36 26 $2,038,626,000 $1 ,744,941,000
1993 22 18 $578,315,860 $551,137,753
1994 50 33 $1,044,541,125 $621,248,511
1995 41 33 $1,234,512,844 $1,114,619,344
1996 98 70 $5,252,905,715 $1,341,898,035
1997 81 67 $2,590,953,306 $768,714,321
1998 176 144 $7,229,154,744 $5,857,672,774
1999 107 95 $2,423,294,502 $2,172,135,868
2000 208 170 $1 1,434,170,431 $5,223,047,298
2001 198 138 $1 ,802,683,640 $1,369,510,437
2002 192 142 $8,589,701,l62 $5,502,616,357
2003 133 99 $1,791 ,915,328 $ 1,275,871,985
2004 218 163 $26,110,658,413 $3,975,214,265
2005 141 111 $2,618,811,630 $1 ,598,003,889
2006 183 136 $29,082,167,202 $6,706,777,535
2007 99 65 $2,244,755,926 $1,951,415,707
2008 127 90 $1 1,102,328,340 $8,407,276,140
2009 10 5 $287,393,960 $248,397,987
TOTAL 2,230 1,683 $122L88,110,937 $54,267,240,364

 

Source: http://www.tpl.or2/tier3 cd.cfm?content_item id=12010&folder id=2386

 

This difference in regional distribution is evident not only on the national level,
but at the state and regional level as well. Figure 1.1displays the tendency of PDR
millage votes to occur in Michigan jurisdictions located in the Southeast and Northwest

comers of the Lower Peninsula.

Two of the most successful states in preserving open space and agricultural land
have been Pennsylvania and Maryland, and two counties within those states, Lancaster

County, Pennsylvania and Montgomery County, Maryland have been particularly

22

successful. “. . .about thirty-six thousand acres have been conserved using TDRs’
[nationally] twenty—four thousand of those acres are in Montgomery County, Maryland.”

(Wright, 1994, p. 382)

Obviously the rest of the country has caught up to Montgomery County’s use of
TDR since the early 1990’s, but Montgomery County is the gold standard for the tool, as
the county had preserved over 51,000 acres as of 2004. (Wison & Paterson, 2002) The
state of Maryland on the whole has been very successful in preserving land, largely due
to its active growth management program. A wealth of land preservation programs have
been established over the last four decades. In 1969, Maryland established Program
Open Space, which authorized a real-estate transfer tax to fund land preservation.
Additionally, the state has established the Maryland Environmental Trust, the
Agricultural Land Preservation Program, and the Rural Legacy Program. (McMahon &

Mastran, 2004)

Montgomery County has taken full advantage of all state programs, in addition to
establishing effective local programs for land preservation. Much of the county’s success
can be attributed to a consistent commitment to long-range planning. As early as 1964
the county council was aware of the negative effects of unplanned growth, and adopted a
regional plan to preserve the area’s quality of life. In 1980, the county created a 93,000
acre Agricultural Reserve, where zoning required 25 acres per dwelling unit. (Wilson &

Paterson, 2003)

Within the Agricultural Reserve there are a variety of programs that purchase

development rights, although development rights transfer is much more prevalent. The

23

Rural Legacy Program, the Montgomery County Agricultural Easement Program, and the
Maryland Agricultural Land Preservation Foundation have all purchased development
rights in the county. Funding sources for the programs vary: the RLP is funded by the
state real-estate transfer tax; the MALPF is funded through the real-estate transfer tax and
an agricultural transfer tax; and the County Program is funded by proceeds from the
agricultural transfer tax and general obligation bonds. (Wilson & Paterson, 2003)
Overall, the success of Montgomery County would seem to be due to the variety of tools

employed to achieve the same goal, preserving open space.

Like Montgomery County, Lancaster County has employed a variety of
techniques to protect open space and has a history of effective long-range planning. In
1975 the County master plan identified 278,000 acres for preservation; this was followed
by townships within the county adopting agricultural zoning on over 300,000 acres.
(Daniels T. , 1998) Lancaster County has also created Agricultural Security Areas,
designated areas where landowners have applied to their respective township for
designation as a security area. The designation does not impose any additional regulation
on land owners; however, it does provide some protection from nuisance suits and
condemnation actions for landowners. Additionally, being located in a security area
makes a landowner eligible for the county purchase of development rights program. As
of 2002, the county had preserved approximately 55,000 acres. (Wilson & Paterson,

2003)

The Lancaster County Purchase of Conservation Easement (equivalent to PDR)
Program is funded by a variety of sources. The county has previously received funding

from the Federal Farm Bill, but the bulk of funding comes from state and county funds.

24

State funds are generated by a two—cent tax on cigarettes. The county matches funds

provided by the state from the general fund or through bonds. (Wilson & Paterson, 2003)

It is important to note the differences between the variety of funding and
programs in place in Lancaster and Montgomery Counties in comparison with the
funding and program options available in Michigan. However, perhaps more important is
the commitment to master planning in both counties that preceded open space

preservation efforts.

Community Characteristics and Open Space Preservation Measures

When attempting to predict/explain the success/failure of land preservation
measures the literature generally considers variables related to three areas, two of which
are commonly combined: 1) Socioeconomic data and political preference, and 2) land use
change and development pressure. Studies considering socioeconomic data and/or land
use change are the most common, and few studies consider one without the other,
however, all studies considering political preference also consider socioeconomic data,
thus the split. The difficulty in attributing political preference to votes taking place over
a series of years, as political preferences may change based on candidates running in a
particular election year makes it difficult to consider political factors alone.
Socioeconomic data is also commonly conceived as a proxy for political preference.
Following, studies are divided based on their primary area of concern, overlap between
categories is unavoidable, but the division is helpful when considering the variety of

literature.

25

Socioeconomic and political data

In one of the earliest studies of the effect of socioeconomic data on voting results,
Deacon and Shapiro constructed an econometric model to explain voter preferences for
public (largely environmental) goods and evaluated it using a regression. The voting
results from two California referenda were the dependent variables, and education,
income, and political preference were among the independent variables. They found
voters largely consider their self-interest, even when voting for collective goods,
“observed voting responses were consistent with self-interest.” (Deacon & Shapiro, 1975,
p. 954) Additionally, it was found that conservative voters were more likely to vote

against the measures, while more educated voters generally were supportive.

Kahn and Matsusaka (1997) evaluated voting patterns on environmental
initiatives on a county-wide basis in California. Among the primary purposes of the
study was determining whether political preference was relevant in such votes, or voters
tended to vote based on self-interest, not political ideology. Variables included in an
initial analysis, not considering political affiliation, included education, importance of
resource-dependent industries, population, and income. In the initial analysis they found,
“income matters for environmental voting. Moreover, the relationship between voting
for environmental goods and income appears to be concave. . .concavity implies that it
[the environment as a good] is normal for low-enough incomes and inferior for hi gh-
enough incomes.” (Kahn & Matsusaka, 1997, p. 151) In order to evaluate the relative
importance of party preference, Kahn and Matsusaka used two variables, results from the
1994 presidential vote and party registration. The presidential vote variable was most

Significant, and “while income and price explain most of the variation, it seems that party

26

preferences can be useful in explaining some of the residual variation.” (Kahn &

Matsusaka, 1997, p. 161)

Soleck, Mason, and Martin, in a review of voter support New Jersey’s 1998 open
space ballot measure, considered an array of variables, including the presence of federally
or state protected lands, a variable unique to their analysis. The influence of location was

significant in their analysis,

“Although strong support was widespread, the highest levels of support were in
northern and central New Jersey, and more specifically in areas known to be
experiencing greater than average growth. In general, support was lowest in older
urban areas and in communities in the southeastern part of the state.” (Soleck,

Mason, & Martin, 2004, p. 631)

Generally, Soleck et al. found voter support to be most highly correlated with higher
socioeconomic status, and low voter support to be correlated with the presence of existing

open space preservation programs, such as the Pinelands National Reserve.

In a study of the differences in voting behavior on environmental ballot measures
between states, Salka evaluated county-wide voting data on state environmental measures
in California, Colorado, Florida, Michigan, and Oregon. The primary question sought to
be answered was whether or not the presence of resource-dependent industries had
significant influence on voting behavior. A regression was used to correlate county
voting data and socioeconomic and political variables. In summary, Salka found support
for the Republican Party was negatively correlated, while a high level of education, and

median income were positively correlated. Resource dependence, the primary subject of

27

the study, was found to be much more influential in western states than in non-western

states.

Arguably the most detailed study was conducted by Nelson, Uwasu, and Polasky
in 2007. They evaluated 718 municipal-level referenda that took place between 2000 and
2004 and, in addition to asking what factors affect the success of open space referenda,
also asked what factors increase the chances of an open space referenda taking place.
Variables considered included population density, tenure, percent change in number of
households, percent change in farmland and the share of votes cast for Bush, Gore, and

Nader in the 2000 presidential election. In summary, Nelson et al. found:

“Our model estimates indicate that richer, more educated, faster growing areas —
both in the municipality and in its surrounding environs — and environmentally
concerned municipalities are much more likely to have open space referenda and
subsequently pass the referenda...1ncreasing taxes to pay for open space
significantly decreased support for open space referendum relative to bond

financing or financing from existing taxes.” (Nelson, Uwasu, & Polasky, 2007, p.

