The Influence of Riparian Setbacks On Private Property Values: Hedonic Price Analysis of Riparian Properties in Jackson County, Oregon

Transcription

1 The Influence of Riparian Setbacks On Private Property Values: Hedonic Price Analysis of Riparian Properties in Jackson County, Oregon Kathryne Maurer Sabrinna Soldavini Under the supervision of Professor Bruce Blonigen Presented to the Department of Economics, University of Oregon, in partial fulfillment of requirements for honors in Economics. June 2013 Acknowledgements: We thank Bruce Blonigen, Robert Parker, and Madeline Phillips for their guidance and assistance in making this paper possible 1

2 The Influence of Riparian Setbacks On Private Property Values: Hedonic Price Analysis of Riparian Properties in Jackson County, Oregon Abstract: Riparian buffers are designated strips of land that border streams and rivers designed to improve water quality. The implementation of riparian ordinances has been controversial as these mandates limit what property owners can do with their land. This paper draws from previous literature on riparian buffers, and employs two models to analyze any implications riparian ordinances have on residential property values using Jackson County, Oregon as a case study. In the city of Shady Cove, located within Jackson County, Oregon, there has been considerable push back against implementing mandatory riparian buffers. Using sales data from Jackson County ranging from two models are employed to determine the effect of riparian ordinances: a hedonic price valuation model properties from Shady Cove, Jacksonville, Phoenix, Talent, and unincorporated Jackson County, and a second hedonic price regression discontinuity model to compare properties in Shady Cove to properties in Unincorporated Jackson County immediately outside of the city boundary. Both models find the effect of mandatory riparian ordinances in Jackson County, Oregon have no significant effect on sales values. Approved: Professor Bruce Blonigen Date 2

3 I. Introduction 1.1 Motivation Tensions between local governments and owners of property along bodies of water have been rising since the United States government first began to regulate pollutants into waterways. In 1974, the Oregon Department of Land Conservation and Development adopted statewide planning goals and guidelines (Oregon Department of Land Conservation and Development, 1996). Goal 5, from the 19 goals, aims to protect natural resources and conserve open spaces, scenic areas, and historic sites. One way in which Oregon has decided to implement Goal 5 is through the use of riparian buffers. Riparian buffers are designated strips of land that border streams and rivers. Riparian buffers improve water quality by providing numerous benefits to the adjacent waterway: increased shade, which cools the water and creates a better habitat for fish, increased water during low flow periods in the summer, and flood and erosion control through protective vegetation ("Riparian Lands Tax Incentive"). By design, local riparian ordinances impose restrictions on the use of land on private properties; however, it is noteworthy that pre- existing structures in riparian buffers are allowed continued use (this is otherwise referred to as structures being grandfathered in ). These restrictions are implemented in order to improve water quality a public good. A natural concern is the extent to which such local ordinances may harm the private value of the affected properties. Yet little research has been done to analyze the effects of riparian buffers on affected property values. The literature that does exist has not yet come to an agreement on the effect riparian buffers have on property values. Mooney and Eisgruber (2001) find that treed riparian buffers reduce market values of streamside residential 3

4 properties in Lane County, Oregon. Bin, Landry and Meyer (2009) observe that riparian properties in Neuse County, North Carolina generally command higher premiums (though no evidence suggests the mandatory buffer rule had a significant effect on property values). Colby and Wishart (2002) observe that property owners in Tucson, Arizona also receive a premium when selling lands with riparian buffers. While the previous literature has begun the initial exploration, more research is necessary to fully understand the effects of riparian buffers on private properties. Shady Cove has faced particular pushback from citizens in response to potential implementation of riparian ordinances. Shady Cove is unique in the sense that Class I and Class II streams run through it, as well as the Rogue River. This project focuses specifically on residential properties in Jackson County along bodies of water. Our research investigates whether changes in riparian ordinances influence residential property values in Jackson County, Oregon. Furthermore, we aim to determine whether properties immediately outside of Shady Cove s city limits have been influenced by riparian ordinances. Our findings suggest riparian ordinances have no significant effect on residential sale values in Jackson County. Further, our findings indicate riparian ordinances have no significant impact on the sales values of residential properties in unincorporated Jackson County within a five mile radius of Shady Cove, Oregon. 1.2 The Evolution of Riparian Ordinances The first piece of legislation enacted to control water pollutants was the Federal Water Pollution Control Act in 1948 ("Digest of federal," 2013). The Federal Water Pollution Control 4

