SHORELAND DEVELOPMENT AND DISTURBANCES: A HEDONIC ANALYSIS OF LAKEFRONT PROPERTIES IN NORTHEASTERN WISCONSIN, USA

University of Wisconsin Milwaukee University of Wisconsin Milwaukee

UWM Digital Commons UWM Digital Commons

Theses and Dissertations

May 2024

SHORELAND DEVELOPMENT AND DISTURBANCES: A HEDONIC SHORELAND DEVELOPMENT AND DISTURBANCES: A HEDONIC

ANALYSIS OF LAKEFRONT PROPERTIES IN NORTHEASTERN ANALYSIS OF LAKEFRONT PROPERTIES IN NORTHEASTERN

WISCONSIN, USA WISCONSIN, USA

SUSAN BORCHARDT

University of Wisconsin-Milwaukee

Follow this and additional works at: https://dc.uwm.edu/etd

Part of the Natural Resource Economics Commons, and the Water Resource Management Commons

Recommended Citation Recommended Citation

BORCHARDT, SUSAN, "SHORELAND DEVELOPMENT AND DISTURBANCES: A HEDONIC ANALYSIS OF

LAKEFRONT PROPERTIES IN NORTHEASTERN WISCONSIN, USA" (2024).

Theses and Dissertations

3458.

https://dc.uwm.edu/etd/3458

This Thesis is brought to you for free and open access by UWM Digital Commons. It has been accepted for

inclusion in Theses and Dissertations by an authorized administrator of UWM Digital Commons. For more

information, please contact [email protected].

SHORELAND DEVELOPMENT AND DISTURBANCES:

A HEDONIC ANALYSIS OF LAKEFRONT PROPERTIES

NORTHEASTERN WISCONSIN, USA

Susan Borchardt

A Thesis Submitted in

Partial Fulfillment of the

Requirements for the Degree of

Master of Science

in Freshwater Sciences

The University of Wisconsin-Milwaukee

May 2024

ABSTRACT

SHORELAND DEVELOPMENT AND DISTURBANCES:

A HEDONIC ANALYSIS OF LAKEFRONT PROPERTIES

NORTHEASTERN WISCONSIN, USA

Susan Borchardt

The University of Wisconsin-Milwaukee, 2024

Under the Supervision of Professor James Price

Shoreland development, encompassing features like boat lifts, manicured lawns, artificial

beaches, and erosion control measures, offers considerable benefits to property owners.

Nevertheless, this development disrupts natural conditions and is associated with increased

sediment and pollutant loading, which negatively impacts aesthetics, recreation, and habitat for

fish and other aquatic species. This thesis conducts two analyses, which respectively quantify the

benefits of shoreland development to homeowners and evaluate the relationship between

shoreland development and lake water quality. In the first analysis, a hedonic property model is

employed to value shoreland development along Wisconsin inland lakes. The model considers

various shoreland development features, using data from 62 lakes surveyed comprehensively

under the Wisconsin Department of Natural Resources (WDNR) Lake Shoreland and Shallows

Habitat Monitoring Program. Results show positive correlations between sales prices and certain

development features, including artificial beaches, erosion control measures, and structures in the

littoral zone, after controlling for housing characteristics and lake fixed effects. Willingness-to-

pay values for these features are derived from the model; these values can be compared to the

welfare loss stemming from sediment and pollutant loading caused by development to inform

iii

shoreland management decisions. In the second analysis, the effect of shoreland development on

lake water clarity is evaluated. Results show that the extent of shoreland development on a lake

is not significantly correlated with water clarity, after controlling for lake characteristic and

nearby land use.

my husband who didn’t tell me that;

going back to school was a

crazy idea.

TABLE OF CONTENTS

SHORELAND DEVELOPMENT AND DISTURBANCES: ......................................................... i

ABSTRACT .................................................................................................................................... ii

LIST OF FIGURES ..................................................................................................................... viii

LIST OF TABLES ......................................................................................................................... ix

LIST OF ABBREVIATIONS ......................................................................................................... x

1. Introduction ................................................................................................................................. 1

2. Hedonic Property Analysis ......................................................................................................... 5

2.1. Introduction to Hedonic Analysis ........................................................................................ 5

2.2. Literature Review ................................................................................................................. 5

2.2.1. Economic Valuation Approaches .................................................................................. 6

2.2.2. Hedonic Property Method ............................................................................................. 8

2.3. Study Area and Data .......................................................................................................... 16

2.3.1. Study Area ................................................................................................................... 16

2.3.2. Housing Data ............................................................................................................... 17

2.3.3. Lake Data ..................................................................................................................... 18

2.3.4. Habitat Assessment...................................................................................................... 18

2.4. Methodology ...................................................................................................................... 21

2.4.1. Theoretical Background .............................................................................................. 22

2.4.2. Hot Spot Analysis ........................................................................................................ 29

2.5. Results ................................................................................................................................ 34

2.5.1. Hedonic Model Results ............................................................................................... 34

2.5.2. Marginal Willingness to Pay ....................................................................................... 37

2.5.3. Robustness Checks ...................................................................................................... 41

2.6. Discussion and Conclusion ................................................................................................ 42

2.7. Future Research .................................................................................................................. 44

3. Lake Water Quality Analysis .................................................................................................... 45

3.1. Introduction ........................................................................................................................ 45

3.2. Literature Review ............................................................................................................... 46

3.2.1. Water Quality Determinants ........................................................................................ 46

3.2.2. Buffer Zones ................................................................................................................ 47

3.3. Data .................................................................................................................................... 48

3.3.1. Land Use/Landcover.................................................................................................... 48

vii

3.3.2. Water Clarity ............................................................................................................... 48

3.3.3. Habitat Assessment...................................................................................................... 51

3.4. Methodology ...................................................................................................................... 51

3.5. Results ................................................................................................................................ 52

3.6. Discussion and Conclusion ................................................................................................ 56

3.7. Future Research .................................................................................................................. 57

4. Conclusion ................................................................................................................................ 58

Bibliography .............................................................................................................................. 61

Appendix ....................................................................................................................................... 66

viii

LIST OF FIGURES

Figure 1 Study lakes and study area for the hedonic models. The size of the study lakes has been

exaggerated for visual clarity. Lake and County shape files were downloaded from the WDNR’s

website (WDNR

, n.d.). ................................................................................................................ 17

Figure 2 The three habitat zones used in the WDNR Habitat Assessment Study (WDNR, 2020)

....................................................................................................................................................... 20

Figure 3 Plan view of the riparian zone in relation to the study parcels and lakes. Lake shape file

from the WDNR. Parcel shape file from the Wisconsin's State Cartographers Office (n.d.). ...... 20

Figure 4 Median sales price of Wisconsin homes from data collected by the Wisconsin Realtors

Association per sales quarter for years 2007-2023, data is updated daily and downloaded from

their website on 2 February 2024 (WRA, 2024) ........................................................................... 27

Figure 5 Normal Distribution Curve (ArcGIS Pro, 2023) ........................................................... 30

Figure 6 Hotspots for the variable of the natural log of the quotient derived when the area of the

lot in acres is divided by the length of the shoreline in feet. ........................................................ 33

Figure 7 Hotspots for the variable of the natural log of the home's gross living area in square

feet................................................................................................................................................. 33

Figure 8 Measuring water clarity with a secchi disk, Image from The Many Faces of Water

(Mierzynski, 2017) ........................................................................................................................ 50

Figure 9 Graph of the relationship between the Lakeshore Disturbance Matrix from the

Wisconsin Department of Natural Resources versus the measurement of how deep a secchi disk

can be observed below the surface on 117 Wisconsin lakes. ....................................................... 56

LIST OF TABLES

Table 2.1 Economic Methods for Measuring Environmental and Resource Values

(Tietenberg & Lewis, 2019) .......................................................................................................... 7

Table 2.2. Hedonic Property Model Variable Labels and Descriptions by Variable Type 28

Table 2.3 Model Specifications with Independent Variables Grouped by Variable Type ... 31

Table 2.4 Descriptive Statistics for Variables Used in the Hedonic Property Model ........... 32

Table 2.5 Hedonic Property Model Results (Dependent Variable: Natural Log of Adjusted

Sales Price) ................................................................................................................................... 39

Table 2.6 Overall Model Statistics (Residuals, Degrees of Freedom, F-Stat, and p-value) for

the Five Hedonic Property Models. ........................................................................................... 40

Table 2.7 Marginal Willingness to Pay Per Lakefront Property in Northeastern Wisconsin

(2022 U.S. Dollars) ...................................................................................................................... 40

Table 2.8 Summary of Shoreland Disturbance Results by Shoreland Habitat Zone ........... 41

Table 3.1 Land Use Classes and Descriptions .......................................................................... 49

Table 3.2 Regression Model Specifications for Lake Water Clarity Analysis (Dependent

Variable: Secchi Depth) .............................................................................................................. 52

Table 3.3 Regression Model Results (Dependent Variable: Secchi Depth) ........................... 54

Table 3.4 Overall Model Statistics for Lake Water Clarity Analysis .................................... 55

Table A.1 Robustness Checks for Hedonic Property Model Results (Dependent Variable:

Natural Log of Adjusted Sales Price). ....................................................................................... 66

LIST OF ABBREVIATIONS

………………..

coefficients

Bypass

………………..

The presence of a feature that allows runoff water to

bypass the filtration effect of the canopy and/or shrubs

Can_Shrub

………………..

Percent of parcel shoreline covered in tree canopy or

shrubs.

CWA

………………..

Clean Water Act

Degrees Of Freedom

EC_Pct

………………..

Percent of parcel shoreline containing artificial erosion

control

EC_Pct

………………..

Percent of parcel shoreline containing artificial erosion

control

ESRI

FRED

………………..

Environmental Systems Research Institute

Federal Reserve Economic Data

GIS

………………..

Geographical Information System

GLA

………………..

Gross Living Area

………………..

number of garage spaces

HPM

………………..

Hedonic property method

HWL

………………..

high-water level

LAKE_NAME

………………..

The name of the lake the property abuts.

LakeA

………………..

Area of the lake in acres

LakeD

………………..

Depth of the lake in feet

LitStruc

………………..

Number of structures in the littoral zone

………………..

natural log

LnASP

………………..

log of the adjusted sales price of the properties

LnGLA

………………..

natural log of the GLA

LULC

MLS

MMI_L

MMI_P

………………..

Land Use Land Cover

Multiple listing service

Shoreland Habitat disturbance metric per WDNR survey

for the lake as a whole

Shoreland Habitat disturbance metric per WDNR survey

for the parcel as a whole

………………..

natural resources

OHWM

………………..

ordinary high-water mark

………………..

property price

………………..

coefficient of determination

RipStruc

………………..

Number of structures in the riparian zone

RR_Pct

………………..

Percent of parcel shoreline using rip rap for erosion

control

SaleQrt

………………..

The quarter the property sold

SaleYear

………………..

The year the property sold

………………..

Standard deviation

………………..

shoreland length

SWIMS

TWTP

Veg

………………..

Surface Water Integrated Monitoring System

Total Willingness To Pay

Unites States

The presence of floating vegetation in the littoral zone

(0,1,2)

VW_Pct

………………..

Percent of parcel shoreline using a vertical wall for

erosion control

WDNR

………………..

Wisconsin Department of Natural Resources

WRA

………………..

Wisconsin Realtors Association

WTP

………………..

Willingness To Pay

………………..

property amenities (variables)

1. Introduction

Wisconsin is rich in lakes, with over 15,000 documented lakes, of which only about 40% have

been named. Most of the state’s lakes are less than 10 acres, with only 2.4% having a surface

area of at least 20 acres. These larger lakes, however, represent 93% of the approximately one

million acres of surface area that Wisconsin inland lakes cover (WDNR, 2009). All these lakes

mean there is a lot of shoreland in the state, and all this shoreland has been zoned in accordance

with standards established by the Wisconsin Department of Natural Resources (WDNR). The

purpose of shoreland zoning, per the WDNR, is to establish minimum standards to limit the

direct and cumulative impacts of shoreland development on water quality; near-shore aquatic

wetland and upland wildlife habitat; and the natural scenic beauty (WDNR, 2017).

Lakeshore development in Wisconsin began in the 1920s and 30s with the advent of better roads

outside the big cities. Agricultural and forest land abutting freshwater lakes was subdivided into

lots to accommodate small cottages (Yanggen & Kusler, 1968). Cottages usually relied on

outhouses or an onsite septic system, both of which allowed domestic waste to seep into the

ground and find its way to nearby surface water. As demand for proximity to recreational waters

grew, the price of existing lakeshore lots increased, and less desirable lots were developed. The

less desirable lots had steep slopes or were low lying with high groundwater levels, making it

challenging to prevent runoff. This development brought much-needed investment to the rural

communities, but it also increased risks to the natural environment. Uncontrolled development

brought erosion problems from road building and grading and increased runoff from the

displacement of shoreland cover and habitat with manicured lawns and impervious surfaces.

To protect the waters and natural beauty of the state's shoreland, Wisconsin enacted its original

shoreland zoning laws in August of 1966 (Yanggen & Kusler, 1968). The WDNR has set

minimum standards that the local zoning authorities can make more, but not less restrictive.

Minimum zoning standards for shorelands are outlined in Chapter NR 115.05 of the Wisconsin

Administrative Code (WDNR, 2017). There are minimum standards for lot size, building

setbacks, vegetation, filling and dredging, use of impervious surfaces, structure height, and the

use of nonconforming structures. There is also a provision that these minimum standards become

the zoning regulation in the case that the county fails to provide zoning standards that meet the

minimums set forth by the WDNR.

The shoreland area, as defined by the WDNR, is land within the following distances from the

ordinary high-water mark (OHWM) of navigable waters: 1,000 feet from a lake, pond, or

flowage; and 300 feet from a river or stream or to the landward side of the flood plain, whichever

distance is greater. The OHWM is the point on the shore where the presence and action of the

water is so continuous as to leave a distinct mark. There are several approved uses of shoreland

property, including parks, agricultural crops, hunting, pasturing of livestock, and both residential

and commercial activities (WDNR, 2017). The zoning laws restrict what an individual property

owner can build or how they can landscape the lot along the shoreline. The ordinances mitigate

negative impacts of development on Wisconsin surface waters by limiting pollution, preserving

fish habitat, and attracting recreation; they are intended to enhance the beauty and value of the

properties with lake frontage (WDNR, 2017). That said, regulations can have a positive or

negative effect on the property’s value. Regulations that put limits on a property's use, or that

limit either directly or indirectly what improvements can be added to the property, can

substantially diminish the land value (Yanggen & Kusler, 1968), while good water quality has

been related to a higher value for lakeshore property in studies by Guignet et al. (2022) and

David (1968).