591)

Additionally, Nelson et al. found no evidence of selection bias among communities
considering an open space preservation measure. “It appears that controlling for

selection bias is relatively unimportant in analyzing referenda results.” (Nelson, Uwasu,

& Polasky, 2007)

Similarly, Kotchen and Powers (2006) considered all national open space

preservation votes occurring between 1998 and 2006, and specifically analyzed all votes

28

taking place within New Jersey and Massachusetts. They tested many of the same
variables tested by Nelson et al., in addition to a few others and unsurprisingly, many of
the same conclusions were drawn. “Across the nation, jurisdictions that have held open-
space referenda differ from national averages in several respects. They tend to have
greater population growth, greater household incomes, greater home values, and greater
home ownership rates.” (Kotchen & Powers, 2006, p. 388) The primary difference
between Kotchen and Power’s analysis and Nelson et a1. is the attention paid to financing
mechanisms by Kotchen and Powers. When evaluating the effect of financing
mechanism, and relative cost to voters, the relationship between support and cost was
found to be concave, not linear. “At the local level—where spillin effects are likely to be
small—higher funding rates tend to decrease voter support. At the state-county-level—
where spillin effects are likely to be large—higher funding rates tend to increase voter

support.” (Kotchen & Powers, 2006, p. 388)

In two studies focused on the influence of the environment as a political issue,
Anderson and Mizak (2006) and Davis and Wurth (2003) consider socioeconomic
variables as well. Anderson and Mizak analyzed the influence of the socioeconomic
characteristics of the constituencies of United States Senators on their voting patterns,

finding,

“the characteristics of districts and states, including the percentage of single
people and the income per capita — items that are more reflective of urban rather
than rural areas — are strong predictors of how a representative will vote. The
strongest predictor that we find is whether or not the elected official is a

Democrat.” (Anderson & Mizak, 2006, p. 155)

29

Davis and Wurth considered similar questions, but used voter survey results following
the 1996 presidential election. In addition to a variety of other predictors of support for
presidential candidates, environmental spending was also considered. “Most important
for our purposes, the results of the logit analysis support the hypothesis that the
environmental spending variable is a statistically significant predictor of voter choice
between Clinton and Dole.” (Davis & Wurth, 2003, p. 737) Additionally, it was found
that pressing for economic sacrifice on behalf of the environment was a dangerous

political strategy.

Land use change and development pressure

Kline and Wichelns 1994 developed an economic model to “describe public
support for farmland preservation programs as a function of local land use patterns and
socio economic data.” (Kline & chhelns, 1994, p. 223) The final conclusion of the
model was development pressure is the best predictor of successful PDR referenda. Two
versions of the model were created, one for Pennsylvania, and one for Rhode Island, the
percentage of “yes” votes in local conservation referenda in was used as the dependent
variable for both. Independent variables used were the percent of land in farms, percent
change in population, and percent change in land and house values. Income and
education variables were dropped from the model because they were not significant.
“Results of this study suggest that PDR funding is most likely to be approved in places
with the greatest increase in population and the greatest increase in land and house

values.” (Kline & chhelns, 1994, p. 232)

30

Pfeffer and Lapping 1995 considered the influence of type of farm on the success
of PDR measures. The change in two types of farms, dairy and fruit was considered, in

addition to change in population, and other socioeconomic variables.

“fruit farms are of greatest relative importance in predicting public PDR support,
followed by diary and then population. The effects of these changes are
independent of the statistically significant control variables in the model: county
population size, percentage of land in farms, whether the planner works as a
consultant to a public or private organization, and whether a PDR program is

available in the county.” (Pfeffer & Lapping, 1995, p. 43)

The fruit farm versus dairy farm predictor indicates that when a jurisdiction provides
residents with an ideal farm use (a beautiful, non-intrusive orchard) support is high,
however, when a less-than desirable dairy farm is considered, preservation is not
supported at the same rate. Pfeffer and Lapping concluded that a large portion of the
popularity of PDR in the northeast stemmed from development pressure in urban fringe
areas. “population growth stimulates demand for PDR, presumably as nonfarm residents
strive to maintain the rural ideal they sought in moving to the area.” (Pfeffer & Lapping,

l995,p.44)

In a national study, Romero and Liserio (2002) attempted to determine whether
sprawling communities were more likely to consider open space conservation measures,
or if open space measures were a result of media “hype.” In order to do so they
compared all communities considering an open space measure with a control group of

“sprawling communities” as defined by the Sierra Club’s 1998 national report on the

31

subject. The final conclusion of the project was that, “smaller, wealthier, and whiter
areas were the most likely to include open-space preservation measures on 1998 and
1999 ballots, regardless of whether actual sprawled development was apparent.” (Romero
& Liserio, 2002, p. 343) However, the primary measure of a community’s level of sprawl
was population density, which posed significant problems when comparing counties to

cities and townships.

In response, Howell-Moroney tested the same hypothesis, using land use change
analysis in addition to population density as a measure of urban sprawl. However, he
limited his study to the Delaware Valley region, and included all municipalities, to
account for the possible selection bias that exists when considering only communities that
voted on a ballot measure. His results were quite different from Romero and Liserio’s.
“. . .population density is negatively related to open-space vote probability and is
statistically significant. This is the exact opposite of Romero and Liserio’s original
result...” (Howell-Moroney, 2004a, p. 175) Howell-Moroney also accounted for the
problem of using density as a comparison between jurisdictions by using land use change
as an independent variable in a regression analysis. The result was that communities
experiencing sprawl as defined by loss of open space and increase in acreage of single
farIlily dwellings were more likely to hold an open space vote. However, Howell-

MOl‘oney does not dismiss the potential for “hype” to influence votes on sprawl.

“. . .it is certainly possible that citizen irrationality can and does exist as it relates
to sprawl, but actual land use patterns and state incentives have the effect of
spuning the planning bureaucracies in communities to initiate the process for

getting open-space votes on the ballot.” (Howell-Moroney, 2004a, p. 178)

32

The influence of local planning bureaucracies and state programs encouraging
preservation programs are often ignored in studies of open space preservation measures,

which Moroney investigated further.

In another study published in 2004, Howell-Moroney again considered the
prevalence of open space votes in the Delaware Valley region. Using logistic regression,
he found “higher levels of community socioeconomic status and higher levels of
community growth were both associated with higher likelihoods of open space policy
adoption.” (Howell-Moroney, 2004b, p. 116) However, Howell-Moroney draws an
additional inference from the data unique to his study. He argues that the prevalence of
local open space preservation policies has a negative effect on efforts to establish
regional planning. “If sprawl has the ability to demonstrate the efficacy of regionalism, it
also has the ability to further entrench the localism of the status quo through locally based

solutions.” (Howell-Moroney, 2004b, p. 11.7)

In addition to studies considering the effect of community characteristics on the
success of open space preservation measures, contingent valuation studies, measuring
People’s willingness to pay must also be considered. All such studies reviewed found at
least some evidence to prove there are public benefits to private landscapes and
environmental amenities or assume that such goods contain public benefit (Lindsey &
Knapp, 1999; Schlapfer, Roschewitz, & Hanley, 2004) These studies are relevant because
they evaluate willingness to pay based on socioeconomic data, “support for greenway
projects, measured as willingness to pay and as willingness to donate. . .was greater
amOng property owners than renters and greater among those who lived in the corridor

than among those who did not.” (Lindsey & Knapp, 1999, p. 309) Other studies found

33

education, gender, urban residence, farm occupation, use of parks, and distance to an
urban growth boundary to be predictive of voting behavior. (Schlapfer, Roschewitz, &
Hanley, 2004; Vossler, Kerkvliet, Polasky, & Gainntdinove, 2003) Additionally, all three
studies cited herein found respondents to contingent valuation surveys consistently
reported a higher willingness to pay than they expressed when given the opportunity to
contribute or vote on a tax increase. “In all observed voting choices we identified as
‘inconsistent,’ stated willingness to pay was not marginally but indeed around 10 to 20
times higher than the costs of the (rejected) actual proposition.” (Schlapfer, Roschewitz,

& Hanley, 2004, p. 14)

34

CHAPTER 3: METHODOLOGY

The hypothesis of this research is that the success of Michigan PDR millage
votes, like the ballot measures discussed in Chapter 2, is tied to the characteristics of the
community considering the millage. Specifically, that higher socioeconomic status and
greater development pressure create support for PDR millages in Michigan. In order to
test this hypothesis a dataset has been created consisting of all minor civil divisions that
have considered a PDR millage in Michigan. A set of independent variables consisting
of statistics relating to community characterstics, development pressure, and political
circumstances was then chosen based on the literature review. The percent of “yes” votes
in each MCD is used as the dependent variable. A bivariate correlation is used to
evaluate the effect each independent variable has on the dependent variable, a stepwise
regression based on select variables and a linear regression using all variables determines
the degree to which the most significant variables as well as the entire model explain the

variance in Michigan PDR millage votes.

Datasets

The most challenging portion of this project was developing a conclusive set of
Michigan jurisdictions that had voted on a PDR millage. In order to ensure no
jurisdiction was missed, the Trust for Public Land’s LandVote database was used in
addition to discussions with state experts in land use and land preservation. The initial
set of jurisdictions was created without assistance from the TPL’s LandVote database,
but the database was used for verification purposes. The database contained two

townships (Berlin and Augusta) that were not part of the original dataset.

35

As of January 2009, eighteen Michigan jurisdictions have voted on nineteen PDR
millage proposals (Peninsula Township approved a millage in 1994 and again in 2002).
Although this is a fair number of millage votes to consider qualitatively, the dataset is too
small to gather significant quantitative information from. The solution to this problem
lies in one of the primary difficulties of working with the data, the variety of jurisdiction

types (Table 3.1).

Table 3.1: Jurisdictions and rate of success

N umber of PDR l\'um her of PDR

 

 

 

 

Jurisdiction 'I’ypc Millage Votes Millages Adopted Success Rate
County 4 l 25%
Township 14 7 50%
City 1 l 100%
Total 19 9 47%

 

 

 

 

 

 

Of the eighteen jurisdictions, four are counties. When county results are
disaggregated, a consistent unit of analysis, Minor Civil Division (MCD), can be used
throughout. All data available for an MCD considering a millage is also available for an
MCD considering a county-wide millage, including vote totals by precinct. When the
county vote results are replaced with individual MCD results, seventy-seven jurisdictions
are added to the dataset (Appendix A lists all county MCDs). This provides distinct
advantages to considering the relationship between percent yes votes and community
characteristics in the initial nineteen jurisdictions. By disaggregating county vote results,
the problems present in Romero and Liserio’s (2002) comparison of density between

counties and cities is avoided and the model’s level of statistical significance increases

36

greatly. The final data set has a population of 92 (the four counties are dropped when all

MCDs within each county are added (77+15=92)).