5 Act has been amended various times in order to authorize additional water quality programs. One of the more notable amendments, The Clean Water Act, passed in 1977 ("Laws, Regulations, Treaties (Watersheds)" March 06, 2012). The Clean Water Act sets the structure for regulating the amount of pollutants that can be discharged into waterways. Section 303(d) of the Act requires states to identify bodies of water that do not, or will not meet quality standards as well as any waters that will be affected by thermal discharges. Additionally, states must calculate a Total Maximum Daily Load (TMDL), a calculation of the total amount of a pollutant that can enter a body of water daily and still meet water quality standards Water quality standards are determined by states, territories, and tribes. Jackson County (located in Oregon) has many different government bodies legislating and mandating riparian ordinances. For instance, each city in the county creates and implements its own riparian ordinance, while properties in unincorporated areas of Jackson County follow county- wide rules. This paper specifically focuses on ordinances that apply to residential properties, rather than focusing on ordinances designed for agricultural properties. Although most ordinances have similar structures, each ordinance varies on the rules regarding the width of the buffers, and on the number of restrictions for development within the designated lot. For example, properties in unincorporated Jackson County must have riparian buffers of at least 75 feet from the Rogue River, and there must also be riparian buffers with a width of 50 feet from any Class I or Class II streams (or other fish- bearing bodies of water) ("Land Development Ordinance (LDO), " 2004). The initial riparian ordinance, established in 1984, mandated that the width of the buffers from Class I and Class II streams be set at a minimum of 25 feet. Jackson County has defines Class I streams as having one or more 5

6 of the following characteristics: direct source of water for domestic use, habitat for spawning, rearing or migration for large numbers of fish, or sufficient discharge to have a major effect on water quality of another stream. Class II streams, on the other hand, must be a habitat for spawning, rearing, or migration of a moderate number of fish, or have a sufficient discharge to have moderate influence on other Class I or Class II streams. Furthermore, each city ordinance was implemented at different points in time. For example, Phoenix implemented its riparian ordinance 1994, Talent passed its ordinance in 1998, and Jacksonville implemented its riparian ordinance in In contrast to these regulations, other cities in Jackson County, like Gold Point, have no riparian ordinance in place. Table 1, included below, gives an overview of the cities in Jackson County with riparian ordinances and their specific regulations. Table 1: Riparian Ordinances in Jackson County, Oregon Ordinance for Rogue River Ordinance for Class I Streams Ordinance for Class II Streams Unincorporated Jackson County 75ft 50ft 50ft Ashland None 50ft 30ft (Center of Stream) Medford None 50ft None Central Point None None 25ft Jacksonville None 50ft 25ft Phoenix None 50ft 25ft Talent None 50ft None 6

7 II. Literature Review Prior literature often uses hedonic pricing methods to investigate the effect that a newly created non- market environmental amenity has on property values. For instance, Kulshreshtha & Gillies (1993) calculate implicit willingness to pay valuations for the aesthetic amenity created by an improvement in riverfront view, such as an increase in vegetation along the riverbanks along the South Saskatchewan River in Saskatoon, Canada. They find the effect of improving a riverfront view is a higher willingness to pay with net annual value to the city of $1.2 million dollars. Similarly, Benson et al. (1998) and Landsord and Jones(1995) estimate the effect a scenic view has on residential property values in Bellingham, Washington and the Highland Lakes in the Colorado River basin in Texas, respectively. Results from both studies indicate consumers are willing to pay a premium for homes with a scenic view, with Lansford and Jones (1995) finding scenic properties command a price premium of 8 percent relative to non- scenic properties. The literature also includes studies conducted on the effect planting trees on a property might have on future property value (as is encouraged with the implementation of a riparian setback in Jackson County, Oregon). Payne (1973), and Anderson and Cordell (1988) study the effects of the presence of trees on a property, finding that the sales value increases as the number of trees on a property increase (up to a certain threshold number of trees) when compared to an identical residential property without trees. They estimate that, on average, planting trees contributes as much as 27 percent to appraised land value (Payne, 1973). However, Anderson and Cordell also find that beyond a threshold number of trees, sales value may actually begin decreasing as number of trees increases. These studies may signal that 7