Despite these standards, there is growing recognition of the need for greater shoreland protection

and restoration; these areas are increasingly recognized as critical habitats for fish of both

ecological and economic importance. Shoreland areas are found to be highly productive habitats

that are favored by juvenile fish because they provide refuge from predation and are thus

considered fish nurseries by ecologists (Munsch et al., 2017). It is estimated that recreational

anglers in Wisconsin annually harvest for food ≈ 4,200 metric tons of fish from ≈ 4,000 inland

lakes (Embke et al., 2020). This estimate, derived from creel surveys, does not include the added

impact from the sport of catch and release fishing. Shoreland habitats, however, are being

transformed to support the demands of home buyers seeking lakefront property, and many of

these shorelands are being modified by armoring (artificial erosion control, vertical walls, rip

rap, etc.) and the addition of structures in both the riparian and littoral zones that further facilitate

waterfront use (Munsch et al., 2017). Per Munsch (2017), “these modifications affect the ecology

of nearshore systems by restructuring, eliminating, and shading shallow waters.”

This thesis conducts two analyses, which respectively quantify the benefits of shoreland

development to homeowners and evaluate the relationship between shoreland development and

lake water quality. The first analysis estimates the value homeowners place on characteristics of

shoreland zoning area—considered by the property owner as improvements but as disturbances

by the WDNR—by evaluating variations in property sales prices. When studies are performed

for a housing market, the sales prices vary with access to various amenities and disamenities. For

example, lakefront properties sell for more than similar properties farther away from the water

body, reflecting the homeowners’ willingness to pay (WTP) for the recreational and aesthetic

benefits of lake access. The recreational and aesthetic benefits are not traded directly in the

market; they are a non-market good. Economists have developed an analytical model to derive

the customer's WTP for this non-market good called the hedonic property method (HPM)

(Rosen, 1974; Champ et al., 2003; Lansford, & Jones, 1995). The HPM has been used to

evaluate a wide range of environmental amenities, including numerous studies that have

estimated WTP for improved water quality. Few studies, however, have evaluated shoreland

habitat condition or environmental zoning. In this analysis, an HPM is employed to value

shoreland development along Wisconsin inland lakes. Shoreland data are obtained from the

WDNR, which has systematically documented shoreland habitat disturbances at numerous

Wisconsin lakes under the WDNR Lake Shoreland and Shallows Habitat Monitoring Program. A

summary of this work is available on their dashboard (WDNR

, n.d.). The HPM in this study

considers various shoreland features, using data from 62 of the surveyed lakes, matched to

household sales data.

Results from the two analyses provide insight into the benefits and costs of shoreland

preservation and restoration efforts in Wisconsin. On one hand, findings from the HPM show

that homeowners place considerable value on some shoreland features, including beaches,

erosion control measures, and recreation-related structures like boat houses. But on the other

hand, the second study on water clarity does not find evidence that shoreland disturbances affect

water clarity. Since the model has a small sample size and is subject to potential omitted

variables bias, this result may not accurately reflect the relationship disturbances and water

clarity. Nevertheless, findings emphasize the need to develop shoreland policies that balance the

benefits and cost of shoreland development.

2. Hedonic Property Analysis

2.1. Introduction to Hedonic Analysis

Water frontage, in general, is considered a desirable amenity, whether on a lake, stream, or pond.

As a result, properties that are adjacent to or near a water body have higher sales prices than

other properties, all else equal (Nicholls & Crompton, 2018). But characteristics of the frontage

must also be taken into consideration. Is there a manicured lawn, a vertical wall, or native shrubs

and herbaceous plants at the water’s edge? Are there accessory structures or impervious

surfaces? Is there a dock or an artificial beach? These features affect the benefits homeowners

receive from their shoreland. Shoreland development in the eyes of property owners means

shoreland improvements (manicured lawns, a dock, a boat lift, a swim raft, etc.) that can enhance

recreational and aesthetic benefits of lakefront property. These same improvements, from the

perspective of the WDNR, are shoreland habitat disturbances that negatively impact overall lake

conditions through increased runoff of sediment, nutrients, and contaminants. An improved

understanding of both the benefits and costs associated with shoreland development is essential

for making informed, efficient management decisions.

2.2. Literature Review

The literature review will examine both economic valuation approaches in general and the

specific application of the HPM to water quality and shoreland features. This review will

highlight research gaps within the literature relating to lakeshore habitat and property values.

2.2.1. Economic Valuation Approaches

Economists have developed methods for estimating the economic value of both market and non-

market goods and services. Market goods are those goods and services that can be traded in

markets, such as food, shelter, and transportation. Non-market goods, on the other hand, are

those goods and services that are not typically traded in markets, such as scenic views, species

diversity, and clean water (Champ et al., 2003). These goods lack a market price, yet they often

possess substantial economic value.

The environmental amenities (or goods) that are the focus of this study are examples of non-

market goods that are not traded in the traditional market. The total economic value of an

environmental good is made up of three main components: non-use, use, and option values

(Tietenberg & Lewis, 2019). Non-use values include bequest value (saving for future

generations) and existence value (just knowing the item exists with full knowledge that you will

never see or use it). The use value is the benefit a person derives by using the good or service,

which is further divided between consumptive use and non-consumptive use. Consumptive use is

that which diminishes the good through use, such as harvesting fish, while non-consumptive use

does not diminish the good, such as the scenic view of the lake. Option value pertains to saving

goods for possible use in the future. Economists have several methods to estimate the value of

non-market goods (Table 2.1).

Table 2.1 Economic Methods for Measuring Environmental and Resource Values

(Tietenberg & Lewis, 2019)

Methods

Revealed Preference

Stated Preference

Direct

Market Price

Contingent Valuation

Simulated Markets

Indirect

Travel Cost

Choice Experiments

Hedonic Property Values

Conjoint Analysis

Hedonic Wage Values

Attribute-Based Models

Avoidance Expenditures

Contingent Ranking

These valuation methods can be divided into two types: revealed preference and stated

preference methods. Stated preference methods are based on hypothetical markets, rather than

real markets. These methods use survey techniques to elicit individuals’ replies within a

structured questionnaire. The survey answers are then used to assess respondents’ willingness to

pay (WTP) for an environmental good or service (Tietenberg & Lewis, 2019). Choice

experiments, conjoint analysis, attribute-based models, and contingent ranking models are

examples of stated preference methods (Table 2.1).

On the other hand, revealed preference methods are based on actual human behavior, where

resource values are inferred from a related marketable good as both goods are internally linked

(Tietenberg & Lewis, 2019). Travel cost, hedonic property or wage values, and avoidance

expenditure models are examples of revealed preference methods (Table 2.1).

In this study, I use a hedonic model to value shoreland conditions, which are environmental

amenities or disamenities that do not have a market value. This method estimates the value of

amenities to homeowners as embodied by the price of the property, thus only the owners’ use

value is quantified. To estimate the total recreational and aesthetic value of the shoreland habitat,

other components would need to be added (Lansford & Jones, 1995). To get closer to the total

value, the following also need to be considered: 1) the value of the shoreland to persons living

outside the area who travel to the lake for recreation and 2) value for the existence, bequest, and

option values that persons who will never visit the lake, but who do place a value on its

shoreland habitat (Lansford & Jones, 1995).

2.2.2. Hedonic Property Method

2.2.2.1. Introduction

The hedonic method of estimating WTP is a revealed preference method that uses multiple

regression to discover the environmental component of value (Tietenberg & Lewis, 2019). This

component of value is a use-value of the homeowner, not the general public. The hedonic model

allows for the measurement of the owner’s marginal WTP for marginal changes in individual

attributes (Tietenberg & Lewis, 2019). The HPM has been used widely in different studies

relating to the value of real estate (David, 1968; Gibbs et al., 2002; Young, 1984; Cordell and

Bergstrom, 1993; Loomis and Feldman, 2003; Kashian & Winden, 2020).

Lakes provide numerous benefits to society, including reduced flooding, biodiversity, and non-

use values (Netusil, 2005). The hedonic method provides a means for quantifying these benefits.

Lakes are also a source of recreation-based income that sustains small-town economies,

including many communities in northern Wisconsin (Macbeth, 1989). But Dziuk and Heiskary

(2003, p. 23) put it simply when they stated, “Those who do not fully appreciate the beauty and

recreational use of lakes may find the substantial income from lakes to be adequate justification

for keeping them in a healthy state.”

The change in property value as related to a change in water quality or proximity to shoreline has

been studied by several researchers. In Wisconsin, early work by David (1968) examined which

attributes of Wisconsin lakes attributed to increased property values for lake property. The study

took as a given that a customer’s WTP was greater for property that had lake access versus one

that did not. Lakes that were either swampy or had steep slopes had a significant negative

relationship with property values; while good water quality, high surrounding population, and a

higher number of nearby lakes had a significant positive relationship with property values

(David, 1968). There was not a statistically significant relationship between sales prices and road

access to the property, the amount of public land nearby, or the fluctuation of the lake level

(David, 1968). More recent studies, however, have found a significant correlation between lake

level fluctuation and property values (Cordell and Bergstrom, 1993; Loomis and Feldman, 2003;

Kashian & Winden, 2020).

The HPM was chosen for this study because the shoreline characteristics mostly affect the

property owner’s wellbeing, the hedonic method is appropriate for estimating these values.

Moreover, it is likely that most of the use-value of smaller lakes accrue to nearby residents,

rather than to people traveling for recreation. The remainder of this section will discuss how the

HPM has been used to determine WTP for variables relating to shoreland development, water

quality, lake proximity, and lake levels.

2.2.2.2. Shoreland Development

Shoreland development has been linked to changes in shoreland habitat and water quality.

Studies have looked at how Wisconsin’s shoreland management program (the nation’s first) has

affected lake shore aesthetics and controlled development. Nicholls & Crompton (2018), in their

literature review of 47 publications relating to the effects of the proximity and views of inland

lakes, found that studies consistently found water views and access were both capitalized into

residential property values. These studies reinforce the DNR’s statement that the purpose of

shoreland development regulations is to maintain safe and healthful conditions; prevent and

control water pollution; protect spawning grounds for fish and other aquatic life; control building

sites and the placement of structures and land uses. The regulations are also in place to preserve

shore cover and natural beauty (WDNR, 2017), which agrees with Klessig (2001) who stressed

both the spiritual, cultural, and emotional value of lakes. To put it simply, the WDNR regulations

were put in place to limit the direct and cumulative impacts of humans on water quality and

aquatic life (WDNR, 2017). Also in Wisconsin, Papenfus & Provencher (2005) in a hedonic

analysis in Vilas County, found that the county's shoreland zoning restrictions requiring

minimum shoreland frontages that were stricter than the state requirements, enhanced property

values. A study in Portland, Oregon, that studied the relationship between environmental zoning

and property values found that the net effect of environmental zoning on a property’s value was

dependent on the type of zoning and the property's location (Netusil, 2005).

In another Wisconsin lakes study, Jennings et al. (2003) found that the quantity of woody debris,

emergent vegetation, and floating vegetation decreased as the density of lakeshore development

increased and that littoral sediments contained more fine particles at these more densely

populated locations. Jennings et al. (2003) goes on to state, “When human activity changes the

physical structure of the environment, it alters the habitat within the lake.” Researchers have

found that by restoring shoreland complexity and minimizing the shade created by structures in

the littoral zone, shoreland habitats that have been compromised by human disturbances can be

restored (Munsch et al., 2017).

In nearly the same region as this current study, the northern forested region of Wisconsin,

Horsch & Lewis (2009) found, using information from 170 lakes, that the presence of the

common aquatic invasive species Eurasian Watermilfoil reduced property values by 13%. In the

same study, using the hedonic method they also found that lakes were more likely to be invaded

with the invasive milfoil if it is popular with recreational boaters (Horsch & Lewis, 2009).

Clapper and Caudill (2014) completed a study in Ontario, Canada, which found that the

shoreland length was positively correlated with lakefront cottage values. If that shoreline is

protected with erosion control measures such as a sea wall or rip rap, the value of the property is

increased by 10 percent per Jin et al. (2015), and Mooney and Eisgruber (2001) found that trees

in the riparian zone decrease property values by approximately 3%. This decrease was found

despite the fact that tree cover provides both shade and filtration. Trees slow surface runoff, and

the sediment carried by the runoff is filtered out before it is carried into the lake by the tree’s root

system. They attribute the decline in value to the decreased view of the waterfront (Mooney and

Eisgruber, 2001). While Netusil (2005), on the other hand, found that tree canopy was not

significantly correlated with property values.

Heinrich & Kashian (2010) found that although lake frontage had a positive effect on the price of

homes in rural Wisconsin, the impact was less noted on shallower lakes. These shallower lakes

usually contain more vegetation in the littoral zone due to sunlight being able to reach the lake

bottom of the shallower waters. Gibbs et al. (2002) found that lakes with a larger surface area

were related to higher property values. It is common knowledge in real estate that homes on

larger lakes sell for higher prices; larger lakes offer better recreational opportunities than smaller

lakes (Schnaiberg et al., 2002).

2.2.2.3. Water Quality

As far back as the early 1970s, we have been aware of the effects of increased runoff of

fertilizers in the form of phosphorus and nitrates in decreasing water clarity and increasing

primary production leading to low oxygen and, in severe cases, fish kills (Schindler et al., 1971).

Several studies have used hedonic techniques to value changes in water quality. Gibbs et al.

(2002) evaluated the effects of lake size and water quality. They found that water clarity was

more important to customers purchasing property on larger surface area lakes in all four markets

that they studied, and that water clarity had a non-linear relationship with price. The other

variables that Gibbs et al. (2002) found significant to the model were age of the structure, gross

living area (GLA), and depending on the market, the distance to lake, density, and tax rate. They

concluded that reductions in lake clarity due to eutrophication reduced property values and in

turn reduced both state and local tax revenues. Walsh (2009) found that water quality not only

affected the value of lakeshore properties but also affected the value of properties in close

proximity to the lake in a declining gradient as the distance to the lake increased.

Young (1984) analyzed the perceived influence of water pollution on seasonal homes on Lake

Champlain in northern Vermont. They used lot size, GLA, quality of construction, presence of a

garage, and presence of a porch, along with water quality factors as determinants of property

value; they found that the presence of a porch actually added substantial value to the property

and was probably acting as a proxy for one or more unobserved amenities. Water quality was

rated from 1-10, with 1 being poor quality and 10 corresponding to excellent quality. The study

found that each point decline in perceived water quality reduced the property’s value by $1,417

or approximately 6.3%. The properties located on the polluted bay with an average water quality

rating of 3.4 were 18.8 to 21% lower in value compared to the properties on the lake with an

average water quality rating of 6.4.