The use of MCDs also creates an opportunity for a unique analysis of select
jurisdictions in Washtenaw County and one in Ingham County. Washtenaw County
considered a county-wide PDR millage in 1998 that was unsuccessful. Since that time
five townships and the City of Ann Arbor have considered a millage to purchase
development rights. The same situation exists for Meridian Township in Ingham County;

it approved a millage in 2000 and again considered a county-wide millage in 2008.

In order to consider development pressure influences, a second dataset is created
at the county-level. Ideally, land use change information would be considered according
to each MCD, but there is no reliable land use change data for the entire state of
Michigan that is applicable to the time period considered by this project. Therefore,
county-level land use change data provided by the Census of Agriculture is used instead.
The mean percent “yes” votes in each county serves as the dependent variable for the

county dataset.

Analysis

The final MCD dataset consisting of 92 jurisdictions is used in a bivariate
correlation and two regression analyses to determine the relative effect of the independent
variables on the percent of “yes” votes in each jurisdiction. Romero and Lisero (2002),
Salka (2003), and Solecki, Mason, and Martin (2004) used linear regressions to evaluate
results. In this case a bivariate correlation and stepwise regression are used to control for

the effect of multicollinearity. Multicollinearity occurs when independent variables are

37

highly correlated with each other, causing regression coefficient values to vary
enatically. Because the socioeconomic data contained herein is intrinsically related,
removing the influence of multicollinearity is a must in order to make reasonable
conclusions about the influence of individual variables on percent “yes.” Importantly,
multicollinearity does not affect the predictive power of a model as a whole, it does
greatly reduce the accuracy of individual coefficients within a model. Appendix B
contains a correlation matrix of all independent variables, and illustrates the high level of
correlation between them. Therefore, significance level, correlation values, coefficient
signs, and coefficient values are used indicate the relative influence of each variable on

support for Michigan PDR millage proposals.

The process of separating the influence of variables in order to obtain a
reasonable model consisted of three steps. This is a method not used in the literature
reviewed for this project. Due to unique characteristics of the data (small sample size,
etc.) compared to other similar studies, multicollinearity must be addressed. First, a
bivariate correlation indicates which independent variables have a significant correlation
with percent “yes.” Second, variables with a significant correlation are then separated
from those without a significant correlation to remove some unnecessary noise in the
data. Third, a linear regression is employed using percent “yes” as the dependent
variable and the independent variables selected in the second step. Table 3.2 contains a
description of each variable, its source, the subject it addresses, and the expected sign of

the correlation value.

Other studies of open space ballot measures and community characteristics have

employed a “log-odds ratio” of the percent of “yes” votes as a dependent variable

38

Variable

Table 3.2 Independent variable description

Sourcc

.-\rca Addressed

thpcctctl

 

Sign

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Total Population 2000 and 1990 US Community Characteristics +
Census

Percent of population 2000 and 1990 US Community Characteristics -

rural Census

Percent of population 2000 and 1990 US Community Characteristics -

living in a different Census

county five years prior

Educational attainment 2000 and 1990 US Community Characteristics +
Census

Median household 2000 and 1990 US Community Characteristics +

income Census

Percent of owner- 2000 and 1990 US Community Characteristics +

occupied housing Census

Median age 2000 and 1990 US Community Characteristics -
Census

Median home value 2000 and 1990 US Community Characteristics +
Census

Percent change in 1980, 1990, and 2000 Development Pressure +

population US Census

Change in Percent Rural 2000, 1990 and 1980 Development Pressure +
Census

Percent change in home 2000, 1990 and 1980 US Development Pressure +

value Census

Percent change in 2000, 1990 and 1980 US Development Pressure +

number of households Census

Population density 2000 and 1990 US Development Pressure +
Census and Michigan
Center for Geographic
Information

Percent change in land 1987-2007 US Census Development Pressure +

in farms" of Agriculture

Percent change in value 1987-2007 US Census Development Pressure +

of farmland“ of Agriculture

Voter tum-out Individual County and Political -
Township Clerks Offices

Percent of voters not Michigan Secretary of Political -

casting a vote for the State

millage

Percent voting for Bush Michigan Secretary of Political -

in 2004 State

 

 

 

*Available only at the County-level (this data is also available for the few major cities included in the
study, but in order to maintain a consistent level of analysis, the county-level is used).

39

(Nelson, Uwasu, & Polasy, 2007; Kahn & Matsusaka, 1997; Kline & chhelns, 1994)
This is a form of a model originally employed by Deacon and Shapiro (1975) that allows
researchers to make inferences about voter preferences from aggregated election results
and employ a logistic regression. (Kotchen & Powers, 2006) The purpose of the present
research is not to understand voter preference. This research is concerned with
characteristics of minor civil divisions that affect the percent of “yes” votes. Individual
voter preference is not considered; therefore the log-odds ratio is not used as the

dependent variable.

In addition, the same three steps are used to evaluate the county-level data to
consider land use change data that cannot be evaluated at the MCD level. The mean
percent “yes” value of all votes in each county in which a PDR vote has taken place is
used as the dependent variable and percent land in farms change from 1987 to 2007 and
change in value of farmland per acre from 1987 to 2007 are used as independent
variables. Table 3.3 displays the mean vote result in each County. The alternative for
aggregating vote results to the county-level is the success rate of votes within each

county; this was not used due to the number of counties with a success rate of zero.

Table 3.3: Development pressure county-level data

Anth 0.430569667
Grand Traverse 0.54445
In 0.53035

0.4136
Leelanau 0.3786
Monroe 0.220984215 0
Washtenaw 0.597914388 0.666666667

 

40

Variables

In addition to using the TPL’s database, voting data was gathered by contacting
the clerks of local jurisdictions to obtain voter registration data and ballot language.
Total ballots cast, used to calculate voter tum—out, was primarily taken from the Michigan
Secretary of State’s office website. The remaining variables were gathered from a
combination of the 2000 and 1990 Census and 1987-2007 Census of Agriculture. (See

Table 3.3)

Based on the literature, three general categories of variables affecting vote
outcome were identified: 1) Community Characteristics, 2) Development Pressure, and 3)
Political Influences. Community Characteristics includes typical socioeconomic statistics
and other statistics like total population and median age that describe community type.
Development pressure is difficult to measure in this context because consistent and
reliable land use/land cover change data is not available for all townships in Michigan
and Census of Agriculture land use data is only available at the county-level.
Consequently, proxies for development pressure based on change in population, number
of households, and home value are used in the MCD model and a separate model based
on county-level development pressure data is used to consider land use and land value
changes. Additionally, the change in percent of population rural is used to illustrate rate
of urbanization, which was not done in the studies reviewed for this research. Of the
three categories, the third is undoubtedly the most difficult to investigate, and nearly
impossible to quantify with any significance due to the relatively small dataset of

nineteen jurisdictions holding a PDR millage vote in Michigan. Unlike census statistics,

41

there is a limited ability to disaggregate political variables because the unit of analysis is

the jurisdiction holding the vote.

As a result of discussions on methodology in Kline and Wichelns (1994) and
Kotchen and Powers (2006) addressing the problem of abstention in open spaces votes,
the percent of voters not casting a vote for a particular millage is added as a variable.
This variable is created by subtracting the number of votes for and against a millage from

the total number of ballots cast.

Some discussion of data sources used for the independent variables in Table 3.3 is
necessary due to the span of time (fourteen years) considered in this analysis. 2000 US
census data is used for all socioeconomic variables except for those associated with the
Peninsula Township 1994 millage vote, for which 1990 data is used. Conversely, the
2000 census is used to evaluate the 1998 Washtenaw County millage because it provides
a more accurate description of the County at the time of the vote than the 1990 census.
Variables related to development pressure all measure the relative change of a value
between 1990 and 2000, with the exception of Peninsula Township 1994, which uses
1980 and 1990 census data to measure change. The age of socioeconomic data compared
to the date of millage vote is a concern for the most recent votes (see Table 1.3);
however, American Community Survey data is only available for places with a

population of 65,000 or more, which eliminates eighty-nine of the ninety-two MCDs.

42

CHAPTER 4: RESULTS

Variable abbreviations were developed in order execute the three step process for
testing the hypothesis that the success of Michigan PDR votes is positively related to
socioeconomic status and development pressure. Table 4.1 explains independent variable
abbreviations used in the analyses. The results of the tests run on each data set are then
detailed in the subsequent sections, first the results of the MCD dataset are given,
followed by the results of the County dataset, and finally the results are given of the

hypothesis as it relates to the three variable areas examined.

Table 4.]: Independent variable abbreviations

 

.\bbrc\iation Variable

 

 

 

 

 

 

 

 

Total Pop. Total population

% Rural Pop. Percent of population rural

% Rural Pop. Change Percent change in rural population 1990-2000

% Mobility Percent of population living in a different county five years
prior

Educational Attainment Percent of population with a bachelors degree or higher

Median Income Median. household income

% Owner Occupied Percent of owner-occupied housing

Housing

Median Age Median age

Median Home Value Median home value

 

% Change Home Value Percent change in home value 1990-2000
% Change Households Percent change in number of households 1990-2000

 

 

 

 

 

 

Density Persons per square mile

% Not Voting Percent of voters not voting on the millage

Voter Turn-out Percent of registered voters voting in the election
Bush Vote Percent of votes for George W. Bush in 2004

% White Percent of White population

 

% Change Farmland Acres Percent change in acres of farmland 1987-2007

 

 

 

 

% Change Farmland Value Percent change of value per acre of farmland 1987-2007

 

43

MCD Dataset Results

Three separate tests were run on the MCD dataset. The results of the first, a
bivariate correlation analysis are contained in Table 4.2. The bivariate correlation
evaluates each variable independently, permitting one to consider the importance of each
relative to the percent of “yes” votes (hereafter simply “Yes”). Of the sixteen variables
considered in the MCD dataset, nine (Total Pop. % Rural Pop., % Mobility, Educational
Attainment, % Owner Occupied Housing, % Change Home Value, Density, Bush Vote,
and % White) are significant at the .01 level and one (Home Value) is significant at the
.05 level. The correlation value for Educational Attainment (.542) is the closest to 1 or -

1, the next-closest is % Owner Occupied Housing at -.472.