8 implementing riparian setback can create a valuable amenity, such as a more scenic view, increasing sales value up to a certain level of setback, but that beyond the threshold setback limit the buffer may begin to create a disamenity (lack of access, blocked view, etc.) that offsets the initial benefit enough to cause property values to decline with the size of the tree buffer. Though much work has been done on willingness to pay for amenity creation resulting from riparian buffer, relatively little of the literature focuses analysis on the effects of riparian setbacks on property valuations. To date, we are aware of just three prior studies, Mooney and Eisgruber (2001), Bin et al. (2009), and Colby and Wishart (2002). All three studies use hedonic pricing methods to estimate the effect of riparian setbacks on property values with results that are mixed at best. Mooney and Eisgruber (2001) use a hedonic valuation model to estimate the effects of a policy which encouraged residential property owners to plant trees along a riparian corridor on property values. This policy encourages planting trees because trees provide shade and cooler temperatures for salmon traveling through the Mohawk watershed. Using data from residential properties within the Mohawk watershed of Western Oregon as a case study, their findings indicate planting a wider riparian buffer, as measured by tree width, led to a decrease in residential property value. They propose this is likely an indication of the diminished riverfront view that is created by planting more trees, thus decreasing the potential buyer s willingness to pay for riverfront property. Other relevant literature, Bin et al. (2009) and Colby and Wishart (2002) both find riparian properties command higher premiums relative to identical non- riparian property. By studying the value of properties along riparian corridors in Northeastern Tucson, Arizona, Colby 8

9 and Wishart (2002) find substantial premiums on private property particularly for property within the first half mile of a riparian corridor, with the benefits extending for private property up to 1 ½ miles from the riparian corridor. Their data show homeowners display a 6% premium on willingness to pay for property within 0.1 miles of a riparian corridor with even higher premiums of 10-27% for undeveloped land within 1.5 miles of a riparian corridor. Bin et al. (2009) also find an increasing willingness to pay for riparian properties by developing a spatial autoregressive hedonic evaluation model to estimate average effects to residential property values resulting from an immediate imposition (introduced in 1997) of a mandatory riparian along the Neuse River Basin in Neuse County, North Carolina. They use the technique of differences- in- differences, a tool which allows for measuring the effect of a treatment at a given period of time between two groups. Through analysis of riparian properties and non- riparian property values both before and after the mandatory buffer law was enacted, they determine that although riparian properties in general command a premium over non- riparian properties, the implementation of a mandatory buffer zone had no significant effect on riparian property values when compared to their control group. The results are attributable to the hypothesis that while a mandatory buffer creates restrictions on private land use (expected to lower property value), this cost may be offset by the creation of new environmental amenities such as improved water quality and aesthetics (which would be expected to raise riparian property values). Bin et al. (2009) also find an increasing willingness to pay for riparian properties through a spatial autoregressive hedonic estimation of the average effects to residential property values resulting from the 1997 immediate imposition of a mandatory riparian buffer in Neuse County, 9

10 North Carolina for properties along the Neuse River Basin. They use a differences- in- differences specification, a statistical technique which allows for measuring the effect of a treatment at a given period of time between two groups. Through analysis of riparian properties and non- riparian property values both before and after the mandatory buffer law was enacted, they determine that although riparian properties in general command a premium over non- riparian properties, the implementation of a mandatory buffer zone has no significant effect on riparian property values when compared to their control group. The results are attributable to the hypothesis that while a mandatory buffer creates restrictions on private land use (expected to lower property value), this cost may be offset by the creation of new environmental amenities such as improved water quality and aesthetics (which would be expected to raise riparian properties values). Due to the inconclusive nature of prior literature, more study in the field is required to identify the true average effect riparian mandates have on private property values. Our paper aims to uniquely contribute to the literature by employing techniques not previously used to analyze the effects of riparian ordinances. This is achieved through a hedonic evaluation of local sales records, as opposed to assessed values, spanning more than 20 years ( ), and a regression discontinuity model (a quasi- experimental design measuring the effects of a treatment between two otherwise identical groups based on a cutoff value) not yet used in evaluating effects of riparian mandates. Using city boundaries as the cutoff value for treatment (riparian ordinance) between otherwise identical properties with sales data along the same waterways should allow for isolation of the effect of the ordinance. It is also our aim to use our research to the inform policy makers in the geographic area of interest: Jackson County, 10