Using water clarity in a hedonic model of 113 lakes across the United States, Moore et al. (2020)

found that on average a 0.1-meter change in Secchi depth led to a 1% change in the property

value of homes near the study lakes. While Clapper and Caudill (2014) found that for each

additional foot of water clarity buyers were willing to pay an additional 2% for lakefront

property. In their meta-analysis, Guignet et al. (2022) used data from 36 unique hedonic studies

to determine elasticity estimates based on water quality metrics. The study found that on average

a 1% change in water clarity produced a 0.1167% change in property value. More recent work in

Wisconsin covering over 100 inland lakes by Wolf et al. (2022) found that decreases in water

quality of approximately 24% led to a loss in value of more than $7500 per parcel. While a

national study, also using the hedonic method, found that a 10% increase in water quality led to a

total increase of between $6 and 9 billion in property value (Mamun et al., 2023). While these

studies concentrated on the relationship between water quality and property value, our study

focuses on the lakeshore disturbances that affect water quality, and their relationship with

property values.

2.2.2.4 Lake Proximity

In a recent literature review, Nicholls & Crompton (2018) found that, in addition to lake

frontage, homeowners are willing to pay for views of and proximity to large lakes and reservoirs.

Thus, the economic value of a lake will be severely underestimated if only the WTP of shoreland

property owners is used for valuation. Accounting for the benefits to property owners in close

proximity to lakes adds millions of dollars to lake values (Nicholls & Crompton, 2018). Not only

is there a positive correlation between the lakes and the surrounding properties' values, but there

is also a correlation between lake attributes and the surrounding properties. However, there is a

point where the distance no longer matters; Lansford & Jones, (1995) found that the marginal

price of proximity falls at a declining rate and that there was little change beyond approximately

2,000 feet from the lake.

2.2.2.5. Lake Levels

Loomis and Feldman (2003) analyzed the effect of lowered water levels on property prices. They

found that lower water levels lead to mudflats between the property and the lake. These mudflats

were both unsightly and reduced access to the lake for boating and swimming. The lost use of the

amenity associated with the property was related to a reduced property value. Lansford & Jones

(1995) also found that lower lake levels, which exposed debris that would have been covered by

higher lake levels, were correlated with lower property values. Therefore, higher lake levels

increased WTP due to both greater accessibility and greater aesthetic value.

Loomis and Feldman (2003) used the hedonic method to determine the economic benefit of

maintaining a high lake level for a longer period of time frame during the summer recreation

season. They found that the number of bathrooms, distance to lake, feet of exposed shoreline,

garage spaces, proximity to a golf course, lake frontage, interest rate, building quality, and lake

view were statistically significant predictors of sales price. The lot size and GLA were only

significant in some model specifications. The study found that lower lake levels that exposed an

additional ten feet of shoreline were related to a 1% decrease in property value.

Kashian & Winden (2020), controlling for three different time periods on four different lakes,

evaluated the relationship between property appreciation and linear foot of shoreline. Their

model also used lot size/frontage, GLA, number of rooms, powder room, and bathroom as

independent variables; only the number of rooms, powder room, and bathroom were found to be

significant. The study concluded that the WDNR’s decision to lower lake levels in Lake

Koshkonong resulted in financial harm to both individual property owners and the surrounding

community.

2.2.2.6. Non-Market Use Value

These studies, most of which used the HPM, only capture the private benefit of the increased

value of each lakeshore property or properties in close proximity to the water body. Lakes also

provide ecosystem services, including reduced flooding, biodiversity, and non-use value, that

benefit society as a whole (Netusil, 2005). Even without taking into consideration the private

benefits and the ecosystem benefits, the recreational income produced by clean lakes sustains

many northern Wisconsin communities (Macbeth, 1989).

The above hedonic studies all find the non-market, non-consumptive use values of lakes and

shoreland amenities. Additional non-use values of water include existence, bequest, and option

values. When these additional non-use values are bundled together, a total non-use value can be

estimated. This non-use value can then be considered in conjunction with the use value to help

policymakers determine the most economical use of a resource. For example, Lansford & Jones

(1995) found that the net present value of water in both municipal and agriculture exceeded the

marginal recreational and aesthetic values derived from the hedonic method alone. Therefore,

they concluded that for a proper comparison, the additional components of non-use values

needed to be added to make a complete assessment (Lansford & Jones, 1995).

2.3. Study Area and Data

2.3.1. Study Area

The study area is located in the northeastern section of the state of Wisconsin in the US (Figure

1). Many of the state’s lakes are bordered by residential development both in the form of year-

round homes and seasonal homes. Some property owners around these lakes have made

substantial changes to the lakeshore habitat, while others have left the shoreland in its natural

state. The WDNR completed an extensive inventory of habitat disturbances on 130 lakes

between 2016 and 2020. A lakeshore disturbance as defined by the WDNR is a disruption or

alteration of the near shore environment, that has been determined to affect water quality. The

availability of the habitat data and sales data from the state’s Multiple Listing Service (MLS)

allow for a hedonic analysis to be performed. The 130 lakes cover an area of approximately 167

(64.5 mi

). The shapefile for the lakes was created from a shapefile of all the lakes in

Wisconsin obtained from the WDNR website (WDNR

, n.d.). The lakes range in size from less

than 13,000 m

(3.2 acres) to over 25,000,000 m

(6178 acres). The study lakes represent a total

lake shoreline of approximately 898,000 meters (558 miles), with the smallest being Bahr Lake

with a shoreline of approximately 724 meters (0.4 miles) and the longest being Tomahawk Lake

with a shoreline of approximately 49,800 meters (31 miles). A statewide geospatial database of

parcels was downloaded from Wisconsin’s State Cartographer’s Office website (Wisconsin State

Cartographer, n.d.). Parcel data was filtered for parcels that have shoreline on the study lakes in

ArcGIS Pro from Esri (ESRI, n.d.).

Figure 1 Study lakes and study area

for the hedonic models. The size of

the study lakes has been exaggerated

for visual clarity. Lake and County

shape files were downloaded from the

WDNR’s website (WDNR

, n.d.).

2.3.2. Housing Data

The housing data was obtained from Wisconsin’s MLS, a database created and maintained by

professional agents, brokers, and other real estate professionals who belong to Wisconsin’s

Realtors Association (WRA, 2024). The database includes details about listed, sold, expired, and

withdrawn properties listed by the members of the WRA from 2001 to 2022. The MLS data also

includes information on lot size, gross living area (GLA), number of bedrooms and bathrooms,

garage spaces, length of water frontage, the size and floor level each of the homes’ rooms are

located, zoning and school district information, the estimated year the home was built and much

more. The data is entered by the listing real estate agent at the time the property is listed for sale,

and updated with the sales information when the property is sold. Although the MLS data is

comprehensive, there are instances where data is missing or incorrect. The MLS data was

cleaned of entries with missing data for date of sale, zero or missing bathrooms, or missing GLA

information. The remaining MLS data was matched spatially to the parcel data using ArcGIS Pro

10.8.2 (ESRI, n.d.). The resulting dataset was then further verified by confirming addresses

matched in both sets of data. The resulting merged dataset contained 847 sold parcels on 62

lakes. Shoreland habitat data was then matched to the MLS sales data using the open-source

statistical computer software program R from the R Project website (R, n.d.).

The next step was to mitigate the temporal influences of inflation, unemployment, seasonal sales,

and mortgage interest rates, by adjusting the sales prices using the quarterly housing price index

for the state of Wisconsin from the Federal Reserve Economic Data (FRED, n.d.). FRED is a

publicly available online database containing economic time series data collected from both

government and private sources by the Research Department of the Federal Reserve Bank of St.

Louis.

2.3.3. Lake Data

Lake-specific variables were incorporated into the dataset and used in some model specifications

to control for observable differences in lake amenities. Specifically, lake area, and lake depth,

were added to the data. Both lake depth and surface area were obtained from the WDNR website

(WDNR

, n.d.).

2.3.4. Habitat Assessment

Shoreland habitat was assessed by volunteers using the Lake Shoreland & Shallows Habitat

Monitoring Field Protocol (WDNR, 2020). The assessment covered three zones for each parcel

on each study lake (Figure 2). The riparian buffer zone begins at the high-water level (HWL) and

extends inland 35 feet (Figure 3). The assessment used the term HWL versus OHWM due to the

potential legal ramifications. The littoral zone extends into the lake from the current water level.

The bank zone lies between the HWL and the bank toe (the place where the lakebed meets the

bank face). A sample of the data sheet the volunteers completed is available in the Protocol

document (WDNR, 2020).

Within the riparian zone, volunteers working under the direction of the WDNR recorded the

percentage of various land cover features; namely, tree canopy, shrubs, impervious area,

manicured lawn, and agricultural land. Percent cover was recorded to the nearest 5% for each

parcel. Volunteers also recorded the number of structures, boats, and fire pits in the riparian

zone, as well as whether the zone contained a point source, channelized flow area, stairway,

sloped lawn, or bare soil area.

In the bank zone, information was obtained on the percentage of the bank with artificial beach,

vertical sea wall, and riprap. The latter two variables are collectively referred to as erosion

control measures. Information on whether the bank was slumping and whether it was more or

less than 1 foot in height was also obtained. In the littoral zone, the number of human structures

was counted, including piers, boat lifts, swim rafts, boat houses, and marinas. The presence of

floating aquatic and/or emergent plants was noted. If the lake levels were lower than normal and

the lakebed was exposed, then observations were made of plants and other disturbances.

Using information from the riparian, bank, and littoral zones, the WDNR calculated a

disturbance index for each property. Specifically, the index was constructed based on the

percentage of manicured lawn, impervious surface, and agricultural land in the riparian zone, the

percentage of altered bank area (i.e., beach and erosion control), and structure density in both the

riparian and littoral zones. Properties with larger index values exhibit greater levels of shoreland

disturbance than properties with smaller values (WDNR, 2023). A lake-level disturbance index is

also constructed using shoreland information for all properties on the waterbody (WDNR, 2023).

The volunteers were instructed to circle the lake in a small boat. From the boat the volunteers

took georeferenced photos and recorded observed habitat disturbances for each parcel. Distances

and percentages were visually estimated for each disturbance recorded. Although the volunteers

were instructed to practice their visual perception of distance by calibrating their eyes using

known measured distances, these estimates still have the potential for bias and imprecision. All

visual estimations, particularly those taken from a distance offshore from a small boat are subject

to a degree of subjectivity, leading to imprecision and error.

Figure 2 The three habitat zones used in the

WDNR Habitat Assessment Study (WDNR,

2020)

Figure 3 Plan view of the riparian zone in

relation to the study parcels and lakes. Lake

shape file from the WDNR. Parcel shape file

from the Wisconsin's State Cartographers

Office (n.d.).

Incorporating variables from the WDNR shoreland study into the hedonic model allows for a

comprehensive analysis of how various aspects of shoreland habitat impact property values. The

inclusion of riparian-zone variables such as tree canopy percentage, manicured lawn percentage,

presence of bypass features, and number of structures provides insight into the aesthetic and

environmental characteristics of the shoreline. Tree canopy cover, for example, not only

enhances the visual appeal of the property but also plays a crucial role in filtering pollutants in

runoff, thereby improving water quality.

The presence of bypass features like channelized flows or point source outlets may impact water

quality by allowing runoff to bypass shoreland vegetation, potentially leading to decreased

nearshore water quality. Additionally, the percentage of manicured lawn in the riparian zone,

while offering recreational space and aesthetics, may lack the ecological benefits provided by

native plants and shrubs, such as reducing erosion and nutrient loading.

Moving to the bank zone, variables related to the presence of artificial erosion control measures

and artificial beach areas are considered. While erosion control measures like vertical sea walls

and riprap may offer protection to shorelines, they can negatively impact fish habitat and

macroinvertebrate communities. Artificial beach areas provide recreational benefits but may also

homogenize habitats, affecting aquatic species.

In the littoral zone, variables such as the total number of boat houses, piers, floats, and lifts are

included, reflecting the recreational amenities associated with waterfront properties. The

presence of floating or emergent vegetation and evidence of plant removal also contribute to

understanding how shoreline vegetation affects property values.

Furthermore, while this study cannot directly evaluate the effects of zoning ordinances on

property values, it can hypothesize that ordinances aimed at improving water quality may

indirectly influence property values by altering recreational and aesthetic benefits. Contrary to

popular belief, restrictive zoning regulations may not necessarily decrease property values, as

they can enhance the environmental quality and attractiveness of shoreland properties in the long

run.

Overall, integrating these variables into the hedonic model provides a nuanced understanding of

how different aspects of shoreland habitat influence property values, balancing aesthetic,

recreational, and ecological considerations. I hypothesize that many of the shoreland

characteristics are positively correlated with housing prices because they enhance recreational

and aesthetic values (e.g., tree canopy, manicured lawn, structures, beach, erosion control).

Conversely, the presence of channelized flows and floating or emergent vegetation is

hypothesized to have a negative correlation.

2.4. Methodology

The goal of this study is to evaluate how various shoreland features affect housing prices using

the HPM and to estimate marginal WTP values for these features. It is assumed that different

shoreland features have diverse effects on property values, just as they have varying impacts on

water quality and near-shore aquatic habitat. In this study, MLS sales data and information from

the shoreland habitat assessments completed by the WDNR are utilized. A theoretical model

used in this analysis is described below.

2.4.1. Theoretical Background

2.4.1.1. Linear vs Non-linear Regression Models

The HPM assumes a perfectly competitive market that determines the equilibrium price for a

home (Rahman, 2021). However, homes are not homogeneous; they are comprised of differing

bundles of structural, neighborhood, and environmental amenities whose quantity and quality

can cause variation in property values. In the hedonic model, differences in property values can

reveal the implicit price of non-market amenities. The implicit price for a marginal change in an

amenity (disamenity), the marginal WTP, is equal to the marginal benefit (loss) to the customer

from that change (Rahman, 2021). The sales price for a home can then be expressed as a multiple

regression equation where the sales price is the dependent variable, characteristics of the

property are independent variables, and estimated parameters represent the amount the price

changes given a change in the independent variables, all else equal. These changes can be either

negative or positive.

The hedonic regression model can take a linear form, as in equation 1, or a nonlinear form, as in

equation 2, where some or all of the independent variables are natural log transformed. It can

also take a nonlinear form where both the dependent variable and independent variables are

natural log transformed, as in equation 3. With the natural log (Ln) transformation, the regression

model can account for a nonlinear relationship. Hedonic property models often exhibit nonlinear

relationships between the dependent variables and the independent variables such as square

footage and lot size (Taylor, 2003). For instance, the value of a marginal increase in GLA

declines as GLA increases. The Ln specification can also help stabilize variance (i.e., reduce

heteroskedasticity), linearize relationships, and normalize skewed data, all of which can improve

a model’s performance (Sakia, 1992).