The MCD stepwise regression used a data set consisting of the above ten
variables. The most significant variable when the entire model is considered is
Educational Attainment; it alone explains 29% of the variance in Yes (Table 4.3). %
Owner Occupied Housing is the only other independent variable chosen for the stepwise

regression, it explained an additional 8% of the variance in Yes.

Finally, the MCD linear regression using all variables produced an adjusted R2
value of .375, compared to the stepwise regression adjusted R2 value of .361. This means
the addition of the remaining fourteen variables explains only 1.4 percent of the variance
in Yes. Tables 4.2 through 4.5 contain results of the MCD dataset analysis. Coefficient
values from the linear regression are not considered due to the high degree of

multicollinearity in the independent variables.

44

Table 4.2: MCD bivariate correlation results

  

Variable Pearson (‘orrclation Sig. (Z—tailcd)

     

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Total Pop. 361 ** .000 92
% Rural Pop. -.423** .000 92
% Rural Pop. Change -.073 .491 92
% Mobility .354** .001 92
Educational Attainment .542** .000 92
Median Income -.064 .544 92
% Owner Occupied Housing -.474** .000 92
Median Age -.156 .139 92
Median Home Value .261* .012 92
% Change Home Value -.370** .000 92
% Change House Holds .140 .184 92
Density .358** .000 92
% No Vote -.036 .731 92
Voter Tum-out .007 .944 92
Bush Vote -.358** .000 92
% White -.271** .009 92

 

 

 

 

 

 

*Correlation is significant at the .05 level. MCorrelation is significant at the .01 level.

Table 4.3: MCD stepwise regression model

('orrclatcd \l(‘l) Sicpnise Regression

 

Change Statistics

 

 

 

 

Adjusted R Std. Error of R Square
Model R R Square Square the Estimate Change F Change
1 .542"1 .294 .286 .08866015065 .294 37.494
2 .613b .375 .361 .08386193587 .081 11.593

 

 

 

 

 

 

 

 

a. Predictors: (Constant), Educational Attainment
b. Predictors: (Constant), Educational Attainment, % Owner Occupied Housing

45

Table 4.4: MCD stepwise regression coefficients

 

(‘orrclatctl .\l('l) Stcpnisc Regression

 

 

 

 

Unstandardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
2 (Constant) .523 .060 N/A 8.649 .000
Educational .253 .054 .422 4.641 .000
Attainment
%Owner Occupied -.217 .064 -.310 -3.405 .001
Housing

 

 

 

 

 

 

 

 

 

Dependent Variable: Yes

Table 4.5: MCD linear regression model

Adjusted R Std. Error of the
Model R R Square Square Estimate

1 .702 .493 .375 .08148296308
Dependent Variable: Yes

 

 

 

 

 

 

 

 

County Dataset Results

The same process used to evaluate the MCD dataset is used on the County
dataset. However, the dependent variable is different. The mean Yes value in each of the
seven counties considered is used as the dependent variable in place of Yes. The
correlation test indicates that one of the two independent variables, % Change in
Farmland Value, is significant. Furthermore, % Change in Farmland Value has a
correlation value very close to -1, of -.816. The other variable, % Change in Farmland
Acres is not significantly correlated, but it too has a negative correlation value. Table 4.6

contains the County dataset correlation values.

46

Table 4.6: County bivariate correlation results

 

 

Variable l’carson ('orrelation Sig; (2-tailctl)
% Change Farmland -.201 .665 7
Acres
% Change Farmland -.816* .025 7
Value

 

 

 

 

 

 

*Correlation is significant at the .05 level

Because only one of the variables in the County dataset was significantly
correlated with Yes, the stepwise regression had only one variable to choose from.
However, % Change Farmland Value was significant, and when used alone in a
regression analysis creates a model significant at the .01 level with an adjusted R2 value
of .598. When % Change Farmland Acres is added in a linear regression the adjusted R2
value increases to .743; meaning although % Change Farmland Acres is not significant it
adds explanatory power to the model, increasing the variance in mean Yes explained by
approximately 15%. Tables 4.7 and 4.8 contain the County stepwise regression results
and Table 4.9 contains the linear regression model results as well as all other model’s

significance values.

Table 4.7: County stepwise regression model

 

('ount) Stcpnisc Regression

Change Statistics

Adjusted R Std. Error of R Square
Model R R Square Square the Estimate Change F Change

1 .816a .665 .598 .0803114 .665 9.937
Predictors: (Constant), % Change F arrnland Value

 

 

 

 

 

 

 

 

 

 

 

47

Table 4.8: County stepwise regression coefficients

('ounty Stcpnisc Regression

 

Unstandardized Standardized
Coefficients Coefficients

 

 

 

Model B Std. Error Beta t Sig.
l (Constant) .783 .1 1 1 N/A 7.032 .001
% Change Farmland Value -.2708 .086 -.816 -3.152 .025

 

 

 

 

 

 

 

 

 

a. Dependent Variable: Mean Yes

Due to population size, the significance level of the MCD linear regression model
is higher than the County linear regression model. However, the adjusted R2 value of the
County linear regression model is .743, and the adjusted R2 value of the County stepwise
regression is .598, both of which are higher than the MCD models (.375 and .361
respectively). The sum of these results is that the MCD model explains less of the

variance in Yes than the County model explains in mean Yes.

Table 4.9: Model significance

 

 

 

 

\lotch 3 3 .ququ R2 ,
Stepwise MCD 92 .375 .361 .000
Stepwise County 7 .665 .598 .001
Linear MCD 92 .493 .375 .000
Linear County 7 .828 .743 .029

 

 

 

 

 

 

 

The above results are mixed as they relate to the hypothesis that Michigan PDR
millage votes are influenced positively by socioeconomic status and increased
development pressure. In the next three sections, the results are evaluated by variable
category based on coefficient values and significance levels in the bivariate correlation

and stepwise and linear regressions.

48

Community Characteristics

In total, nine community characteristic variables were considered. Of those,
seven (Total Pop., % Rural Pop., % Mobility, Educational Attainment,% Owner
Occupied Housing, Median Home Value, and % White) have a significant correlation
with Yes. (Table 4.10) However, when the hypothesis that Yes is positively correlated
with higher socioeconomic status is tested based solely on coefficient signs it fails.
Values typically positively related to high socioeconomic status; % Owner Occupied
Housing, Median income, and % White, are negatively correlated with Yes.
Additionally, Median Age and % Mobility were positively correlated with Yes, a result

that was not anticipated.

Table 4.10: Community characteristics variables

 

 

 

 

 

 

 

 

 

Variable Correlation (,‘orrclation Stepwise Expected
Sig. Coefficient Sign

Educational Attainment** .000 .542 .253 +
% Owner Occupied .000 -.474 -.217 +
Housing"

Total Pop.” .000 .361 N/A +
% Pop. Rural""II .000 -.423 N/A -
% Mobility“ .001 .354 N/A -
% White" .009 -.271 N/A +
Home Value* .012 .261 N/A +
Median Age .139 -.156 N/A -
Median Income .544 —.064 N/A +

 

 

 

 

 

 

 

*Significant at the .05 level. "Significant at the .01 level.

Of the nine socioeconomic variables, two are significant in the stepwise

regression, each at the .01 level (Table 4.10 and 4.4). Interestingly, the correlation values

49

of the two significant variables provide conflicting conclusions when testing the
socioeconomic status hypothesis. If Yes is correlated with higher socioeconomic status,
it should have risen with Educational Attainment and % Owner Occupied Housing.
Instead, there is a significant negative correlation between % Owner Occupied Housing
and Yes. For every 1% increase in owner-occupied housing the percent Yes is predicted
to decrease by .217%. Educational Attainment, in contrast, has the highest level of
significance, and it is positively correlated. Every 1% increase in Educational Attainment

causes a .253% increase in Yes.

Development Pressure

Development pressure variables are included in the MCD dataset as well as the
County dataset. In the MCD bivariate correlation, three of the four variables relating to
development pressure exhibit the expected sign. All variables are meant to positively
measure development pressure, meaning as they increase so too does the pressure to
develop land. The unexpected and significant negative correlation exhibited by %
Change Home Value brings the hypothesis that development pressure is positively related

to PDR millage support into question. (Table 4.11).

No development pressure variables were included in the MCD stepwise
regression and of the four, only % Change Home Value and Density were included in the
dataset used for the MCD stepwise regression (because they were the only significantly
correlated variables). However, the significant correlation values of % Change Home

Value and Density are of worth noting.

50

Table 4.11: Development pressure MCD regression variables

 

 

 

 

Variable Correlation Correlation Stepwise Expected
Sig. Coefficient .‘ g
% Change Home Value** .000 -.455 -.082 +
Density** .000 .358 N/A +
% Change Households .169 .145 N/A +
% Change Rural Pop. .491 .073 N/A +

 

 

 

 

 

 

 

”Significant at the .01 level

Development pressure in terms of land use change is measured using data from
the Census of Agriculture. The change in acres of farmland in each county is negatively
correlated with mean Yes, but not significant; whereas the change in value of farmland
by acre is negatively related to mean Yes in the stepwise regression and significant at the
.05 level, an unexpected result (Table 4.12). For each 1% increase in the percent change
of farmland by acre between 1987 and 2007 Yes is predicted to decrease by .270%. In
other words, as the rate of increase of farmland value becomes greater, support for PDR

millages lessens in Michigan.