11 Oregon. We believe our research will lead to particularly relevant results for County Planners and policy makers seeking to determine whether the net benefits of changes to, or the implementation of new riparian ordinances in the future, outweigh the costs to residential property owners. The regression discontinuity model should prove especially relevant to residents of Shady Cove. Assuming all relevant controls are valid and omitted variable bias is not present in the model, it should be the case that properties immediately outside of Shady Cove along the same waterways should be directly comparable to those located immediately inside city limits thereby estimating the effect that instating riparian ordinances in Shady Cove would have on local property owners. III. Economic Theory The motivation behind this paper is to address owner s concerns regarding devaluation of property due to the implementation of a riparian ordinance. We expect that residential property owners will experience costs and benefits from implementing riparian buffers. However, it is not clear what the magnitude of these costs is, or how frequently property owners must face these costs. New amenities produced by a riparian buffer may also come with positive and negative benefits. Economic theory lends itself as an explanation for the various outcomes that could take place. I. Costs incurred as a one- time expenditure If property owners only have to pay a one- time fixed expenditure, then that fee will be a one- time sunk cost that will not affect the future property value. For instance, the cost of moving a garden back to prevent fertilizer from entering a water source will only occur once. 11

12 This suggests that even if there is a one- time fee, which may be cumbersome for property owners, that a riparian ordinance will not significantly affect property values. II. Costs as a recurring fee If property owners expect to pay a regular fee (such as maintenance cost) due to the riparian buffer, that cost will be taken into account as a decrease to the property value. An example of such a cost is the cost of removing noxious weeds from the setback every year. Consequently, the introduction of a riparian ordinance would result in a decrease in property sales values. III. Features that are a benefit to the property Features that enhance the property and environmental characteristics (increase in fish stock, increased water quality), will be positively valued in the marketplace. Both Bin et al. (2008), and Colby and Wishart (2002) find properties with riparian buffers command higher premiums when sold. This suggests that the added benefit of increased water quality (the ultimate goal of riparian buffers), or some other positively valued characteristic, is preferred by buyers. However, it is likely that positively valued characteristics in some regions might be devalued in other regions. For example, shade in Arizona (where sun is abundant) might be positively valued, while shade in Oregon (where sunlight is more limited) is negatively valued. IV. Features the hinder property value Features that diminish positively valued qualities of the property (for instance, a diminished view of the stream due to trees) may negatively affect the property value. Mooney 12

13 and Eisgruber (2001) find that a higher number of trees on a riparian buffer tended to lower the property value. This suggests that some characteristics introduced by riparian buffers are negatively valued by property owners. V. Smaller properties are more negatively affected Riparian ordinances implement the same regulations on to all properties, regardless of shape and size. Consequently, larger properties will have less of their property dedicated to the riparian buffer. Smaller properties, on the other hand, will have less ability to choose how to develop their property. In this case, the riparian ordinance would negatively impact sales values of smaller properties relative to larger properties. VI. Owners of properties with higher quality homes less will value positive features Owners of properties with higher quality homes are likely to have more wealth. Costs associated with implementing and up keeping the riparian buffer are not as burdensome for property owners with more wealth. Furthermore, property owners with more wealth may also have a strong preference for positive features, such as cleaner water and more fish, and thus are willing to pay more to receive these benefits. Consequently, owners of properties with higher quality homes will positively value the implementation of riparian buffers. VII. Properties with structures are negatively affected relative to vacant land Vacant properties have minimal (if any) established structures that need to be modified to fit a riparian ordinance. Consequently, vacant lands will face fewer costs to implement a riparian buffer than properties with existing structures. Furthermore, Colby and Wishart (2002) 13

14 find riparian ordinances generated a proportionally greater premium for underdeveloped lands. This suggests properties with structures will be negatively affected by a riparian ordinance relative to vacant properties. V. Methodology Through use of a hedonic pricing model, this paper will use data before and after changes in riparian setback ordinances in Jackson County, Oregon to identify the average effect of a riparian setback mandate on private property values. This choice of model comes from Rosen (1974) who develops the hedonic price method by extending Lancaster s theory of consumer demand. Hedonic evaluation models decompose the price of private property based on its characteristics. Market value is thus the combined sum of implicit prices potential buyers place on each of the characteristics of a property. Our models will use the following base function: (Si)= f ( Ci ) Where i is the ith sale observation, (Si) is the ith property s sale value, and S is a denoted as a function of the characteristics (C) that make up the property. Identifying the model as such will allow for the differentiation between effects of multiple variables which could impact the valuation of private property. Specifically, controlling for other influences on property values, we can examine how riparian regulation changes over the period of interest ( ) have affected property values. In order to fully understand the effects that riparian ordinances have had in Jackson County, we will introduce the technique of difference- in- differences analysis into our hedonic 14