= β

+ β

+ e Equation 1

= β

+ β

ln(X

) + β

+ β

ln(X

) + e Equation 2

lnP

= β

+ β

ln(X

) + β

+ β

ln(X

) + e Equation 3

Where: P = property price

β = coefficients

X = property amenities

e = error term

The preferred model form is context-specific. Young (1984) found that the linear model

(equation 1) had greater predictive ability and produced coefficients with higher levels of

statistical significance, while Loomis and Feldman (2003) found that a non-linear model

performed better. Both studies compared results from linear and non-linear models estimated

with the same data, testing the sensitivity of results to the model form, and evaluating the

appropriateness of the independent variables included in the model.

2.4.1.2. Box-Cox Model

The Box-Cox model can be used to help identify an appropriate model form. Box and Cox

(1964) proposed a nonlinear transformation method that could make model residuals more

normal and less heteroskedastic. The transformation, which can be applied to any set of strictly

positive dependent and independent variables, is in the form of equation 4.

(λ)

= (Y

– 1)/λ Equation 4

While the estimated parameter lambda can take any value, when it is 1 the transformation is

equivalent to using a linear specification, and when it is 0 the transformation is equivalent to a

natural log specification. For this analysis, I estimated the Box-Cox model with transformations

on the sales price and several structural characteristics. Results suggest that a model with a log-

transformed dependent variable and log-transformed independent variables is preferred to a

linear model. Thus, this study employed the model form depicted in equation 3, where the

natural log of the dependent variable was regressed against the property amenities, some of

which were log-transformed.

2.4.1.3. Model Variables

Ideally, all determinants would be included in the hedonic model, but such a model would be

unmanageable due to the large number of factors that affect home prices (Young, 1984).

Furthermore, information for many factors is either not available or of poor quality. In this study,

information on the age of the structure was missing for a large portion of the sales transactions,

so it was not used in the model. An additional reason for limiting the number of independent

variables used in the model is multicollinearity, which occurs when two or more variables are

highly correlated. The inclusion of multicollinear variables can increase estimated standard error

and add little to the predictive ability of the model (Young, 1984). For example, Loomis and

Feldman (2003) omitted the number of bedrooms in their hedonic model due to high correlation

with the number of bathrooms and the GLA. The subset of predictors used in the model becomes

proxies for the related unused factors that also contribute to property values (Loomis and

Feldman, 2003).

The dependent variable used in this study is the natural log of the adjusted sale price (LnASP).

The LnASP ranged between 10.15 and 14.56, with a mean of 12.75 and a standard deviation of

0.54 (Table 2.4). Independent variables used in the model can be divided into three broad

characteristics (structural, neighborhood, and environmental) (Rahman, 2021). For structural

amenities, this study used the log of the home’s GLA (LnGLA), the number of garage spaces

(GS), total number of bathrooms (Tot_Bath), the presence of a dishwasher (DshWshr), the

presence of a washer (Wash), the presence of a dryer (Dry), the natural log of the area lot area

per GIS (LnAGIS), and the natural log of the ratio of lot area to shoreland length

(LnAcresPerSL). AcresPerSL is similar to the lot size per lake frontage variable found to be

significant in Kashian & Winden (2020). The average property in this study has 1.7 garage

spaces and 1.5 baths; its GLA is 1687 square feet, and it sits on a 1.2-acre lot that has a shorter

shoreline than its lot depth. Approximately 44% of the study properties have a washer and dryer.

The variables chosen for this study are a combination of those found to be significant in previous

literature and those that were found to be complete in the dataset. The statistics for each variable

are listed in Table 2.4.

The neighborhood amenities are controlled for using lake-level fixed effects. Identical properties

on different lakes can sell for a different price due to differences in lake qualities not addressed

in the model (lake area, depth, water quality, fish stocks, etc.). The fixed effects account for the

overall effect of each lake’s characteristics on home prices, assuming lake qualities remain

relatively constant over the study period. Furthermore, because most lakes in the study are small,

the lake fixed effects also control for the location-specific factors like distance to larger

population centers, school district, and nearby environmental amenities. In some model

specifications, the area of the lake in acres (LakeA) and the depth of the lake in feet (LakeD) are

used in place of the lake fixed effects. The area of the study lakes ranges between 13 and 6215

acres, with the average just over 2000 acres. The average lake depth of the study lakes was

between 8 and 84 feet, and the mean average lake depth was just over 32 feet (Table 2.4).

The environmental amenities are obtained from the shoreland habitat assessment. The

characteristics in the riparian zone that were used were the percentage of forested cover

(CANOPY_PCT), the presence of a bypass (Bypass_Pres), the percentage of manicured lawn

(LAWN_PCT), and the number of structures (RipStruc). In the bank area, the characteristics that

were used were the percent of the shoreline that contains artificial erosion control (TotalEC_Pct),

the presence of shoreline erosion (EROSION_PRES), and the percent of shoreline that contains

an artificial beach (BeachPerc). Lastly, in the littoral zone, the characteristics used were the

presence of vegetation (Veg_Pres), evidence that vegetation was removed (Veg_Rem), and the

number of structures (LitStruc). The average number of structures in the littoral and riparian

zones is 1.7 and 1.5, respectively. Some model specifications also use a property-level index of

shoreland disturbance developed by the WDNR (MMI_P). This index is created from numerous

shoreline features, where higher values indicate properties with a greater degree of shoreland

disturbance. The index is a sum of the percentages of shoreland disturbances in the riparian,

littoral and bank zones (WDNR

, n.d.). The index ranges between 0 and 400, with the average

disturbance being 214.19 and a standard deviation of 112.32.

In addition to the structural, neighborhood, and environmental variables, additional variables

were used to control for time fixed effects. A similar property with the same bundle of attributes

may sell at a different price in the spring than in the fall of the same year or at a different price

the following spring due to demand, interest rates, etc. (Rahman, 2021). These differences can be

seen in Figure 4, which depicts the median sales value of homes in Wisconsin by sale year and

quarter. This study used fixed effects for both the year (SaleYear) and the quarter (SaleQrt) that

the property was sold, determined by the proprietary sales date listed in the MLS dataset. The

independent variables used in this study are listed in Table 2.2, and the descriptive statistics for

each variable are presented in Table 2.4.

Figure 4 Median sales price of Wisconsin homes from data collected by the Wisconsin Realtors

Association per sales quarter for years 2007-2023, data is updated daily and downloaded from

their website on 2 February 2024 (WRA, 2024).

2.4.1.4. Model Specifications

This study estimated five different hedonic model specifications. The first model consisted of the

structural variables and the lake, year-, and quarter-fixed effects. The second model used the

structural variables, fixed effects, and the property-level shoreland disturbance index developed

by the WDNR. In the third model, the disturbance index was replaced by the shoreland

characteristics in the riparian, bank, and littoral zones. The fourth model included the same

structural variables and shoreland variables as the third model but replaced the lake-fixed effects

with the lake area and lake depth variables. Lastly, the fifth model is a parsimonious

specification that removes some of the variables that were not statistically significant in the other

specifications. The models were estimated in the program R using lm_robust found in the

estimatr package (R, n.d.), which generates robust standard errors. A list of the variables

included in each model is provided in Table 2.3.

$100000

$150000

$200000

$250000

$300000

1 2 3 4

Median Sales Price in Dollars

Sales Quarter

Median Sales Price of Wisconsin Homes

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Sales Year

Table 2.2 Hedonic Property Model Variable Labels and Descriptions by Variable Type

Variable Type

Variable

Description

Structural

LnGLA

Natural log of the buildings gross living area in feet

squared

Number of garage spaces

Tot_Bath>2

Presence of more than 2 baths (0 if not present, 1

otherwise)

Tot_Bath2

Presence of 2 baths (0 if not present, 1 otherwise)

Tot_Bath1

Presence of 1 bath (0 if not present, 1 otherwise)

DshWshr

Presence of a dish washer (0 if not present, 1

otherwise)

Washer

Presence of a clothes washer (0 if not present, 1

otherwise)

Dryer

Presence of a clothes dryer (0 if not present, 1

otherwise)

LnAGIS

Natural log of the area of the parcel per GIS in acres

LnAcresPerSL

Natural log of the Parcels area in acres divided by

shore length in feet

Neighborhood

LakeA

Area of the lake in acres

LakeD

Depth of the lake in feet

Environmental

MMI_P

Shoreland habitat disturbance index

CANOPY_PCT

Percent of parcel shoreline covered in tree canopy.

LitStruc

Number of structures in the littoral zone

RipStruc

Number of structures in the riparian zone

TotalEC_Pct

Percent of parcel shoreline containing artificial erosion

control.

EROSION_PRES

Presence of indication of shoreland erosion (0 if not

present, 1 otherwise)

BeachPerc

Percent of parcel shoreline containing artificial beach.

LAWN_PCT

Percent of parcel in the riparian zone containing a

manicured lawn.

Bypass_Pres

The presence of a feature that allows runoff water to

bypass the filtration effect of the canopy (0 if not

present, 1 otherwise).

Veg_Rem

Presence of indication of plant removal activity (0 if

not present, 1 otherwise)

Veg_Pres

The presence of floating and/or emergent vegetation in

the littoral zone (0 if not present, 1 otherwise)

Fixed Effects

Variables

SaleYear

The year the property sold (2001 – 2022)

SaleQrt

The quarter the property sold (1,2,3 or 4)

LAKE_NAME

The name of the lake the property abuts.

2.4.2. Hot Spot Analysis

Variation in sales prices and housing characteristics across space is known as spatial

heterogeneity. Anyone working with real estate knows that location matters; the three most

important attributes of real estate are “location, location, location”. The First Law of Geography,

according to Waldo Tobler (1970), is "everything is related to everything else, but near things are

more related than distant things”. This is especially true for property values, where a real estate

professional determines the value of a property by finding nearby properties (location) with

similar attributes to the subject property that have sold recently (time). The professional then

compares these properties to the subject property to make an estimate of its value. Therefore, if

the most important attribute is location, the second most important attribute is time. To evaluate

patterns in spatial heterogeneity in property attributes, I used Hot Spot Analysis.

Hot Spot Analysis can determine if the property characteristics are clustered spatially using the

Getis-Ord Gi* statistic, which is calculated and mapped using the Hot Spot Analysis tool in

ArcGIS Pro (n.d.). The analysis assesses whether high or low values for a variable are

concentrated spatially using the Gi* statistic, which is a z-score (ArcGIS Pro, n.d.).

The G

statistic looks at each feature within its own neighborhood to determine not only does it

have a high or low value but is it statistically significant. To be significant a feature must not just

be high or low, but it must be surrounded by similarity high or low neighbors. The features

value, along with its neighbors’ values are summed and compared with the proportionate sum of

all the features to determine if the sum is too large or small to be the result of just random chance

(ArcGIS Pro, n.d.).

Figures 6 and 7 depict the spatial variation of z-scores for two variables: the natural log of GLA

(LnGLA) and the ratio of lot area to shoreland length (AcresPerSL). These variables were

selected because the size of the home and the length of the shoreline are recognized as two

primary factors influencing a customer’s willingness to pay for lake property in the real estate

industry. The statistics for these variables are detailed in Table 2.4.

The z-scores indicate the number of standard deviations (SD) an observed value deviates from

the mean of the variable's distribution. A z-score falling between ±1.65 and 0 suggests random

attribute variation, resulting in a Gi_Bin or confidence level score of 0. Conversely, if the score

is < -1.65 or > 1.65, the variation is considered significant, leading to a Gi_Bin score ranging

between ± 1 to ± 3. These bins represent the number of SDs the attribute value deviates from the

mean, as illustrated in the example normal distribution curve in Figure 5.

In Figures 6 and 7, the Gi_Bin scores for each parcel are mapped at their respective locations.

Remarkable spatial clustering of property characteristics is observed, both around lakes and

groups of lakes. These findings suggest the presence of unobserved factors related to lakes or

lake groups that may impact sales prices. Hence, this supports the utilization of lake-level fixed

effects in the regression model to account for these factors.

Figure 5 Normal Distribution Curve (ArcGIS Pro, 2023)

Table 2.3 Model Specifications with Independent Variables Grouped by Variable Type

Model

Structural

Environmental

Location

Fixed Effects

LnGLA

Tot_Bath

DshWshr

Wash

Dry

LnAGIS

LnAcresPerSL

SaleYear

SaleQrt

LAKE_NAME

LnGLA

Tot_Bath

LnAGIS

LnAcresPerSL

MMI_P

CANOPY_PCT

Veg_Pres

Veg_Rem

SaleYear

SaleQrt

LAKE_NAME

LnGLA

Tot_Bath

LnAGIS

LnAcresPerSL

CANOPY_PCT

LitStruc

RipStruc

TotalEC_Pct

EROSION_PRES

Bypass_Pres

LAWN_PCT

BeachPerc

Veg_Pres

Veg_Rem

SaleYear

SaleQrt

LAKE_NAME

LnGLA

Tot_Bath

LnAGIS

LnAcresPerSL

CANOPY_PCT

LitStruc

RipStruc

TotalEC_Pct

EROSION_PRES

Bypass_Pres

LAWN_PCT

BeachPerc

Veg_Pres

Veg_Rem

LakeA

LakeD

SaleYear

SaleQrt

LnGLA

Tot_Bath

LnAGIS

LnAcresPerSL

LitStruc

TotalEC_Pct

EROSION_PRES

BeachPerc

SaleYear

SaleQrt

LAKE_NAME

Table 2.4 Descriptive Statistics for Variables Used in the Hedonic Property Model

Max.

Min.

Avg.

Std. Dev.

Dependent Variable

LnASP

14.563

10.151

12.748

0.536

Structural Variables

1.721

1.160

LnGLA

9.055

5.768

7.320

0.466

Tot_Bath

1.534

0.499

DshWshr

0.388

0.488

Dry

0.440

0.497

Wash

0.445

0.497

LnAGIS

3.959

-2.641

-0.572

0.944

LnAcresPerSL

-1.762

-6.907

-5.118

0.648

Environmental Variables

MMI_P

400

214.187

112.315

Veg_Pres

0.419

0.494

Veg_Rem 1 0 0.152 0.360

CANOPY_PCT

100

50.248

35.691

Bypass_Pres

0.649

0.477

TotalEC_Pct

100

34.536

43.842

EROSION_PRES 1 0 0.095 0.293

LAWN_PCT

100

56.954

37.462

BeachPerc

100

4.317

18.253

LitStruc

1.721

1.250

RipStruc

1.543

2.075

Location Variables

LakeA

6215

2159.711

2754.362

LakeD

32.411

10.961

Fixed Effects Variables

SaleYear

2022

2001

2014.64

4.917

SaleQrt

2.519

1.033

Figure 6 Hotspots for the variable of the natural log of the quotient derived when the area of the

lot in acres is divided by the length of the shoreline in feet.

Figure 7 Hotspots for the variable of the natural log of the home's gross living area in square

feet.