Table 4.12: County-level development pressure variables

 

 

Variable Correlation Correlation Stepnise Expected
Sig. Coefficient Sign
% Change Farmland Value .025 -.816 -.270 +
% Change Farmland Acres .665 —.201 N/A +

 

 

 

 

 

 

 

“Significant at the .05 level.

51

Political Variables

Three political variables were considered in the bivariate correlation. Of the
three, % No Vote and Voter Tum-out exhibited the expected negative relationship (Table
4.13). Bush Vote is negatively correlated with Yes, and is significant. No specific
hypothesis was formulated for the political variables, but the significant relationship with
Bush_Vote is of interest because the results of previous studies are mixed. Some found
political party affiliation to be a significant determinate for support of land preservation
(Anderson & Mizak, 2006; Deacon & Shapiro, 1975; Salka, 2003) while others found
political preference to be unrelated (Kahn & Matsusaka, 1997; Nelson, Uwasu, &

Polasky, 2007).

Table 4.13: Political variables

 

 

 

 

 

 

 

 

 

Variable Correlation Correlation Stepwise Expected
Sig. Coefficient Sign
% Not Voting .731 -.036 N/A -
Voter Tum-out .944 .007 N/A -
Bush V0te** .000 -.358 .291 -
“Significant at the .01 level.

 

Although the sum of these results rejects the initial hypothesis that Yes will
increase as development pressure and socioeconomic status increase, by no means are the
results inconclusive. The final chapter discusses these results, makes suggestions for

PDR millage campaigns based on them, and identifies future areas of research.

52

CHAPTER 5: DISCUSSION

This chapter consists of four parts. It begins with a general discussion of the
results detailed in Chapter Four and possible limitations of those results, areas for future
research are discussed next, and the final section contains an overall conclusion from the

project.

Discussion of Results

The hypothesis that Yes is positively related to high socioeconomic status and
development pressure must be rejected based on the information provided herein. This
indicates one of two things, either the factors affecting PDR support in Michigan are
different from those in other parts of the country or this study’s methodology is different

enough from previous studies to produce conflicting results.

The significant community characteristic variables in the bivariate correlation
indicate that larger communities with little rural population, a relatively high rate of
mobility, a well educated population, a low rate of owner occupancy, or high median
home value have been most likely to support PDR millages in Michigan. Significant
development pressure variables in the MCD and County datasets indicate that
communities with home values increasing at a slow rate and high population density
located in counties where land values are rising relatively slowly have been most likely to
support PDR millages. The lone significant political variable, Bush Vote, indicates that
communities that support democratic candidates for president have been more likely to

support PDR millages. No specific hypothesis was formulated for the political variables,

53

but the significant relationship with Bush Vote is of interest because the results of
previous studies are mixed. Some found political party affiliation to be a significant
determinate for support of land preservation (Anderson & Mizak, 2006; Deacon &
Shapiro, 1975; Salka, 2003) while others found political preference to be unrelated (Kahn

& Matsusaka, 1997; Nelson, Uwasu, & Polasky, 2007).

The most important variables when all MCDs are considered are Educational
Attainment and Owner Occupancy, which combined explain 36% of the variance in Yes.
Whereas % Change in Farmland Value explains 59% of the variance in the aggregated

County dataset.

While many of these variables reflect the expected relationship, the negative
relationship of % Mobility, % Owner Occupied Housing, % Change Home Value, and %
Change Farmland Value do not. This is indicative of PDR millage support in Ann Arbor
in the city and Washtenaw County votes (67 and 54% yes) and in East Lansing in the
Ingham County vote (71% yes), two cities that exhibit all of the aforementioned
characteristics. The only other city with a population over 40,000, Lansing, also

supported a PDR millage (52% yes).

In the MCD and County stepwise regressions these results are repeated.
Educational Attainment is positively related to Yes (as anticipated), is the most
significant variable, and explains the most variance. Conversely, the negative
relationship of % Owner Occupied Housing to Yes is unexpected. The negative
relationship between Yes and % Owner Occupied Housing in the MCD stepwise

regression may be indicative of a broader trend signifying urban support for PDR

54

millages. Total Population and Density are positively correlated with Yes, and % Owner
Occupied Housing, Bush Vote, and % White are negatively correlated, all of which
typifies urban characteristics. Additionally, these five of variables are significantly
correlated with each-other, with correlation values above .6. This emphasizes the need
for controlling multicollinearity in the model, and supports the use of Owner Occupancy
as a variable in the MCD stepwise regression because it measures the same variance as

other urban indicators.

An additional qualitative analysis of the County dataset indicates that proximity to
an urban area may have a positive effect on Yes. The three counties with a mean Yes
value above fifty percent, Ingham, Washtenaw, and Grand Traverse contain two cities
with a population over 100,000 (Lansing and Ann Arbor) and three cities with a
population between 10,000 and 100,000 (Ypsilanti, East Lansing, and Traverse City).
Whereas the four counties with a mean Yes vote under fifty percent contain only one city

with a population over 10,000 (Monroe City in Monroe County).

The results of tests run on the County dataset, appear to conflict with the urban
hypothesis, as farmland prices generally increase with proximity to urban areas but are
negatively correlated with Yes. This means the greater the increase in value of farmland
in a county the less support exists for PDR millages. In fact, the correlation value
between land value change and percent “yes” was the correlation value closest to one or
negative one (.-816, Table 4.6). Alone, this variable explains almost 60% of the variance
in mean percent yes on PDR votes in the seven counties considered. However, a look at
the data indicates the more urban counties studied in this context saw a smaller increase

in farmland value than more rural counties. Washtenaw and Ingham counties saw an

55

increase in farmland per acre between 1987 and 2007 of between 80 and 90 percent,
whereas more rural counties like Monroe and Antrim saw increases of over 150 percent
(Table 5.1). The tendency of jurisdictions experiencing increases in farmland value not
to support PDR millages reflects a common perception that farmers are the most
burdened by property tax increases used to pay for the PDR programs, and therefore do

not vote in favor of tax increases.

The influence of urbanity on the datasets prompted an additional analysis. While
East Lansing and Ann Arbor are unique, in many ways their respective counties, Ingham
and Washtenaw, are also unique when compared to the rest of the data. Both Ingham and
Washtenaw counties have significantly higher populations than the other five counties
considered (see Table 5.1), and the vast majority of the MCDs located within each can be
considered urban or suburban, or at least in close proximity to a true urban area; whereas
the remaining five counties are predominantly rural. Therefore, the MCD dataset was
divided into two separate datasets. The Urban MCD dataset consists of all MCDs in
Washtenaw and Ingham County (N=52) and the Rural MCD dataset consists of all

remaining MCDs in Monroe, Lapeer, Grand Traverse, Leelanau, and Antrim Counties

Table 5.1: Farmland value change

Jurisdiction Mean Ves Population "/6 Change Farmland

 

 

 

 

 

 

 

 

 

Value
Washtenaw 0.5979 322,895 80.0%
Ingham 0.5303 279,320 89.2%
Lapeer 0.413 87,904 101.3%
Grand Traverse 0.5444 77,654 119.8%
Leelanau 0.3786 21,119 139.6%
Antrim 0.4305 23,110 164.7%
Monroe 0.2209 145,945 180.7%

 

 

 

 

56

(N=40). This division will evaluate the influence of the urban areas on the data to

determine if different factors affect Yes in urban and rural jurisdictions.

The same three-step methodology was then applied. Table 5.2 contains all

variables that exhibited a significant correlation with percent Yes in the MCD, Urban

MCD, and Rural MCD datasets. The mean population in the Urban MCD dataset is

Table 5.2: MCD datasets comparison

 

 

 

 

 

 

 

 

 

 

 

 

 

\ ariablc \l(‘l) Dataset l rban “(1) Dataset Rural \l('l) Dataset
Pearson Sig. Pearson Sig. Pearson Sig.
Correlation (2-tailed) Correlation (2-tailed) Correlation (2—tailed)

Total Pop. .361** .000 .420“ .002 .271 .099
% Mobility .354** .001 .311* .025 .271 .099
% White -.271** .009 —.327* .018 -.028 .866
% Owner -.474** .000 -.485** .000 -.222 .181
Occupied
Housing

edian Home .261* .012 .296* .033 .037 .826
Value
% Change -.370** .000 -.595** .000 -.342* .035
Home Value
Density .358" .000 .395** .004 .273 .097
Bush Vote -.358** .000 -.523** .000 .243 .141
Educational .542** .000 .663** .000 .260 .114
Attainment
% Rural Pop -.423** .000 -.477** .000 -.367* .024
% Rural Pop. -.073 .491 .114 .420 .452” .004
Change
No Vote -.036 .731 -. 162 .252 323* .048

 

 

 

 

 

 

 

 

* Significant at the .05 level. MSignificant at the .01 level.

57

 

15,144.06 and the mean Yes value is .42, whereas the mean population in the Rural MCD
is 3,568.97 and the mean Yes value is .41. Compared to a mean population of 10,064.93
and mean Yes value of .42 in the MCD dataset. Appendix C contains descriptive

statistics for all datasets.

The result of the Urban MCD and Rural MCD correlations indicate an additional.
phenomenon of the data; that rural MCDs are subject to different influences than urban
MCDs. The Urban MCD correlation finds the same 10 significant variables found in the
MCD dataset, all of which maintain the same sign. In contrast, the Rural MCD
correlation finds 4 variables to be significant, two of which, % Rural Pop. Change and No

Vote, were not significant in either of the other MCD datasets.

The regressions ran on the Rural and Urban datasets also produced results worth
noting (see Table 5.2). Both produced higher linear regression adjusted R2 values, and
the stepwise Urban regression produced a higher adjusted R2 value than the MCD
stepwise regression. The linear Rural MCD regression explains the most variance; it
explains 49.9% of the variance in Yes in rural jurisdictions. The fact that educational
attainment is the only variable to be included in two stepwise regressions is illuminating;
it suggests that Yes in each dataset is subject to different influences. Overall, the
divergent characteristics of the Rural MCD and Urban MCD correlations and regressions
indicate that the level of urbanity in a community has played a significant role in the

approval of PDR millages in Michigan.