15 regression model. By using data before and after the imposition of riparian setbacks in Jackson County we can measure the average effect of the setback requirements on riparian property values. This type of analysis requires the study of a treatment and a control group to distinguish the effect of the riparian setback changes on property values from the effect of other exogenous variables. For instance, changes in land use alterations may affect riparian property values over time. To control for this issue we include properties that were sold over time in locations that either do not, or did not at time of sale have to maintain a riparian buffer. The treatment group will be the properties which were influenced by a change in policy over our period of interest. Model One: To test whether the implementation of different riparian ordinances over time and location in Jackson County have had an effect on residential property values in Jackson County, we first illicit the use of a simple hedonic regression to run a series of regressions including all of the observations collected (omitting Shady Cove as the reference group). The hedonic method does not have a specified form associated with it. Therefore, through analysis of the data we chose to use a semi- logarithmic model to allow for interpretation of coefficients as percentages changes. Hence our first model is: ln(saleprice! ) = β! + β! RegTalent + β! RegPheonix + β! RegJVille + β! RegUninc84 + β! RegUninc A(h! ) + L(! c)+! T(y) + ε 15

16 Where A(h! ) represent attributes associated with the property (acreage, lnsqft, beds, baths, yearbuilt, vacantland, ln_assesed_land_value), L(c! ) represent locational controls (city, class of waterway the tax lot lies on), D(a! ) represent distances to local amenities (hospitals and schools),t(y! ) represents the year of sale controls (sold1980- sold2013) to control for exogenous changes in real estate markets and thus take into account inflation over time making sales values throughout the time period comparable, and α! represents the error term. A description of all variables is provided in section A.2 of the appendix. The variables of particular interest are RegJVille, RegPheonix, RegTalent, RegUninc84, and RegUninc2004 which each denote different riparian regulation controls which took place between 1984 and in 2004 in Jackson County. The coefficients on these variables make it possible to estimate the effects that riparian regulations had on residential property values. In this paper, the ordinances of interest are: Talent s riparian ordinance, which passed in 1998; Phoenix s ordinance, which passed in 1994; Jacksonville s ordinance, which passed in 2008; and both the 1984 and 2004 ordinances implemented in unincorporated Jackson County. The semi- logarithmic model allows us to interpret the coefficients on our variables of interest in percentage terms, meaning that the coefficients will describe the percent change in sale value as a result of the riparian regulation. From this model further restrictions are made to test for differences in differing lot sizes (less than 10 acres and less than 4 acres). Model Two: Base Hedonic Property Evaluation Model The second series of regressions restricts the data set to include only observations located along the Rogue River and Class I and Class II streams in Shady Cove, and to include 16

17 observations located in the unincorporated area immediately outside Shady Cove within a radius of 5 miles. This model introduces the idea of a regression discontinuity into the basic hedonic price evaluation method. Shady Cove does not currently have a riparian ordinance in place, while the homes immediately outside city limits fall under Jackson County jurisdiction. Therefore, by imposing the city limits of Shady Cove as the cutoff for treatment determining whether or not a property is under the influence of a riparian ordinance should allow for determining the true effect of the riparian ordinance. Therefore, our second model is: ln(saleprice) = β! + β! Unin84 + β! Uninc A(h! ) + L(c) + T(y! ) + I(y! ) + D(a! ) + B(d! ) + ε! Where again A(h! ) represents attributes associated with the property, L(c! ) are location controls, T(y! ) represents year sold control variables, and D(a! ) represent variables locating distance to local amenities. New to the discontinuity model are interaction variables represented by I(y! ), and dummy variables B(d! ), that indicate whether a property in unincorporated Jackson County falls within a 1.5, 3, or 5 mile radius of Shady Cove city limits. Interaction controls are introduced to control for and capture any trends in sales values or public goods provisions for properties in unincorporated Jackson County relative to properties in Shady Cove. The buffer distances have been included to measure any effects of moving further away from the cutoff value. Similar to the first model, the variables of interest are RegUninc84 and RegUninc2004, which indicate changes in riparian ordinances in Jackson County in 1984 and Again, a semi- logarithmic model is used to interpret the coefficients on these variables as the percentage change in property value, as caused by the first time riparian ordinances were 17