2.5. Results

This chapter will discuss the results from the five hedonic model specifications (models 1-5). All

five hedonic models used the natural log of the adjusted sales price (LnASP) as the dependent

variable, and all models exhibited statistically significant F-statistics (p-value of 2.2e-16),

indicating that the models provide a better fit to the data than the sample mean.

2.5.1. Hedonic Model Results

Parameter estimates are reported in Table 2.5, and overall model statistics (i.e., error term

distribution, F-statistic) are provided in Table 2.6. A summary of results for shoreland variables

by shoreland habitat zone is provided in Table 2.8. All models contained fixed effects for the

sales year and quarter, while models 1-3 and 5 also included fixed effects for the 62 study lakes.

Models were estimated with robust standard errors in R. At their largest, residuals ranged

between -2.3 and 1.1, with a median value of 0.05. In all five models, the residuals were

approximately normally distributed around 0 without any extreme outliers. The F-Statistics of

models 1-3 and Model 5 ranged between 12.52 and 13.38, with degrees of freedom (DF)

between 746 and 753. Model 4, which did not include the lake fixed effects, had an F-statistic of

18.22 and DF of 805.

The first model (Model 1) included each property’s structural variables and fixed effects for the

year of sale, quarter of sale, and the lake the property was abutting. Model 1 has an r

of 0.6205,

meaning that the model explains approximately 62.1% of the variability in the log of property

values. The variables for the presence of appliances (Wash, Dry, and DshWash) are not

statistically significant. However, the variables describing the lot (LnAGIS and LnAcresPerSL)

were both significant, with LnAGIS at p < 0.01 and LnAcresPerSL at p < 0.05. The variable

LnAGIS was positively correlated with property value, indicating that larger lots sold for more,

all else being equal. The variable LnAcresPerSL was negatively correlated with property value,

implying that properties with less shoreland per acre sold for less. The variables GS and LnGLA

were also significant at p < 0.01, showing a positive correlation with property value, meaning

that larger homes and homes with more garage spaces had higher sales prices. Homes with more

than two bathrooms (Tot_Bath > 2) also had higher sales prices. The fixed effects accounted for

the differences of both time and location with varying degrees of significance and correlation.

The next model (Model 2) incorporated several environmental variables: each property’s

shoreland disturbance index (MMI_P), the percent of canopy cover (CANOPY_PCT), and the

variables indicating the presence and removal of vegetation in the littoral zone (Veg_Pres,

Veg_Rem). As previously described, the habitat disturbance index is determined using the

percent of the riparian zone that is covered by a manicured lawn, the density of structures in both

the riparian and the littoral zones, and the percent shoreline with bank alterations. The variables

for the presence of appliances (Wash, Dry, and DshWash) were dropped from this model due to

a lack of statistical significance. Model 2 has an r

of 0.6204, which is slightly less than the

previous model. Therefore, the addition of the MMI_P variable, along with the canopy and

vegetation in the littoral zone variables, did not meaningfully affect the model’s overall fit to the

data. The adjusted r

also decreased from 0.5736 in Model 1 to 0.5729 in Model 2. The adjusted

makes an adjustment for the number of variables in the model. The MMI_P variable, however,

is statistically significant to the model at p < 0.01 and has a positive correlation with the log of

the property’s value, meaning that properties with greater levels of shoreland disturbance sold for

more, all else being equal. The sign and significance of the structural variables in this model

were similar to those from Model 1.

Model 3 replaced the shoreland disturbance index with measures of the index's various

components, while still including variables on canopy coverage and vegetation. The r

for Model

3 increased slightly over Model 1, explaining 62.7% of the variability in property value. The

addition of structures in the littoral zone added to the value of lake property at a significance

level of p < 0.05, while the number of structures in the riparian zone was not significant. The

percentage of the shoreline that had erosion control, such as a vertical wall or rip rap, had a

positive effect on value at a significance level of p < 0.01, while the presence of erosion was

negatively correlated with price at p < 0.05. The percentage of the riparian area with tree canopy

and the percentage with manicured lawn was again not significant. The coefficient pertaining to

the presence of a stormwater bypass was negative but also not significant. However, the

percentage of artificial beaches did have a positive and statistically significant effect on property

value at p < 0.05. In the littoral zone, the presence of floating and/or emergent weeds did not

have a significant effect on value, and neither did evidence of the removal of vegetation. The

sign and significance of the structural variables remained unchanged Table 2.5.

In Model 4, lake fixed effects were replaced with variables describing the lake area (LakeA) and

lake depth (LakeD). The fixed effects were removed to evaluate the robustness of results to a

model specification that did not control for location-specific factors, like distance to a larger

population center, school district, and nearby environmental amenities. Without the lake fixed

effects, the r

was only 0.4813, effectively reducing the model’s predictive abilities by 15%. The

lake depth and lake area were both significant at p < 0.01. Lake depth was positively correlated

with sales price, while lake area was negatively correlated with the property value. The removal

of the lake fixed effects also affected the significance and magnitude of some of the

environmental and structural variables. Most notably, the lot size and ratio of lot size to shoreline

were no longer significant. Moreover, the absolute value of the coefficients for the percentage of

shoreline with a beach, percentage of shoreline with erosion control, presence of a bypass, and

number of structures in the littoral zone increased.

The final model (Model 5) is a parsimonious specification; it is similar to Model 3 but excludes

shoreland variables that were not statistically significant (i.e., LAWN_PCT, CANOPY_PCT,

RipStruc, Bypass_Pres, Veg_Pres, and Veg_Rem). Excluding these variables reduced the r

value from 0.6239 to 0.6230, but due to the reduced number of variables, the adjusted r

value

increased from 0.5741 to 0.5765. The included structural and shoreland variables are all

significant, and the sign and magnitude of the coefficients are all similar to those in Model 3.

Model 5 is the preferred model specification; it includes those variables that influence the value

of lake shore properties along sampled lakes. The resulting regression equation is:

lnASP = 8.95 + (0.48) LnGLA + (0.07)GS + (0.13)TotBath>2 + (0.09)LnAGIS + (-

0.08)LnAcresPerSL + (0.002)BeachPerc + (0.03)LitStruc + (0.001)TotEC_Pct + (-

0.11)ERSION_PRES0 + β(LAKE_NAME) + β(SaleYear) + β(SaleQrt) + e Equation 5

Where: β = coefficients

LAKE_NAME = Lake fixed effects

SaleYear = Sale year fixed effects

SaleQrt = Sale quarter fixed effects

e = The error term

Results of the five regression runs are shown in Table 2.5. The table consists of the coefficients

and p-values for each variable included in each of the models along with r

and adjusted r

for

each of the models. The statistics defining the residuals for each of the models are in Table 2.6.

2.5.2. Marginal Willingness to Pay

In Table 2.7, I present the estimated change in the adjusted value of a property given a one-unit

change in an independent variable (i.e., the marginal WTP values). These estimates are based on

Model 5 and are calculated using sample averages. The average value of a property in the study

is $392,700.04. Marginal WTP values for the structural variables are reasonable and consistent

with industry norms: an additional square foot of GLA increased the property value by $111.20,

and an additional garage space increased the property value by $26,271.63. Compared to a

property with less than two bathrooms, having more than two bathrooms increased the property

value by $49,126.78. An additional acre in lot size added $31,069.18 to the value of the property.

Most of the shoreland variables were positively correlated with property value, with the

exception of the presence of erosion in the riparian zone, which exhibited a negative relationship.

The presence of erosion decreased the property value by $42,450.87. An additional percent of

artificial beach added $628.32 to the property value. Since the average percent of artificial

beaches is 4.3 percent, artificial beaches add (4.3 * $628.32) $2,701.78 to the value of the

average study property, assuming constant marginal WTP. An additional structure in the littoral

zone adds $11,113.41 to the property value, and each additional percent of erosion control adds

$431.97 to the value. Since the average percent of erosion control per property is 34.5 percent,

erosion control measures add approximately (34.5 * $431.97) $14,902.97 to the value of the

average property. Likewise, the average number of structures in the littoral zone per property is

1.72, and the increase in property value due to these structures is (1.72 * $11,113.41), which is

$19,115.07 for the average property.

Although the amount that the presence of erosion correlates to reduced WTP is high at $-

42,450.87, erosion is only observed at a small fraction of properties, leading to a small effect on

the average property (0.096 * $-42,450.87) at $-4,075.28. In summary, the value added from

features classified as shoreland disturbances is equal to ($2,701.78 + $14,902.97 + $19,115.07),

which is $36,719.82 on average for each property, and subtracting the effects of erosion, the total

becomes ($36,719.82 - $4,075.28), which is $32,644.54.

Table 2.5 Hedonic Property Model Results (Dependent Variable: Natural Log of Adjusted

Sales Price)

Variable Model 1 Model 2 Model 3 Model 4 Model 5

LnGLA

0.4847***

(0.0000)

0.4832***

(0.0000)

0.4798***

(0.0000)

0.4324***

(0.0000)

0.4778***

(0.0000)

Garage Spaces

0.0613***

(0.000003)

0.0694***

(0.00000)

0.0657***

(0.000001)

0.0528***

(0.0001)

0.0669***

(0.000001)

Total Bath >2

0.0971*

(0.0821)

0.1211**

(0.0302)

0.1258**

(0.0245)

0.2506***

(0.00003)

0.1251**

(0.0195)

-0.0161

(0.6551)

0.0009

(0.9797)

-0.0002

(0.9958)

0.0360

(0.3591)

0.0016

(0.9636)

Presence of a Dryer

0.1080

(0.2020)

Presence of a Washer

-0.0148

(0.8608)

Presence of a Dishwasher

0.0010

(0.9688)

Ln of Lot Acres per GIS

0.0819***

(0.0020)

0.1215***

(0.00003)

0.0973***

(0.0007)

0.0193

(0.4647)

0.0927***

(0.0006)

Ln of Acres/Shore Length

-0.0731**

(0.0431)

-0.1008***

(0.0066)

-0.0916**

(0.0157)

0.0379

(0.2672)

-0.0816**

(0.0247)

Shoreland Metric

0.0004***

(0.0028)

Manicured Lawn Percent

-0.0002

(0.6354)

-0.0003

(0.5805)

Canopy Percent

0.0005

(0.2738)

0.0002

(0.5734)

0.0003

(0.5209)

Beach Percent

0.0017*

(0.0314)

0.0020**

(0.0206)

0.0016**

(0.0241)

Structures Lit Zone

0.0296**

(0.0128)

0.0732***

(0.0000)

0.0283**

(0.0151)

Structures Rip Zone

-0.0052

(0.4819)

-0.0115

(0.1108)

Erosion Control %

0.0012***

(0.0018)

0.0023***

(0.0000)

0.0011***

(0.0022)

Erosion Present

-0.1138**

(0.0120)

-0.1493***

(0.0028)

-0.1081**

(0.0152)

Bypass_Pres

-0.00051

(0.6063)

0.0049

(0.8818)

Vegetation Present

0.0280

(0.3649)

0.0210

(0.5020)

0.0231

(0.4649)

Vegetation Removed Pres

- 0.0070

(0.8632)

-0.0084

(0.8373)

-0.0565

(0.1751)

Lake Depth

0.0072***

(0.000003)

Lake Area

-0.00002**

(0.0038)

Constant

9.0197***

(0.0000)

8.7218***

(0.0000)

8.8934***

(0.0000)

9.6564***

(0.0000)

8.9529***

(0.0000)

Year fixed effect

Yes

Quarter fixed effect

Yes

Lake fixed effect

Yes

Observations

847

0.6205

0.6204

0.6270

0.4813

0.6260

Adjusted r

0.5736

0.5729

0.5769

0.4548

0.5792

Note: Robust standard errors reported in parentheses.

*p<0.1; **p<0.05; ***p<0.01

Table 2.6 Overall Model Statistics (Residuals, Degrees of Freedom, F-Stat, and p-value) for

the Five Hedonic Property Models.

Statistic

Model 1

Model 2

Model 3

Model 4

Model 5

Residual Min

-2.03476

-2.09613

-2.0811

-2.28197

-2.08227

Residual 1Q

-0.11983

-0.12016

-0.1215

-0.17524

-0.13037

Residual Median

0.03804

0.04247

0.0410

0.04926

0.03814

Residual 3Q

0.19739

0.20149

0.1954

0.23040

0.19638

Residual Max

0.73206

0.79275

0.7957

1.08034

0.79747

Residual Standard Error

0.3499

0.3502

0.3485

0.3957

0.3476

753

752

746

805

752

F-Stat

13.24

13.07

12.54

18.22

13.39

p-value

2.2e-16

Table 2.7 Marginal Willingness to Pay Per Lakefront Property in Northeastern Wisconsin

(2022 U.S. Dollars)

Variable

Model 5

Ave. Property

ASP

392,700.04

LnGLA

0.4778*** (0.00000)

111.20

Garage Spaces

0.0669*** (0.000001)

26,271.63

Total Bath >2

0.1251** (0.0241)

49,126.78

Ln of Lot Acres per GIS

0.0927*** (0.0006)

31,069.18

Ln of Acres/Shore Length

-0.0816** (0.0247)

-1,791.35

Beach Percent

0.0016** (0.0241)

628.32

Structures Littoral Zone

0.0283** (0.0151)

11,113.41

Erosion Control Percent

0.0011*** (0.0022)

431.97

Erosion Present

-0.1081** (0.152)

-42,450.87

Table 2.8 Summary of Shoreland Disturbance Results by Shoreland Habitat Zone

Shoreland Habitat Zone

Environmental Variable

Statistical Significance

(p<0.05)

Littoral

Structures

Yes Positive

Vegetation Present

Vegetation Removed

Bank

Erosion

Yes Negative

Percent Erosion Control

Yes Positive

Presents of a Bypass

Riparian

Percent of Manicured Lawn

Canopy Present

Structures

Percent Artificial Beach

Yes Positive

2.5.3. Robustness Checks

To check the robustness of results, several additional model specifications were estimated using

various subsets of the data. The Model 3 specification was used for these robustness checks.

First, to confirm that results are not driven by older sales occurring before the shoreland surveys

were conducted, two models were estimated using shorter sales periods. Model 3a was estimated

using property sales from 2010-2022 rather than 2001-2022. Model 3b was estimated using

property sales from 2015-2022. Additionally, Model 3d was estimated for the period 2001-2022,

using homes with at least 750 square feet of GLA, thus removing several small dwellings from

the analysis. The results for these models are reported in the Appendix. With a few exceptions,

the variables that were significant in Model 3 remained significant in models 3a, 3b, and 3d. In

particular, the variables Total Bath>2, LnAcresPerSL, and LitStruc were not significant in one or

more of the models. The remaining structural and shoreland variables remained statistically

significant in all models. The coefficients for these variables in models 3a, 3b, and 3d varied

minimally from those in Model 3.