In addition to separating urban and rural MCDs, the additional analysis separated

jurisdictions that supported PDR millages (urban) from those that did not (rural). The

58

Table 5.3: MCD datasets regression comparison

 

 

 

 

 

 

\lodcl _ ' ’ Adjusted R2 ,
Stepwise MCD* 92 .375 .361 .000
Linear MCD 92 .493 .3 75 .000
Stepwise Urban” 52 .491 .470 .000
Linear Urban 52 .61 1 .433 .001
Stepwise Rural 40 .359 .352 .000
Linear Rural 40 .715 .499 .006

 

 

 

 

 

 

 

*Predictors: Educational Attainment and Owner Occupancy. ”Predictors: Educational
Attainment and % Change Home Value. ***Predictors: % Rural Pop. Change and No
Vote

Rural MCD dataset contains 5 MCDs, 13% of the dataset, that had a Yes value above
50% (Acme Township, Traverse City, Peninsula Township (twice), and Lapeer City),
while the Urban MCD dataset contains 10 MCDs, 19% of the dataset, with a Yes value
above 50% (Ann Arbor City (twice), Ann Arbor Township (twice), East Lansing City,

Lansing City, Meridian Township (twice), Scio Township, and Webster Township).

The analysis of urban and rural MCDs permits three conclusions. 1) Of all 92
MCDs considered in this project, only 15 (16%) have a Yes value over 50%. This further
emphasizes the role urban areas have played in passing PDR millages in Michigan.
Moreover, in reality, only 11 jurisdictions have supported a PDR millage in Michigan
due to the necessary double-count of jurisdictions that have considered a PDR millage
twice (the Limitations and Future Research sections discuss this in more depth) 2) The
variables considered in the MCD datasets have a different effect on Yes in rural
jurisdictions than in urban and suburban jurisdictions. Specifically, % Rural Pop. Change
and No Vote have the most influence on Yes in rural jurisdictions, but have no significant

influence on Yes in urban and suburban jurisdictions. For each 1% increase in Change in

59

% Rural Pop. Change, Yes is predicted to increase by .249% and each 1% increase in No
Vote causes a 828% increase in the predicted Yes value in rural jurisdictions (see Table
5.2). 3) The only variables to exhibit a significant correlation in all three datasets were %
Change Home Value and % Rural Pop., which are both negatively correlated with Yes,
and significantly positively correlated with each other (see Appendix B). This result
makes obvious sense, areas with a high percentage of rural population did not experience

increases in home value like those seen in suburban and urban jurisdictions in the 1990’s.

Limitations

Before one begins advocating for PDR millage campaigns in and around every
urbanized area in Michigan, there are significant limitations to the above conclusions that
must be discussed. Local ballot proposals are anything but an exact science. First and
foremost, it must be stressed that this research seeks to explain why PDR millages in
Michigan have passed; it does not predict support for them in the future. By no means do
the regression models contained herein provide a model for choosing jurisdictions likely
to pass such millages in the future (although this will make for interesting research when

additional PDR millages occur in Michigan).

To firrther stress the explanatory, not predictive nature of the research, the data set
is somewhat peculiar, and to a degree self-selecting. In order to vote on a PDR millage
some level of support must already exist, otherwise a millage proposal would never be on
the ballot. In addition, it is unlikely that the MCD dataset consisting of 92 local units of
government is representative of the over 1800 local units of government in Michigan.

For an example, East Lansing and Ann Arbor, the proverbial poster children for PDR

60

support in Michigan, are extremely unique cities. Due to the influence of Michigan State
University and the University of Michigan both contain exceptionally large student
populations and have community characteristics not replicated anywhere else in the state.
Although Nelson et a1. (2007) found no evidence of selection bias in their research; this
data is likely subject to such a bias due to its smaller population size and state-specific
nature. Therefore while this information characterizes jurisdictions that previously
considered PDR millages, it may not be useful as a predictor of potential support for a

PDR millage in a particular jurisdiction.

In addition to the explanatory focus of this research, some shortcomings of the
MCD and county datasets must also be considered. A slight possibility exists that the
MCD dataset is incomplete. Regardless of the vigor of the research effort, there is
nothing stopping a sparsely populated township in an out-of-the-way county from quietly
voting on a millage to purchase development rights to land, voting the millage down, and
no one knowing about it other than citizens and township and county clerks. The
Michigan Secretary of State office does not keep election data on local ballot proposals,
nor does it collect information on local or special elections. Therefore it is conceivable
that some jurisdiction has been missed. The County dataset has one obvious issue, the
dependent variable is aggregated. Although the results of the correlation and regressions
ran on it are significant, the aggregation of vote results to the county level does not
consider the jurisdictions that have not considered a PDR millage. For example, Berlin
Township is the only MCD in Monroe County to consider a PDR millage and while the
county land use statistics may be applicable to individual townships, the level of support

for PDR in Berlin Township is not applicable to the county as a whole.

61

A final note on the dataset is that some double-counting necessarily occurred. Six
jurisdictions in this dataset considered both a county and local PDR millage. As a
consequence, the values of all variables are repeated in these instances, with the

exception of Yes, Voter Tum-out, and % No Vote.

The purpose and scope of PDR millages in Michigan vary; in this case the
Meridian Township and the City of Ann Arbor millages are atypical. All other measures

considered in this context have a significant farmland preservation component that is

—-..- r..-“— m1,

absent in the Meridian Township and Ann Arbor programs. The Ann Arbor measure

stated its purpose as, “[to] provide funds for preservation and protection of parkland,

 

open space, natural habitats and City sourcewaters by the acquisition and management of
land and land rights both within and outside the City of Ann Arbor.” The Meridian
Township millage purpose is, “to provide funding for the purchase of land and/or
conservation easements, improvement of natural habitat and management of land for the
permanent preservation of open green spaces and natural features...” Although these
proposals are different from the rest in that they lack a farmland preservation component,
they list the purchase of rights to develop land (conservation easements in the Meridian
case) as a purpose. This separates them from typical land acquisition millages because
their interest is beyond the fee simple purchase of land. For this reason they are included,
but it must be noted that a millage without a farmland preservation component is likely

subject to different political factors than proposals with a farmland preservation focus.

The final limitation of the research is unavoidable. The results of the correlations
and regressions describe the likelihood of particular communities to support a PDR

millage, not particular people. Community characteristics describe an entire community,

62

and voter tum-out is never 100%. Therefore, characteristics like educational attainment
may describe a community’s population, but have no bearing on who shows up to cast a

ballot.

Future Research

An eight-hundred pound gorilla sits in the comer of this research. The political
factors that influence PDR millage results are largely ignored. The various campaign
strategies employed by pro and anti-millage campaigns are not considered. The different
campaign strategies available in urban and rural jurisdictions may also contribute to the
divergence in results between the Urban and Rural MCD datasets. This problem was also

identified by Nelson, et al. (2007),

. .many referendum details that could have affected voter behavior, including
how the referenda was placed on the ballot, clearness of ballot language, degree
of organized support/opposition to the measure, amount of information regarding
the referendum distributed to the public, and the mood of the electorate at the time

of the vote, are not readily observable.” (Nelson, Uwasu, & Polasky, 2007)

These same short-comings exist within this research, but are more readily observable due

to the small data set of Michigan jurisdictions considering a PDR millage.

One of the previously mentioned limitations of the data, double counting of
jurisdictions voting on both a county and local PDR millage, may provide an opportunity
for future research. In effect, these millages control for community characteristics so that

the influence of political variables like campaign strategy can be isolated.

63

Some basic information about these unique jurisdictions can be considered in this
context. All six of the jurisdictions that considered a local and county PDR millage (five
in Washtenaw County: Ann Arbor City, Ann Arbor Township, Scio Township,
Bridgewater Township, and Augusta Township, and one in Ingham County: Meridian
Township) supported the local millage proposal at a higher rate. In Peninsula Township,
the only jurisdiction to consider a local PDR millage twice, support for PDR increased in
the second election even though the millage rate more than doubled. In 2002, prior to
expiration of the 1994 millage proposal, Peninsula Township voters approved a new PDR
millage of 2 mills, a 1.25 mill increase on the 1994 millage. Table 5.4 displays the
results of these millage votes, the “2” following the jurisdiction name denotes a county

millage.

Table 5.4: Double-counted jurisdictions

Ann Arbor Ci Local/S
Ann Arbor Ci 2 State
Ann Arbor T '
Ann Arbor T ' State
T ' Local/S
A T ' State
' T ' National

T ' State

Meridian T ' National
Meridian T ' '
Peninsula T ' National
Peninsula T ' State
Scio T ' National
Scio T ' State
Webster T ' Local/S
Webster T ' State

 

64

 

In addition to further investigating the affect of political factors, data reflecting
land use/cover change would provide a significantly better measure of development
pressure than the proxies employed here. Additional research involving all Michigan
MCDs, using a binary independent variable, could also create a model with greater

coefficient values and less multicollinearity.

Final Conclusions

The sum of this research is that PDR millages in Michigan have been more likely
to pass in urban and suburban communities with a high level of educational attainment
and low rates of owner occupancy that are located in counties where farmland value did
not increase rapidly between 1987 and 2007. Other variables that were found to have a
significant influence on the percent of “yes” votes for a PDR millage are total population,
change in home value, home value, rural population, mobility of population, population
density, percent population white, and support for George W. Bush in the 2004
presidential election. However, these conclusions are tempered by the disproportional
influence of urban jurisdictions, the small number of Michigan jurisdictions to consider a

PDR millage, and the high degree of multicollinearity of the variables employed.