A fixed effects model for school districts was also considered (Model 3c, see Appendix). Only

one of twenty-three school districts was found to have a statistically significant effect on home

prices, relative to the reference district. This finding was expected because most lakes are in a

single school district, with only three out of sixty-one lakes having properties in two different

school districts. For this reason, the school district fixed effects were not used in the preferred

model specifications. Models 3a-e verify the robustness of the hedonic models because the main

findings did not change when potentially biasing factors like older or smaller homes are excluded

from the analysis, or the time frame of sale, or when the effect of school districts are considered.

2.6. Discussion and Conclusion

This study analyzed how shoreland features classified as habitat disturbances by the Wisconsin

Department of Natural Resources (WDNR) affect the value of residential property along some of

Wisconsin’s inland lakes. These disturbances are often considered amenities by property owners.

For example, the presence of a tree canopy, which improves water quality by providing both

shade and filtration, is a shoreland disturbance when it is removed, but its presence can also

obstruct the lake view (Mooney and Eisgruber, 2001). The presence of a forested canopy was

found to be insignificant in predicting changes to the shoreland property values, which is in

agreement with previous literature by Netusil (2005). On the other hand, Mooney and Eisgruber

(2001) found that trees in the riparian zone decrease property values by approximately 3%.

Although homeowners and the WDNR share some objectives regarding shoreland conditions,

their interests do not always align. Homeowners seek to manage shorelands to maximize

recreational and aesthetic benefits, while the WDNR is primarily concerned with protecting

shoreland habitat and surface water quality. Results from this study suggest these differences

may be more pronounced for some shoreland features than others. For instance, the study found

that the presence of floating and/or emergent plants in the littoral zone did not have a significant

correlation with property value, and neither did the removal of that plant life. The WDNR finds

plant life in the littoral zone to be beneficial for fish habitat. Likewise, the percentage of

manicured lawn in the riparian zone is not significantly correlated with property value, even

though the substantial resources expended to install and maintain a manicured lawn suggest it

would increase homeowner utility. Lawns are a potential source of degradation for the habitat of

near-shore fish and other aquatic species. The presence of major erosion reduces property values

and degrades shoreland habitat; it represents a shoreland feature where homeowners and the

WDNR have an unequivocal, shared interest in addressing.

In contrast, artificial beaches, erosion control, and structures in the littoral zone, including piers,

boat lifts, and swim rafts, are positively correlated with property value. These features can lead

to increased sediment loading and interfere with fish habitat by limiting feeding and reproduction

opportunities (Munsch et al., 2017). In particular, artificial erosion measures known as armoring

have been shown to lead to less beach spawning and the consumption of less epibenthic and

terrestrial prey (Munsch et al., 2017). Thus, there is a negative externality associated with the

installation of shoreland features for recreational and aesthetic purposes.

Although habitat disturbances have been found to decrease water quality and shoreland habitat

conditions, they can be either positively or negatively correlated with increases in property

values. The improvements that owners have made to increase the value of their properties and

that have been found to be significant in the model (i.e., artificial beach, erosion control

measures, and structures in the littoral zone) accounted for a total of $36,719.82 of the average

property’s value. A study of lakes across the country found that a 0.1-meter (~ 4 inches) increase

in the Secchi depth measurement increased property values by 1% (Moore et al., 2020). Since the

percentage that the lakeshore disturbances increased value is equal to 9.35%

((36,719.82/392,700.04)*100) we can determine the following. To match the effect that the

significant disturbance variables have on property values, clarity would need to be improved by

(9.35 *0.1) 0.94 meters or ~ 3 feet. Secchi depths on the study lakes range from 2.57 to 20.17

feet with an average of 9.28 feet; a 3-foot increase in depth would represent a 32% improvement

in water clarity for the average lake. However, using Clapper and Caudill’s (2014) findings that

willingness to pay increases 2% for every additional foot of clarity, (9.35/2) a 4.68-foot

improvement in water clarity would be required. Lastly, using the data compiled by Guignet et

al. (2022), who reviewed studies nationwide and determined that in the Midwest on average a 1

percent increase in water clarity will lead to an 0.165% increase in lakeshore property value, it

can be estimated that a (9.35/0.165) 56.67% increase would be required. For the average lake in

the study this would be equal to (9.28 * 0.5667) 5.26 feet of additional clarity. Averaging the

three aforementioned studies (3, 4.68, 5.26) we can estimate that to match the effect that the

significant disturbance variables have on property values, clarity would need to be improved by

4.31 feet or approximately ((4.31/9.280) *100) 46.48%.

2.7. Future Research

The WDNR is continuing its survey work on additional lakes, and as this new data becomes

available, this study should be updated to include the additional data. Future research topics

could also evaluate whether shoreland development affects the value of neighboring parcels (i.e.,

whether shoreland development features positively or negatively affect the recreational and

aesthetic benefits of neighboring properties) and how lake regulations such as no-wake hours

interact with shoreland disturbances, water quality and habitat conditions, and homeowner WTP.

Although many of these factors are captured in the lake fixed effects, it would be interesting to

break out a few of these variables in future work.

3. Lake Water Quality Analysis

3.1. Introduction

Increased shoreland disturbances can lead to an increase in runoff reaching surface waters,

resulting in elevated levels of silts, salts, and clays, thereby contributing to reduced water clarity

and increased sediment accumulation (Van Hengstum et al., 2007). In the first chapter I analyzed

how these disturbances affected property values, in this chapter I will look at how shoreland

disturbance and nearby land use affect lake water quality.

Previous studies have attempted to answer this question in different ways. For example, core

samples extracted from the lake bottom can be analyzed for both physical composition and age

to determine historical runoff levels reaching the lake (Van Hengstum et al., 2007). These

sediment samples offer insights into the water clarity of the lake over time, dating back to

periods before European settlement. A study of four Wisconsin lakes, analyzing sediment cores

from their bottoms, revealed that sedimentation rates rose alongside changes in land use,

transitioning from pre-settlement land cover to seasonal homes, and finally to year-round

residences (Garrison et al., 2000). The transition to year-round homes saw a notably greater

increase in sedimentation compared to the shift to seasonal homes, likely attributable to the

proliferation of manicured lawns, accessory structures, and impermeable surfaces, all of which

contribute to heightened runoff when small seasonal cottages are replaced with more intensively

utilized year-round residences (Garrison et al., 2000).

This study aims to utilize water clarity data, land use information, and shoreland disturbance data

to ascertain which land use categories impact changes in water clarity across 164 Wisconsin

lakes. Additionally, it seeks to determine whether an overall lakeshore disturbance index can

elucidate variations in water clarity, while controlling for differences in land use.

In a subsequent analysis, this study delves into whether shoreland disturbances correlate with

lake conditions, as evidenced by water clarity. While prior research has established that land use

surrounding water bodies affects water quality, limited information exists regarding the

quantitative relationship between water quality and shoreland disturbances. This analysis

employs a regression model to evaluate the relationship between lake water clarity (a proxy for

water quality) and an aggregate measure of shoreland disturbance developed as part of the

WDNR Lake Shoreland and Shallows Habitat Monitoring Program, while also controlling for

lake attributes and nearby land use.

3.2. Literature Review

3.2.1. Water Quality Determinants

Wisconsin, like other states, is mandated by the 1972 federal Clean Water Act (CWA) to assess

the quality of its surface water and disclose the findings to the public. The primary objective of

the CWA was to render all US waters suitable for fishing and swimming by 1983. While this

standard was not met within the desired timeframe, the CWA remains committed to advancing

toward this goal. One of the measures utilized to gauge water quality is water clarity, assessed

using either a Secchi disk or a transparency tube, both of which measure the vertical distance

through the water that a black and white disk can be observed. Greater clarity indicates clearer

water. Water clarity has served as a proxy for water quality due to its sensitivity to both

atmospheric and terrestrial processes (Topp et al., 2021). Pollution control efforts during the

1970s and 80s notably reduced point source pollution, resulting in significant improvements in

water clarity. Topp et al. (2021) observed, in their study encompassing over 14,000 lakes, that

water clarity has increased on average by 0.52 cm annually since 1984.

Despite notable improvements in water clarity attributable to reductions in point source

pollutants, managing non-point source pollutants remains challenging. Runoff from urban

streets, parking lots, and agricultural fields continues to transport pollutants to lakes and streams.

Through regression analysis, Zampella et al. (2007) determined that models describing the

relationship between watershed disturbance and water quality must incorporate both urban land

cover and upland agricultural land use. Conversely, planted riparian buffer zones have been

identified as mitigators of the impacts of upland land use/land cover (LULC).

3.2.2. Buffer Zones

In a study of nine riparian buffer zone schemes, Parkyn et al. (2003) found that visual water

clarity was significantly better in buffer zones compared to upstream unbuffered areas. The

buffer zones in this study varied in width from 3.5 meters to 75 meters (11.5 – 246 feet). In

addition to employing the hedonic model, this study assesses whether shoreland disturbance

significantly influences water clarity after accounting for LULC within buffer zones surrounding

the study lakes. Another study in southern Ontario found that the landscape characteristics of a

100-meter (328 foot) buffer zone had only slightly less influence on the surface water quality

than the landscape characteristics of the entire water shed in three different study areas (Sliva &

Williams, 2001). In addition, a study in China used 6 buffer zones between 500m and 1800m

(1640ft to 5906ft) to determine which LULC classes effected water quality. The study found that

the amount of forested land in all 6 buffer zones had the greatest effect on water quality and that

LULC was better able to explain variation in water quality in the first two buffer zones (500m,

800m) (1640ft, 2625ft) than in the remaining four zones (Huang et al., 2020). This study chooses

to use two buffer zones, one similar to the Ontario study, and one similar to the first two buffer

zones of the China study. This analysis provides further insights into the potential effects of

shoreland preservation and restoration, aiding in evaluating the trade-offs between shoreland

management practices and water quality benefits.

3.3. Data

3.3.1. Land Use/Landcover

The Land Use/Land Cover (LULC) data was obtained from the Sentinel-2 10M Land Use/Land

Cover downloader (Esri, 2023). This data is generated annually using Impact Observatory’s deep

learning AI land classification model (Karra et al., 2021). Specifically, the LULC data for the

year 2022 covering areas 15T and 16T, which encompass the state of Wisconsin, was

downloaded, and subsequently merged in ArcGIS Pro. The land uses are categorized into 9

classes, as outlined in Table 3.1.

To ascertain the proportion of each land use class surrounding the study lake, buffers were

created at two distances from the shoreline: 0-500 feet (Buffer 1) and 500-2500 feet (Buffer 3).

These buffers were established to analyze the distribution of land use classes in the immediate

vicinity of the study lake.

3.3.2. Water Clarity

Water clarity serves as both an indicator of water quality for aquatic species and possesses

aesthetic value that contributes to the desirability and value of properties with lake frontage.

According to the Wisconsin Department of Natural Resources (WDNR, 2021), water clarity

impacts the depth at which aquatic plants can establish roots, thereby influencing the foraging

capabilities of fish, loons, and other wildlife. Additionally, the WDNR (2021) highlights the

significance of water clarity aesthetically, as it can influence both recreational activities on the

lake and the property values of surrounding areas.

Table 3.1 Land Use Classes and Descriptions

Class Number

Class Name

Description

Water

Area where water is predominantly present

throughout the year

Trees

Any significant clustering of tall dense vegetation

of 15m or more

Flooded Vegetation

Any type of vegetation with intermixing of water

throughout a majority of the year

Crops

Human planted crops

Build Area

Human build structures and transportation networks

Bare Ground

Rock or soil with very sparse to no vegetation

Snow/Ice

Large homogenous areas of permanent snow or ice

Clouds

No landcover information available

Rangeland

Areas of homogenous grasses with little or no taller

vegetation

Water clarity was chosen as an indicator of water quality in this study because it is readily

available for all the study lakes, and because it is often associated with concentrations of other

pollutants (e.g., phosphorus) via unfiltered runoff of organic and inorganic matter. In addition,

evidence suggests that water clarity is the water quality attribute most relevant to lakeshore

homeowners (Moore et al., 2020). Since the first chapter of this study concentrated on how

shoreland disturbances affected property value, the use of water clarity in this analysis

establishes a clear link between the two studies. However, water clarity is not the only indicator

of water quality, and very clear water can also be an indicator of an unhealthy lake environment,

such as the invasion of an invasive species as in the case of zebra mussels. Changes in water

temperature and chemical makeup also are indicators of an aquatic system that is out of balance.

Water clarity is typically assessed using a Secchi Disk, as described by the University of

Wisconsin Stevens Point (UWSP, n.d.). This disk, used in Wisconsin, comprises an 8-inch

diameter disk divided into 4 quadrants alternating between white and black (Figure 8). Attached

to a graduated rope and equipped with a bottom weight to maintain horizontal orientation, the

disk is slowly lowered into the water. The process involves marking the point where the disk is

no longer visible by affixing a clothespin on the rope at the water's surface. Subsequently, the

disk is lowered another 2 feet, gradually raised until visible, and marked with a second

clothespin. The Secchi depth recorded is the average between the two markings (WDNR, 2021).

Figure 8 Measuring water clarity with a secchi disk, Image from The Many Faces of Water

(Mierzynski, 2017)

Water clarity data for each lake is collected by volunteers multiple times during the summer

months and is accessible on the WDNR website (WDNR

, n.d.). The data utilized in this study

analysis (SecchiDepth) represents the average of all summer-time measurements spanning from

2010 to 2020.

3.3.3. Habitat Assessment

The first chapter of this study provides an overview of the Wisconsin Department of Natural

Resources' (WDNR) shoreland habitat assessment. In summary, this assessment utilizes data

from the riparian, bank, and littoral zones to compute a disturbance index for each property.

Properties with higher index values are indicative of greater levels of shoreland disturbance

compared to those with lower values (WDNR, 2023).

This chapter of the study focuses on a lake-level disturbance index (MMI_L) constructed using

shoreland information from all properties surrounding the lake, amalgamating this data to

generate a single index representative of the lake's overall disturbance level (WDNR, 2023).

3.4. Methodology

Regression models were employed to assess whether shoreland disturbances affect water quality,

while controlling for lake characteristics and nearby land use. Five model specifications were

considered. All five followed the general regression equation format represented in equation 6

with the measurement of Secchi depth representing water quality used as the dependent variable.

= β

+ β

+ e Equation 6

Where: SD = Secchi Depth

β = coefficients

X = LULC percents and Lake type

e = error term

In the first model (Model A), Secchi depth (SecchiDepth) was regressed on the percentage of

each land use class in Buffer 1 to examine the relationship between land use classes and water

clarity. Lake depth and lake area were included as independent variables, following the approach

of some hedonic model specifications. Additionally, a factor variable was introduced to

distinguish between three different lake types: DRAINAGE, SEEPAGE, and SPRING. As

defined by the WDNR

(n.d.), a seepage lake lacks surface inflow or outflow, a drainage lake

experiences surface water flow into and/or out of the lake, and a spring lake, or headwater lake,

receives water from subsurface sources and has a watershed draining less than four square miles

into it. Given that a drainage lake receives surface water inflow, it is expected to exhibit lower

water clarity due to upstream runoff.

The second model (Model B) mirrored Model A, except that it utilized the percentage of each

land use class in Buffer 3. A third model (Model C) regressed SecchiDepth against the lake's

average MMI_L to explore the relationship between variations in shoreland disturbances and

water clarity. Finally, Models D and E were developed by incorporating the MMI_L variable

into the variables used in Models A and B, respectively.

Table 3.2 Regression Model Specifications for Lake Water Clarity Analysis (Dependent

Variable: Secchi Depth)

Model

Secchi

Buffer LULC

MMI_L

Lake Variables

0-500

500-2500

0-500

500-2500

3.5. Results

This chapter presents the findings from the five regression models (Models A-E), all utilizing the

depth at which the Secchi disk remains visible below the water's surface (SecchiDepth) as the

dependent variable. Results from the Secchi depth regressions are detailed in Table 3.3, while

overall model statistics are summarized in Table 3.4.

Model A incorporated land use characteristics within a 500-foot buffer zone surrounding each

lake, along with lake area, lake depth, and indicators for lake type (e.g., seepage, spring). The r

value for Model A was 0.4964, with only the lake type and lake depth variables emerging as

significant predictors of lake clarity. Both spring lakes and seepage lakes were associated with

greater Secchi depth measurements relative to drainage lakes, with statistical significance at

p<0.01. Moreover, lake depth exhibited a positive correlation with water clarity, also significant

at p<0.01. Surprisingly, land use variables within the buffer zone were not found to be

significantly related to water clarity, contrary to expectations from previous studies (Zampella et

al., 2007; Parkyn et al., 2003; Garrison et al., 2000).

Model B utilized the same variables as Model A but based land use variables on a buffer zone of

500 – 2500 feet. While the r

and adjusted r

values increased slightly from 0.4964 and 0.4604 to

0.5041 and 0.4687, respectively, the only significant variables remained the lake characteristics.

Thus, this model did not identify a relationship between land use/land cover (LULC) variables

and water quality.

Model C was a simple linear regression between lake disturbance indices, developed by the

WDNR, and Secchi depth measurement. The r

value was 0.0139, indicating minimal

explanation of variation in lake water clarity. Although Figure 9 depicts a positive correlation

between Secchi depth and the level of shoreland disturbance around the lake, this relationship

lacked statistical significance, with a p-value of 0.2082.

In Model D, the lake disturbance index variable (MMI_L) was added to Model A. This addition

increased the r

value from 0.4964 to 0.5471, a rise of approximately 5.1%. The lake type and

depth variables maintained their positive relationship with lake clarity at a significance level of

p<0.01, with notable changes observed in the LULC variables. Specifically, the percentage of

land covered by tree cover and built area showed a positive and statistically significant

relationship with Secchi depth relative to rangeland area. Although the correlation between tree

cover and water clarity aligned with expectations, the unexpected positive relationship between

built area and water clarity requires further investigation.

Finally, Model E incorporated the MMI_L variable into Model B. Similar to Model D, the

addition of MMI_L increased the r

value. This model also demonstrated changes in the LULC

variables, with tree cover and built area being positively and significantly related to water clarity

at p<0.1. However, the disturbance index in both models D and E was not significant, suggesting

its inclusion did not enhance the model's ability to explain variation in water clarity.

Table 3.3 Regression Model Results (Dependent Variable: Secchi Depth)

LULC Class No.

Model A

Buffer 1

Model B

Buffer 3

Model C

Model D

Buffer 1

Model E

Buffer 3

Dependent

Variable

Secchi

Depth

Secchi

Depth

Secchi

Depth

Secchi

Depth

Secchi

Depth

MMI_L

0.0065

(0.2082)

-0.0015

(0.7473)

-0.0003

(0.9463)

Lake Type

SEEPAGE

4.1943***

(0.0000)

4.4460***

(0.0000)

5.1108***

(0.0000)

5.4380***

(0.0000)

Lake Type

SPRING

2.6180***

(0.0017)

2.6955***

(0.0013)

2.3855***

(0.0092)

2.6697***

(0.0036)

Lake Depth

0.1522***

(0.0000)

0.1593***

(0.0000)

0.1527***

(0.0000)

0.1496***

(0.0000)

Lake Area

-0.0007

(0.1674)

-0.0006

(0.1948)

-0.0005

(0.3238)

-0.0004

(0.4366)

-0.0238

(0.7486)

0.0276

(0.6179)

-0.0386

(0.6299)

0.0560

(0.3853)

0.0449

(0.2433)

0.0611

(0.2210)

0.0738*

(0.0548)

0.1033*

(0.0606)

-0.0309

(0.7051)

-0.1298

(0.1298)

0.0860

(0.3309)

0.0186

(0.9275)

-0.0389

(0.4947)

0.0164

(0.7464)

-0.0170

(0.7722)

0.0557

(0.3161)

0.0453

(0.2780)

0.0606

0.3056 ()

0.0745*

(0.0915)

0.1109*

(0.0928)

-3.9529

(0.5178)

-0.3200

(0.9689)

13.4740

(0.4605)

1.3658

(0.8636)

Constant

-0.0493

(0.9893)

-2.0135

(0.6646)

8.6602***

(0.0000)

-2.7803

(0.4466)

-6.1124

(0.2252)

Observations

151

116

114

0.4964

0.5041

0.0139

0.5633

0.5471

Adjusted r

0.4604

0.4687

0.0052

0.5162

0.4983

Note: Robust standard errors reported in parentheses.

*p<0.1; **p<0.05; ***p<0.01

Table 3.4 Overall Model Statistics for Lake Water Clarity Analysis

Statistic

Model A

Model B

Model C

Model D

Model E

Residual Min

-8.4663

-8.0323

-6.4706

-8.1519

-7.9951

Residual 1Q

-2.0847

-1.8789

-3.6508

-1.9495

-1.7434

Residual Median

-0.0126

-0.3231

-0.9313

-0.1601

-0.2434

Residual 3Q

1.7358

1.6264

2.9341

1.5536

1.2433

Residual Max

13.2645

12.5957

11.1883

7.2229

7.2724

Residual

Standard Error

3.3949

3.3688

4.4906

3.1589

3.2168

140

114

102

F-Stat

13.7992***

14.2318***

1.6020

1109588***

11.2013***

p-value

< 2.2e-16

0.2082

4.233e-14

2.38e-13

Figure 9 Graph of the relationship between the Lakeshore Disturbance Matrix from the Wisconsin

Department of Natural Resources versus the measurement of how deep a secchi disk can be

observed below the surface on 117 Wisconsin lakes.

3.6. Discussion and Conclusion

The study also analyzed if the lakes overall shoreland habitat disturbance value, as assessed by

the WDNR, was a predictor of water clarity, after controlling for the effect of surrounding land

use. Water quality and clarity have been found to affect property values in previous studies (e.g.,

Gibbs et al., 2002; Young, 1984; Moore et al., 2020). In addition, other studies have found that

there is a relationship between LULC and water quality and/or clarity (e.g., Zampella et al.,

2007; Parkyn et al., 2003). The results of this study were not significant for either the LULC or

the MMI_L variables, but additional modeling is warranted.

The study does have several limitations, first is that the study sample size was small relative to

many other HPM analysis. The WDNR studies additional lakes each year, therefore the available

y = 2.1363x + 76.271

R² = 0.0139

100

150

200

250

300

350

400

0 5 10 15 20 25

Lake Shore Disturbance Matrix

Secchi Depth in feet

Lake Disturbance Matrix vs. Secchi Depth

MMI_L

Trend Line

data will continue to grow in the future. But, even with additional data availability from the

WDNR, matching sales data from the MLS is still a limiting factor. Turnover, or sale of these

lakeshore properties is often limited due to the frequency of the properties being passed from

generation to generation without ever becoming for sale in the open market. The study was

further limited by the quality of data available from the MLS, leading to potential omitted

variable bias. MLS data is inputted by real estate agents or their assistants and at times it is

incomplete, or there are errors Many hedonic models have incorporated the age of the property in

the model (Netusil, 2005; Gibbs et al., 2002; Mooney and Eisgruber, 2001), but due to the many

sales records that were missing this information, this variable was not used. Probably more

important that property age would be a variable for property condition, unfortunately property

condition is not an available attribute in MLS data. Condition is important because even very old

properties, that have been extensively remodeled, can be in next to new condition. This study

should be expanded as more information becomes available from both the WDNR and the MLS.

3.7. Future Research

As with the previous chapter, these models should be updated as more data becomes available

from the WDNR. Future research could separate the artificial erosion control disturbance from

the remaining disturbances to evaluate if, as might be expected, they have opposing effects on

water clarity. Additionally, future research could evaluate a broader set of water quality

parameters, such as temperature and phosphorus levels. These additional models may

demonstrate a link between shoreland disturbances and water quality that could inform shoreland

management decisions.

4. Conclusion

Compared to some previous hedonic analyses in Wisconsin, this study examines the effects of

shoreland habitat disturbances (the environmental variables) on lakefront housing prices. The

results reveal that certain shoreland disturbances negatively impact property values, while others

are associated with an increase in property value. These mixed findings are expected, as many of

these disturbances offer improved recreational and aesthetic benefits to property owners, despite

their negative effects on shoreland habitat and water quality. It is crucial to acknowledge that

lower water quality can diminish property values.

Shoreland development, in the form of erosion control, beaches, and littoral-zone structures,

provide private benefits to lakeshore property owners, but there is an externality associated with

these features that negatively affects the public good through reduction in overall lake water

quality and aquatic habitat conditions. These private benefits total $36,720 for the average

property in the study sample. Multiplying this by the number of lakeshore properties in

Wisconsin, I estimate $43.5 billion (1.2M x $36,719.82) in benefits to homeowners from

shoreland development features. This estimate should be viewed cautiously as it assumes the

sample properties and lakes are representative of those across the state; nevertheless, it highlights

the considerable value of shoreland development and, conversely, the considerable loss

associated with converting shoreland back to a more natural condition.

Whether it makes economic sense to restore some shoreland areas to a more natural condition

depends on the benefits associated with improved water quality and habitat. Although the second

analysis in this study does not find a statistically significant correlation between water clarity and

shoreland disturbances, there is ample evidence to suggest that natural shoreland reduces

pollutant loading and provides critical habitat for fish populations (Munsch et al., 2017). From a

policy perspective, it is necessary to balance the benefit and costs of shoreland development.

To mitigate the impact of shoreland disturbance on the habitat of fish and other aquatic species,

it is imperative for policymakers and regulators to educate the public about conservation efforts

that enhance both water quality and shoreland property values (Munsch et al., 2017). These

actions are particularly viable when they align with both societal goals, such as erosion control

and recreational access, and ecological goals, such as water quality and habitat management.

However, it's essential to recognize that education outreach may have limited impact on

conservation efforts, as many shoreland disturbance features offer substantial non-market

benefits to homeowners.

It is also possible to incentivize homeowners to restore shoreland using subsidies. The WDNR is

currently spending $1.53 million on such subsidies, along with a social media project to educate

property owners on the positive outcomes of a more natural shoreland area. But results from this

study suggest that benefits of shoreland development are substantially greater than the available

subsidies. Estimated WTP from the hedonic analysis could be used to develop more effective set

of subsidies to help property owners transform their shorelines from disturbed, to a more natural

and environmentally friendly shoreland with native plantings.

To establish an efficient subsidy program, it will be necessary for the WDNR to better quantify

the public benefits of shoreland restoration. The WDNR can show the value of the public good

by presenting evidence that more natural shorelands will improve water quality and promote a

more diverse aquatic habitat, which will promote greater fishing opportunities. These changes

will lead to increased recreational, aesthetic, and non-use benefits. By quantifying and

monetizing these benefits, the WDNR can compare the benefits and costs of shoreland

restoration. A study in Iowa, for instance, found that the recreational benefits related to water

quality on inland lakes were greater than the cost of lake restoration (Otto et al., 2007). Likewise,

an analysis of benefits of shoreland restoration in Wisconsin, coupled with the WTP values

estimated from this analysis, can be used to evaluate where the benefits of shoreland restoration

are likely to exceed costs to the homeowner.

In conclusion, Wisconsin boasts approximately 15,000 lakes and 1.2 million lakeshore

properties. This hedonic analysis investigates the effects of various shoreland disturbances on

shoreline housing prices. The study demonstrates that shoreland disturbances are linked to higher

property values but also lower lake water quality, as evidenced by other studies showing a

reduction in property values. The findings underscore the importance of evaluating trade-offs

between homeowner preferences for shoreland features (e.g., erosion control, boathouses) and

near-shore water quality and habitat conditions when considering shoreland management

practices.

Bibliography

ArcGIS Pro (2023) https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-statistics/what-

is-a-z-score-what-is-a-p-value.htm, last access 29 October 2023.

ArcGIS Pro (n.d.) https://pro.arcgis.com/en/pro-app/3.1/tool-reference/spatial-statistics/h-how-

hot-spot-analysis-getis-ord-gi-spatial-stati.htm, last accessed 2 April 2024.

Box, G. E., & Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical

Society Series B: Statistical Methodology, 26(2), 211-243.

Champ, P. A., Boyle, K. J., & Brown, T. C. (Thomas C. (2003). A primer on nonmarket

valuation / edited by Patricia A. Champ, Kevin J. Boyle, and Thomas C. Brown. Kluwer

Academic Publishers.

Clapper, J., & Caudill, S. B. (2014). Water quality and cottage prices in Ontario. Applied

Economics, 46(10), 1122–1126. https://doi.org/10.1080/00036846.2013.851778

Cordell, H. K., & Bergstrom, J. C. (1993). Comparison of recreation use values among

alternative reservoir water level management scenarios. Water resources research, 29(2), 247-

258.

David, E. L. (1968). Lakeshore property values: a guide to public investment in recreation.

Water Resources Research, 4(4), 697-707.

Dziuk, H., & Heiskary, S. (2003). Local economic impact of healthy lakes. LakeLine, 23(3), 21-

23.

Embke, H. S., Douglas Beard Jr, T., Lynch, A. J., & Vander Zanden, M. J. (2020). Fishing for

food: quantifying recreational fisheries harvest in Wisconsin lakes. Fisheries, 45(12), 647-655.