Overall, this project does not paint a clear picture of the determinants of success
for PDR millages in Michigan. Basic characteristics of the data make this nearly
impossible. This may be a result of poor design of the project; or it may indicate the
importance of other factors not considered. Specifically, political campaign strategies,
the status of planning institutions in the area, and land use/land cover change within each

jurisdiction may exhibit a greater influence on success of PDR millages than community

65

 

characteristics, development pressure, and political factors (most of which are out of the
control of local advocates and local officials). If factors other than those examined herein
do in fact exhibit the greatest influence on PDR millages, this research may be a success
simply because it dismisses the popular perception that growth management and land
preservation is reserved for affluent communities. At least in Michigan, it may be
possible to influence support for land preservation in less-affluent communities using

proven campaign strategies (future research is needed to identify these strategies).

One final philosophical note on this research is necessary. This is an advocacy
piece; the inherent assumption is that the passage of PDR millages is generally a good
thing for Michigan. While this may be an obvious conclusion for many, some literature
(Howell-Moroney, 2004b) and case studies indicate otherwise. In Lancaster and
Montgomery Counties, two places where land preservation efforts have been particularly
successful, master planning and preservation efforts are one and the same. While
preserving land has intrinsic value; doing so without consideration of future land use
planning wastes an opportunity to permanently direct growth to desired areas. For this
reason a PDR millage at the local level, while a positive step in itself, should consider a
regional, or at least county master plan in its selection criteria. Preserving an occasional
parcel, or unsustainable blocks of farmland will only result in vacant farmland as
increasing incompatible uses result in nuisance lawsuits and exacerbate the
impermanence syndrome (Sherman, Milshaw, Wagner, & F reedgood, 1998). There are
far too many benefits of considering future land use goals and land preservation together

to not do so.

66

APPENDIX A:
COUNTY MINOR CIVIL DIVISIONS

Table A.l: County MCDs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ingham County
Alaiedon TownshiL Locke Township
Aurelius Township Mason City
Bunker Hill Township Meridian Townslm)
Delhi Township Onondaga Township
East Lansing Stockbridge Township
Ingham Township Vevay Township
Lansing City Wheatfield Township
Lansing Township White Oak Township
Leroy Township Williamston City
Leslie City Williamston Township
Leslie TownsM)

Lapeer County
Almont Township Imlay City
Arcadia Township Imlay Township
Attica Township Lapeer City
Burlington Township Lapeer Township
Burnside Township Marathon Township
Deerfield Township Mayfield Township
Dryden Township Metamora Township
Elba Township North Branch Township
Goodland TownshiL Oregon Township
Hadley Township Rich Township

 

 

Leelanau County

 

 

 

 

 

 

 

Bingham Township Kasson Township
Centerville Township Leelanau Township
Cleveland Township Leland Township
Elmwood Township Solon Township
Empire Township Suttons Bay

Glen Arbor Traverse City

 

 

67

 

Table A.1 continued

 

 

 

 

Washtenaw County
Ann Arbor City Pittsfield Township
Ann Arbor Township Salem Township
Augusta Township Saline City

 

Bridgewater Township

Saline Township

 

Dexter Township

Scio Township

 

Freedom Township

Sharon Township

 

 

 

 

 

 

 

 

Lima Township Superior Township
Lodi Township Sylvan Township
Lyndon Township Webster Township
Manchester Township York Township
Milan City Ypsilanti City
Northfield Township Ypsilanti Townshfir

 

 

68

 

 

APPENDIX B: CORRELATION MATRIX

Table 8.1: MCD dataset independent variable correlations

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

% Rural Change Rural
Yes Total Pop. Pop. POE % Mobility

Yes Pearson 1 .361" -423" 0.073 .354"
Correlation

Total Pop. Pearson .361“ 1 4557" -0023 .275"
Correlation

% Rural Pop. Pearson -.423" -557" 1 -.238' -.226‘
Correlation

% Change Rural Pearson 0.073 -0023 -238" 1 -.211‘

Pop. Correlation

% Mobility Pearson .354" .275" -.226' -.211 1
Correlation

Educational Pearson .542" .422‘T -497" .226' .264’

Attainment Correlation

Median Income Pearson -0.064 -O.168 0.092 .499" -382"
Correlation

% Owner Occupied Pearson -.474" -.603'T .741 " .290" -.532"

Housing Correlation

Median Age Pearson -0.156 -453" .477" 0.049 -.2l3'
Correlation

Home Value Pearson .261' -0.034 -0.073 .278" 0.124
Correlation

% Change Home Pearson -.3701r -.360'r .515" -.321rr -0.11

Value Correlation

% Change Pearson 0.14 —.211TT 0.133 -0204 .409"

Households Correlation

Density Pearson .358" .741“ -704" -0125 .389"
Correlation

% No Vote Pearson -0.036 0.05 -0.201 0.009 0.107
Correlation

Turn-out Pearson 0.007 -0157 0.15 0.134 .223‘
Correlation

Bush Vote Pearson -.358" -.692" .667" 0.039 4259C
Correlation

% White Pearson -271" -537" .630" —0.039 -.298"
Correlation

 

 

 

 

 

 

 

 

"Significant at the .01 level.

*Significant at the .05 level.

69

Table 3.] continued

Educational Median . Median Home
Attainment Income Occupied Age Value

Correlation
Correlation
Correlation
Correlation
Correlation
Attainment Correlation
Correlation
Correlation
Correlation
Correlation
Value Correlation
Households Correlation
Correlation
Correlation
Correlation
Correlation

Correlation
at .01
*Significant at the .05 level.

 

70

Table 8.] continued

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

% Change % Change % No
Home Households Density Vote Tum-out
Value

Yes Pearson «370* 0.14 .358" -0.036 0.007
Correlation

Total Pop. Pearson -.360W -.21 1' .741" 0.05 0157
Correlation

% Rural Pop. Pearson .515" 0.133 -704" -0.201 0.15
Correlation

% Change Rural Pearson -321" -0204 -0125 0.009 0.134

Pop. Correlation

% Mobility Pearson -0.11 .409" .389" 0.107 .223'
Correlation

Educational Pearson -.516" 0.106 .353" -0072 .408"

Attainment Correlation

Median Income Pearson -0.201 0.012 —.372" -.220' 0.131
Correlation

% Owner Pearson .240' -0.04 -.825" -0.199 0.142

Occupied Housing Correlation

Median Age Pearson 0.126 .325" -640" -.257' .629“
Correlation

Home Value Pearson -0.154 .389" -0.17 -.260' .540"
Correlation

% Change Home Pearson 1 .238' -351" -0.058 .0079

Value Correlation

% Change Pearson .238' 1 -.265r -.209' .353“

Households Correlation

Density Pearson -351" -265' 1 277* -.248'
Correlation

% No Vote Pearson -0.058 -.209' .277" 1 0.1
Correlation

Turn-out Pearson -0079 .353“ -.248' -0.1 1
Correlation

Bush Vote Pearson .328" 0.169 -743" —0.069 0.076
Correlation

% White Pearson .310" 0.135 -.607" -0.171 0.078
Correlation

 

 

 

 

 

 

 

 

"Significant at the .01 level.

*Significant at the .05 level.

71

Table 3.] continued

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Bush Vote %
White

Yes Pearson -.358" -.271"
Correlation

Total Pop. Pearson -.692" «537“
Correlation

% Rural Pop. Pearson .667" .630"
Correlation

% Change Rural Pearson 0.039 -0.039

Pop. Correlation

% Mobility Pearson -2595 -298"
Correlation

Educational Pearson -.540" -.42 l "

Attainment Correlation

Median Income Pearson 0.129 0.1 1 1
Correlation

% Owner Pearson .726" .65?

Occupied Correlation

Median Age Pearson .478" .494"r
Correlation

Home Value Pearson -0.125 -0.086
Correlation

% Change Home Pearson .328fi .310"

Value Correlation

% Change Pearson 0.169 0.135

Households Correlation

Density Pearson -743" -.607“
Correlation

% No Vote Pearson -0.069 -0.171
Correlation

Turn-out Pearson 0.076 0.078
Correlation

Bush Vote Pearson 1 .788'r
Correlation

% White Pearson .788" 1
Correlation

 

 

"Significant at the .01 level.
*Significant at the .05 level.

72

 

 

APPENDIX C: DESCRIPTIVE
STATISTICS

Table C.1 Rural MCD descriptive statistics

\tll'itllllc \Iinimum \l;1\imum \lcaln Std. lhwintion

       

40

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Yes .22 .59 .41 .07
Total Pop. 40 788.00 9072.00 3568.97 2065.36
%Pop. Rural 40 .00 1.00 .87 .26
Change % Pop. 40 .00 .65 .07 .15
Rural

Migration 40 .12 .38 .23 .05
Educational 40 .06 .54 .2 1 . 13
Attainment .
Median Income 40 32436.00 72381.00 50608.92 9058.26
% Owner 40 .52 .98 .87 .09
Occupied

Age 40 32.40 52.60 39.43 5.07
Median Home 40 91400.00 298000.00 152184.21 41121.52
Value

% Change Home 40 .43 1.73 1.10 .32
Value

% Change House 40 .06 .57 .30 .11
Holds

Density 40 22.31 1570.21 169.31 306.69
No Vote 40 .01 .20 .06 .04
Turn Out 40 .15 .86 .44 .26
Bush Vote 40 .43 .68 .58 .06
% White 40 .74 .99 .96 .04

 

73

 

Table C.2: Urban MCD descriptive statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Minimum \ltnimum \lczlll Std. Deviation
Yes 52 .27 .77 .42 .12
Total Pop. 52 1177.00 114321.00 15144.06 27327.35
%Pop. Rural 52 .00 1.00 .56 .43
Change % Pop. 52 -.O3 .52 .10 .16
Rural
Migration 52 .10 .57 .21 .10
Educational 52 .09 .79 .34 .19
Attainment
Median Income 52 28217.00 90830.00 60480.98 15517.20
% Owner 52 .32 .95 .79 .17
Occupied
Age 52 21.70 41.30 36.27 4.09
Median Home 52 73200.00 325300.00 166826.92 56571.41
Value
% Change Home 52 .45 1.25 .83 .21
Value
% Change House 52 -.03 .66 .22 .17
Holds
Density 52 32.29 5090.37 818.00 1291.75
No Vote 52 .00 .22 .06 .04
Turn Out 52 .10 .84 .37 .19
Bush Vote 52 .20 .64 .50 .1 1
% White 52 .61 .99 .90 .10

 

 

 

 

 

 

74

 

 

Table C.3 MCD descri

tive statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Std.
Minimum Maximum .‘lcan Deviation
Yes 92 .22 .77 .42 .10
Total P0P- 92 149.00 114321.00 10064.93 21314.28
%Pop. Rural 92 .00 1.00 .69 .39
Change % Poo. 92 -.78 .65 .07 .18
Rural
Migration 92 .10 .57 .23 .09
Educational 92 .06 .79 .29 .18
Attainment
Median Income 92 19423.00 90830.00 55693.73 14519.31
% Owner 92 .32 .98 .81 .15
Occupied
Age 91 21.70 52.60 37.69 4.81
Median Home 92 73200.00 325300.00 161430.43 50778.38
Value
% Change 92 .43 1.73 .95 .30
Home Value
% Change 92 —.03 1.86 .27 .23
House Holds
Density 92 22.31 5090.37 539.54 1037.32
No Vote 92 .00 .22 .06 .04
Turn Out 92 .10 .88 .41 .22
Bush Vote 92 .20 .68 .53 .10
% White 92 .61 .99 .93 .09

 

 

 

 

 

 

75

 

 

BIBLIOGRAPHY

Adelaja, S., Lake, M. B., Colunga-Garcia, M., Hamm, M., Bingen, J ., Gage, S., et a1.
(2006). Acreage and Funding Goals for Farmland Preservation in Michigan:
Targeting Resiliency, Diversity, and Flexibility. East Lansing: The MSU Land
Policy Institute.

Anderson, W. L., & Mizak, D. A. (2006). Politics of Environmental Law: Political
Ideology, Elitism or Urban-le Interests? Public Choice , 131-157.

Bidwell, D., Westphal, J ., Wunsch, J ., & Berton, V. (1996). Forging New Protections:
Purchasing Development Rights to Save Farmland. Washington DC: American
Farmland Trust.

Daniels, T. (1998). The Purchase of Development Rights, Agricultural Preservation and
other Land Use Policy T ools-- T he Pennslyvania Experience. Retrieved June 1 l,
2009, from Increasing Understanding of Public Problems and Policies:
http://pur1.umn.edu/1 7934

Daniels, T. (1991). The Purchase of Development Rights: Preserving Agricultural Land
and Open Space. Journal of the American Planning Association , 421-431.

Daniels, T., & Bowers, D. (1997). Holding Our Ground: Protecting America 's Farmas
and Farmland. Washington DC: Island Press.

Daniels, T., & Lapping, M. (2005). land Preservation: An Essential Ingredient in Smart
Growth. Journal of Planning Literature , 316-327.

Davis, F ., & Wurth, A. (2003). Voting Preferences and the Environment in the American
Electorate: The Discussion Extended. Society and Natural Resources , 720-740.

Deacon, R., & Shapiro, P. (1975). Private Preference for Collective Goods Revealed
Through Voting on Referenda. American Economic Review , 943-955.

Duke, J. M., & Lynch, L. (2007). Gauging Support for Innovative Farmland Preservation
Techniques. Policy Sciences , 132-155.

Furseth, O. J. (1987). Public Attitudes Toward Local Farmland Protection Programs.
Growth and Change , 49-61.

Gardner, D. B. (1977). The Economics of Agricultural Land Preservation. The American
Journal of A gricultural Economics , 1027- 1036.

Governor's Task Force on the Future of Agriculture. (1970). Final Report. Lansing: State
of Michigan.

Howell-Moroney, M. (2004). Community Characteristics, Open Space Preservation and
Regionalism: Is there a Connection? Journal of Urban Affairs , 109-118.

76

 

Howell-Moroney, M. (2004). What are the Determinants of Open-Space Ballot
Measures? An Extension of the Research. Social Science Quarterly , 169-182.

Kahn, M., & Matsusaka, J. (1997). Demand for Environmental Goods: Evidence from
Voting Patterns on California Initiatives. Journal of Law and Economics , 137-
173.

Kline, J ., & Wichelns, D. (1996). Public Preferences Regarding the Goals of Farmland
preservation Programs. Land Economics , 538-549.

Kline, J ., & Wichelns, D. (1996). Public Preferences Regarding the Goals of Farmland
Preservation Programs. Land Economics , 53 8-549.

Kotchen, M., & Powers, S. (2006). Explaining the Apearance and Success of Voter
Referenda for Open-Space Conservation. Journal of Environmental Economics
and Management , 373—390.

Kotchen, M., & Powers, S. (2006). Explaining the Appearance and Success of Voter ,,
Referenda for Open-Space Conservation. Journal of Environmental Economics i

 

and Management , 373-390.

Lindsey, G., & Knapp, G. (1999). Willingness to Pay for Urban Greenway Projects.
Journal of the American Planning Association , 297-314.

Lonick, B., & Garfield, M. (2008, July 15). How to Initiate a Millage Proposal. Purchase
of Development Rights Training Seminar . Lansing, MI: Michigan Department of
Agriculture.

McMahon, E. T., & Mastran, S. S. (2004). Better Models for Development in Maryland.
Arlington: The Conservation Fund.

Michigan Farmland Agriculture Development Task Force. (1994). Policy
Recommendations and Options for the Future of Michigan Agriculture. Lansing:
State of Michigan.

Michigan Land Use Leadership Council. (2003). Michigan 's Land, Michigan's Future:
Final Report of the Michigan Land Use Leaderhsip Council. Lansing: State of
Michigan.

Myers, P. (1999). Livability at the Ballot Box: State and Local Referenda on Parks,
Conservation, and Smarter Growth, Election Day 1998. Retrieved May 15, 2009,
from Brookings Institution Web Site:
http://www.brookings.edu/reports/ 1999/01metropolitanpolicy_myers.aspx

Nelson, E., Uwasu, M., & Polasky, S. (2007). Voting on Open Space: What Explains the
Appearance and Support of Municipal-level Open Space Conservation Referenda
in the United States? Ecological Economics , 580-593.

77

Nelson, E., Uwasu, M., & Polasy, S. (2007). Voting on Open Space: What Explains the
Appearance and Support of Municipal-Levelo Open Space Conservation
Referenda in the United States. Ecological Economics , 580-593.

Noregaard, K. (1999). T wenty-F our Years of F armland Preservation in Michigan, PA
116. East Lansing: Department of Agricultural Economics, Michigan State
University.

Norris, P., & Deaton, B. J. (2002). Public Opinions about Farmland and Farmland
Preservation: Results fiom a Survey in Kent County, Michigan. East Lansing:
Department of Agricultural Economics, Michigan State University.

Norris, P., & Deaton, B. J. (2001). Understanding the Demand for Farmland
Preservation: Implications for Michigan Policies. East Lansing: Michigan State
University Department of Agricultural Economics.

Norris, P., Soule, J ., Weissert, C., Stuart, G., & Skole, D. (n.d.). Michigan's Opportunities
and Challenges: MS U Faculty Perspectives. Retrieved ll 23, 2008, from
Michigan State University Extension:
http://web1 .msue.msu.edu/msue/iac/transition/papers/ManLandUse.pdf

Pfeffer, M., & Lapping, M. (1995). Public and Farmer Support for Purchase of
Development Rights in the Metropolitan Northeast. Journal of Soil and Water
Conservation , 30.

Romero, F., & Liserio, A. (2002). Saving Open Spaces: Determinants of 1998 and 1999
"Antisprawl Ballot Measures. Social Science Quarterly , 341-352.

Salka, W. (2003). Determinants of Countywide Voting Behavior on Environmental
Ballot Measures: 1990-2000. Rural Sociology , 253-277.

Schlapfer, F ., Roschewitz, A., & Hanley, N. (2004). Validation of Stated Preferences for
Public Goods: A Comparison of Contingent Valuation Survey Response and
Voting Behavior. Ecological Economics , 1-16.

Sherman, R. L., Milshaw, S., Wagner, R. C., & Freedgood, J. (1998). Investing in the
Future of Agriculture: The Massachusets Farmland Protection Program and the
Permanence Syndrome. American Farmland Trust: Northhampton.

Skjaerlund, D., & Sheridan, S. (2003, February). Designing Farmland Preservation
Programs to Complement Flamed Growth. Planning and Zoning News , pp. 10-
13.

Soleck, W. D., Mason, R. J ., & Martin, S. (2004). The Geography of Support for Open-
Space Initiatives: A Case Study of New Jersey's 1998 Ballot Measure. Social
Science Quarterly , 624-639.

USDA National Agricultural Statistics Service. (2009, F eburary 3). Table 8: Farms, Land
in Farms 2007 and 2002. Retrieved June 11, 2009, from The Census of
Agriculture:

78

 

http://www.agcensus.usda. gov/Publications/2007/Full_Report/Volume_1 ,_Chapte
r_2_US_State_Level/st99_2_008_008.pdf

Vossler, C., Kerkvliet, J ., Polasky, S., & Gainntdinove, O. (2003). Externally Validating
Contingent Valuartion: An Open-Space Survey and Referendum in Corvallis
Oregon. Journal of Economic Behavior and Orgainzation, 261-277.

Wilson, R. H., & Paterson, R. (2003). Innovative Initiatives in Growth Management and
Open Space Preservation: A National Stuofv. Austin: LBJ School of Public
Affairs.

Wison, R. H., & Paterson, R. (2002). State Growth Management and Open Space
Preservation Policies. Austin: LBJ School of Public Affairs.

Wright, J. B. (1994). Designing and Applying Conservation Easements. Journal of the
American Planning Association, 380-388.

79