ESRI (2023) Sentinel-2 10m Land Use/Land Cover Time Series

https://livingatlas.arcgis.com/landcoverexplorer/#mapCenter=-

3.286%2C31.34%2C3&mode=step&timeExtent=2017%2C2021&year=2022&downloadMode=t

rue last accessed 26 January 2024

ESRI (n.d.) ArcGIS Pro https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview last

accessed 31 January 2024

FRED (n.d.) Federal Reserve Economic Data https://fred.stlouisfed.org/ last accessed 31 January

2024

Garrison, P. J., & Wakeman, R. S. (2000). Use of paleolimnology to document the effect of lake

shoreland development on water quality. Journal of Paleolimnology, 24, 369-393.

Gibbs, J. P., Halstead, J. M., Boyle, K. J., & Huang, J. C. (2002). An hedonic analysis of the

effects of lake water clarity on New Hampshire lakefront properties. Agricultural and Resource

Economics Review, 31(1), 39-46.

Guignet, D., Heberling, M. T., Papenfus, M., & Griot, O. (2022). Property values, water quality,

and benefit transfer: A nationwide meta-analysis. Land economics, 98(2), 191-218.

Heinrich, J., & Kashian, R. (2010). Pricing the Homebuyer’s Proximity to Open Land. The

Journal of Applied Business and Economics, 11(1), 80–.

Horsch, E. J., & Lewis, D. J. (2009). The Effects of Aquatic Invasive Species on Property

Values: Evidence from a Quasi-Experiment. Land Economics, 85(3), 391–409.

https://doi.org/10.3368/le.85.3.391

Huang, W., Mao, J., Zhu, D., & Lin, C. (2020). Impacts of Land Use and Land Cover on Water

Quality at Multiple Buffer-Zone Scales in a Lakeside City. Water (Basel), 12(1), 47-.

https://doi.org/10.3390/w12010047

Jennings, M. J., Emmons, E. E., Hatzenbeler, G. R., Edwards, C., & Bozek, M. A. (2003). Is

Littoral Habitat Affected by Residential Development and Land Use in Watersheds of Wisconsin

Lakes? Lake and Reservoir Management, 19(3), 272–279.

https://doi.org/10.1080/07438140309354092

Jin, D., Hoagland, P., Au, D. K., & Qiu, J. (2015). Shoreline change, seawalls, and coastal

property values. Ocean & Coastal Management, 114, 185-193.

Karra, Kontgis, et al. (2021) “Global land use/land cover with Sentinel-2 and deep learning.”

IGARSS 2021-2021 IEEE International Geoscience and Remote Sensing Symposium. IEEE,

2021.

Kashian, R., & Winden, M. (2020). An Assessment of Lakefront Property Values Based on a

Decline in Water Levels: It’s Impact on Value and Taxes.

Klessig, L. L. (2001). Lakes and society: The contribution of lakes to sustainable societies. Lakes

& Reservoirs: Research & Management, 6(2), 95-101.

Lansford, N. H., & Jones, L. L. (1995). Recreational and Aesthetic Value of Water Using

Hedonic Price Analysis. Journal of Agricultural and Resource Economics, 20(2), 341–355.

https://doi.org/10.22004/ag.econ.30776

Loomis, J., & Feldman, M. (2003). Estimating the benefits of maintaining adequate lake levels to

homeowners using the hedonic property method. Water Resources Research, 39(9).

Mamun, S., Castillo-Castillo, A., Swedberg, K., Zhang, J., Boyle, K. J., Cardoso, D., Kling, C.

L., Nolte, C., Papenfus, M., Phaneuf, D., & Polasky, S. (2023). Valuing water quality in the

United States using a national dataset on property values. Proceedings of the National Academy

of Sciences - PNAS, 120(15), e2210417120–e2210417120.

https://doi.org/10.1073/pnas.2210417120

Macbeth, E. J. (1989). The relationship of shoreland zoning elements to the aesthetics of

developed lakeshores in Wisconsin (Doctoral dissertation).

Mierzynski, M. (2017). The Many Faces of Water,

http://www.themanyfacesofwater.eu/index.php/2017/05/19/measuring-water-clarity-with-secchi/

last accessed 1 May 2024

Mooney, S., & Eisgruber, L. (2001). The influence of riparian protection measures on residential

property values: The case of the Oregon Plan for Salmon and Watersheds. The Journal of Real

Estate Finance and Economics, 22(2–3), 273–286. https://doi.org/10.1023/A:1007899716050

Moore, M. R., Doubek, J. P., Xu, H., & Cardinale, B. J. (2020). Hedonic price estimates of lake

water quality: Valued attribute, instrumental variables, and ecological-economic

benefits. Ecological Economics, 176, 106692.

Munsch, S. H., Cordell, J. R., & Toft, J. D. (2017). Effects of shoreline armouring and overwater

structures on coastal and estuarine fish: opportunities for habitat improvement. Journal of

Applied Ecology, 54(5), 1373-1384.

Netusil, N. R. (2005). The effect of environmental zoning and amenities on property values:

Portland, Oregon. Land Economics, 81(2), 227-246.

Nicholls, S., & Crompton, J. L. (2018). The contribution of scenic views of, and proximity to,

lakes and reservoirs to property values. Lakes & Reservoirs: Research & Management, 23(1),

63-78.

Otto, D., Monchuk, D., Jintanakul, K., & Kling, C. (2007). The economic value of Iowa’s natural

resources. Iowa state University.

Papenfus, M. M., & Provencher, B. (2005, November). A hedonic analysis of environmental

zoning: Lake classification in Vilas County, Wisconsin. In Unpublished manuscript, University

of Wisconsin-Madison. Retrieved from http://www. aae. wisc. edu/seminars/papers/2005%

20Fall% 20papers/AppEcon/provencher (Vol. 12).

Parkyn, S. M., Davies‐Colley, R. J., Halliday, N. J., Costley, K. J., & Croker, G. F. (2003).

Planted riparian buffer zones in New Zealand: do they live up to expectations?. Restoration

ecology, 11(4), 436-447.

R (n.d.) The R Project for Statistical Computing https://www.r-project.org/ last accessed 31

January 2024

Rahman, R. (2021). Economic Value of Water Quality Improvements in Ontario (Doctoral

dissertation, University of Saskatchewan).

Rosen, S. (1974). Hedonic prices and implicit markets: product differentiation in pure

competition. Journal of political economy, 82(1), 34-55.

Sakia, R. M. (1992). The Box-Cox transformation technique: a review. Journal of the Royal

Statistical Society Series D: The Statistician, 41(2), 169-178.

Schindler, D. W., Armstrong, F. A. J., Holmgren, S. K., & Brunskill, G. J. (1971).

Eutrophication of Lake 227, Experimental Lakes Area, northwestern Ontario, by addition of

phosphate and nitrate. Journal of the Fisheries Board of Canada, 28(11), 1763-1782.

Schnaiberg, J., Riera, J., Turner, M. et al. Explaining Human Settlement Patterns in a

Recreational Lake District: Vilas County, Wisconsin, USA. Environmental Management 30, 24–

34 (2002). https://doi.org/10.1007/s00267-002-2450-z

Sliva, L., & Williams, D. D. (2001). Buffer zone versus whole catchment approaches to studying

land use impact on river water quality. Water research, 35(14), 3462-3472.

Taylor, L. O. (2003). The hedonic method. In A primer on nonmarket valuation (pp. 331-393).

Dordrecht: Springer Netherlands.

Tietenberg, T., & Lewis, L. (2019). Natural resource economics: The essentials. Routledge.

Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit

region. Economic geography, 46(sup1), 234-240.

Topp, S. N., Pavelsky, T. M., Stanley, E. H., Yang, X., Griffin, C. G., & Ross, M. R. (2021).

Multi-decadal improvement in US lake water clarity. Environmental Research Letters, 16(5),

055025.

UWSP (n.d.) The Secchi Disk and Our Eyes - Working Together To Measure Clarity of Our

Lakes chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://www3.uwsp.edu/cnr-

ap/UWEXLakes/Documents/resources/newsletter/vol31-

vol35/FurtherReading/Secchi%20Article%20-%20Full.pdf last accessed 12 March 2024.

Van Hengstum, P. J., Reinhardt, E. G., Boyce, J. I., & Clark, C. (2007). Changing sedimentation

patterns due to historical land-use change in Frenchman’s Bay, Pickering, Canada: evidence

from high-resolution textural analysis. Journal of Paleolimnology, 37, 603-618.

Walsh, P. (2009). Hedonic property value modeling of water quality, lake proximity, and spatial

dependence in central Florida. University of Central Florida.

WDNR (2023) Shoreland Habitat Disturbance in Wisconsin

https://www.arcgis.com/home/item.html?id=6b5e408d2ecf4b45b9958c9bfab0e607 last accessed

19 October 2023.

WDNR (2021) Secchi Disk Monitoring Procedure

https://dnr.wisconsin.gov/topic/SurfaceWater/Monitoring.html last accessed 25 January 2024

WDNR (2020) Lake Shoreland & Shallows Habitat Monitoring Field Protocol

https://dnr.wisconsin.gov/topic/SurfaceWater/Monitoring.html last accessed 23 September 2023

WDNR (2017) Chapter NR 115 https://docs.legis.wisconsin.gov/code/admin_code/nr/100/115

last accessed 19 September 2023

WDNR (2009) Wisconsin Lakes https://dnr.wi.gov/lakes/lakebook/wilakes2009bma.pdf last

accessed 19 September 2023

WDNR

(n.d.) Shoreland Habitat Disturbance Viewer

(https://www.arcgis.com/apps/dashboards/e60c681037ff491389dd9199d6427d73 last accessed

29 January 2024

WDNR

(n.d.) https://data-wi-dnr.opendata.arcgis.com/search?collection=Dataset last accessed

31 January 2024

WDNR

(n.d.) Find A Lake https://apps.dnr.wi.gov/lakes/lakepages/Default.aspx last accessed

31 January 2024

WDNR

(n.d.) Surface Water Integrated Monitoring System (SWIMS) Database

https://dnr.wisconsin.gov/topic/SurfaceWater/SWIMS last accessed 31 January 2024

WDNR

(n.d.) Wisconsin’s Riverine and Lake Natural Communities

https://dnr.wisconsin.gov/topic/Rivers/NaturalCommunities.html#:~:text=If%20there%20is%20s

urface%20water,classified%20as%20a%20drainage%20lake. Last accessed 2 February 2024

WRA (2024) Wisconsin Realtors Association https://www.wra.org/ last accessed 31 January

2024

Wisconsin State Cartographers Office (n.d.). Statewide Parcel Map Initiative.

https://www.sco.wisc.edu/parcels/data/ last accessed 31 January 2024.

Wolf, D., Klaiber, H. A., & Gopalakrishnan, S. (2022). Beyond marginal: Estimating the demand

for water quality. Resource and Energy Economics, 68, 101299.

Yanggen, D. A., & Kusler, J. A. (1968). Natural resource protection through shoreland

regulation: Wisconsin. Land Economics, 44(1), 73-86.

Young, C. E. (1984). PERCEIVED WATER QUALITY AND THE VALUE OF SEASONAL

HOMES 1. JAWRA Journal of the American Water Resources Association, 20(2), 163-166.

Zampella, R. A., Procopio, N. A., Lathrop, R. G., & Dow, C. L. (2007). Relationship of Land‐

Use/Land‐Cover Patterns and Surface‐Water Quality in The Mullica River Basin 1. JAWRA

Journal of the American Water Resources Association, 43(3), 594-604.

Appendix

Table A.1 Robustness Checks for Hedonic Property Model Results (Dependent Variable:

Natural Log of Adjusted Sales Price).

Variable

Model 3

Model 3a

Model 3b

Model3c

Model3d

LnGLA

0.4798***

(0.0000)

0.4827***

(0.0000)

0.5029***

(0.0000)

0.4575***

(0.0000)

0.5172***

(0.0000)

Garage Spaces

0.0657***

(0.000001)

0.0636***

(0.00001)

0.0531***

(0.0024)

0.0645***

(0.000001)

0.0649***

(0.000001)

Total Bath >2

0.1258**

(0.0245)

0.1373**

(0.0281)

0.1344*

(0.0843)

0.1540***

(0.0054)

0.0920

(0.1124)

-0.0002

(0.9958)

0.0027

(0.9452)

-0.0029

(0.9519)

0.0225

(0.5279)

-0.0181

(0.6228)

Ln of Lot Acres per GIS

0.0973***

(0.0007)

0.0945***

(0.0028)

0.1010**

(0.0118)

0.1150***

(0.00005)

0.0861***

(0.0033)

Ln of Acres/Shore Length

-0.0916**

(0.0157)

-0.1002**

(0.0142)

-0.0833

(0.1040)

-0.1204***

(0.0014)

-0.0721*

(0.0661)

Manicured Lawn Percent

-0.0002

(0.6354)

0.0001

(0.8755)

0.00005

(0.9436)

-0.0001

(0.9145)

-0.0004

(0.4045)

Canopy Percent

0.0002

(0.5734)

0.0003

(0.4842)

0.0002

(0.7820)

0.0002

(0.7100)

0.0003

(0.5570)

Beach Percent

0.0017*

(0.0314)

0.0019**

(0.0377)

0.0020*

(0.0961)

0.0006

(0.4483)

0.0020**

(0.0133)

Structures Lit Zone

0.0296**

(0.0128)

0.0233*

(0.0781)

0.0082

(0.6268)

0.0235**

(0.0441)

0.0339***

(0.0056)

Structures Rip Zone

-0.0052

(0.4819)

-0.0075

(0.3628)

-0.0085

(0.4178)

-0.0072

(0.3142)

-0.0055

(0.4632)

Erosion Control Percent

0.0012***

(0.0018)

0.0009**

(0.0350)

0.0011**

(0.0311)

0.0011***

(0.0028)

-0.1216***

(0.0092)

Erosion Present

-0.1138**

(0.0120)

-0.1127**

(0.0206)

-0.1310**

(0.0235)

-0.0551

(0.2253)

-0.1216***

(0.0092)

Bypass_Pres

-0.00051

(0.6063)

-0.0063

(0.5855)

-0.0090

(0.5255)

-0.0053

(0.5829)

-0.0008

(0.9391)

Vegetation Present

0.0210

(0.5020)

0.0418

(0.2339)

0.0356

(0.4244)

0.0282

(0.3593)

0.0093

(0.7756)

Vegetation Removed

-0.0084

(0.8373)

0.0060

(0.8951)

-0.0156

(0.7891)

-0.0148

(0.7135)

-0.0095

(0.8223)

Constant

8.8934***

(0.0000)

8.7032***

(0.0000)

8.4864***

(0.0000)

8.9000***

(0.0000)

8.6952***

(0.0000)

Year fixed effect

Yes

Quarter fixed effect

Yes

Lake fixed effect

Yes

School District fixed effect

Yes

Observations

847

681

463

847

802

0.6270

0.6306

0.6377

0.6445

0.6157

Adjusted r

0.5769

0.5735

0.5607

0.5946

0.5608

Note: Robust standard errors reported in parentheses.

*p<0.1; **p<0.05; ***p<0.01