18 introduced in Jackson County in 1984, and the 25 foot increase in setback distance (from ) for properties along the Rogue River in 2004 respectively. Further, assuming that all relevant controls are valid and omitted variable bias is not present in the model, it should be the case that properties immediately outside of Shady Cove along the same waterways should be comparable to those located immediately inside city limits, thereby estimating the effect of riparian ordinances in Jackson County. VI. Data Our area of study is restricted to private properties in Jackson County, Oregon. Specifically, the relevant data includes only those properties close enough to a waterway to be affected by a change in riparian setback mandates. For the purposes of this case study, such properties are those located adjacent to Class I and II streams as well as the Rogue River in Jackson County, Oregon. Further, we restricted our sample to properties in the cities of Jacksonville, Phoenix, Shady Cove, Talent, and the unincorporated area immediately outside of Shady Cove within a three mile radius. The primary source of data used in our analysis was obtained from Jackson County Geographic Information Systems (JCGIS), and the Jackson County Assessment office who work in accordance to produce the Jackson County Online Property Database containing all sales transactions by account and tax lot, in addition to records of Real Market Assessed values (assessed land and improvement values), type of property, sale year, acreage, and owner for all tax lots within Jackson County. 18

19 The tax lot dataset from JCGIS includes over 91,000 observations with information on each tax lot including data of property classification, acreage, and geocoded locations. In order to identify only the properties affected by implementation of riparian setbacks the dataset was imported into ArcMAP 10.1 (a geographic information system created by ArcGIS which allows for the use of spatial analysis). County and city boundaries data layers, as well stream and river coverage datasets were also obtained from JCGIS and overlaid to create an overview map of tax lots by city and geocoded location. The geoprocessing tools were employed to spatially select all tax lots within Jackson County city limits that intersected waterways within a specified distance of the stream bank (75ft from the Rogue River, 50 feet from Class I streams, and 25 feet from Class II streams). Unincorporated properties on the Rogue River and Class I and Class II waterways within a three mile radius of Shady Cove were also selected for our sample. This was done through geoprocessing distance buffers of 1.5, 3, and 5 miles outside of Shady Cove. These buffers were used to differentiate properties further outside city limits, which would allow us to control for any differences that might occur as properties become located further from public benefits attributed to being in an incorporated area. After identifying all relevant tax lots in the study area located along waterways, geocoded feature layers of hospitals and schools in Jackson County were obtained from JCGIS and added to the maps. Using the geoprocessing near command in ArcMAP, distance variables (in feet) were created to calculate the distance to the nearest hospital and school in the county. This allowed for controls of distance to local amenities that one might expect to be important for potential residential property buyers. 19

20 All selected tax lots were then matched by account number, and sales transactions records (sale value, year of sale) as well as property characteristics (square feet, beds, and baths) were manually found through the Jackson County Property Data Online. These sales records were then cross- referenced with property transaction records from Zillow (an online real estate database) when possible. Matching sales by account allows for the possibility of multiple observations per tax lot; however, only the three most recent sales transactions were recorded. After dropping all accounts without sales records and restricting the sample to include sales data from , our sample included 794 sales observations along the waterways affected by riparian setbacks. While a large sample of close to 800 sales observations within the study area were obtained for analysis, there a few issues with the data that should be recognized before proceeding. Namely, the data used is not a random sample of sales records within Jackson County, and there almost certainly is measurement error present in both the Jackson County Property and the data set assembled by the authors of this paper both issues potentially leading to a systematic bias in the results. An ideal data set for this study would have included only a spatially randomized selection of residential sales transactions with Jackson County along all waterways affected by riparian ordinances. Unfortunately, no such data set exists and time and budget constraints prevented us from creating such a robust data set. As such, the properties included suffer from a problem referred to in geostatistics as spatial dependence, and characterized by Waldo R. Tobler s first law of geography that "Everything is related to everything else, but near things are more related than distant things" (Tobler 1970). Spatial dependence lends itself to the idea of 20

21 spatial autocorrelation based on the probable assumption that price of recent sales in an area likely influence the property value of similar, nearby properties leading to endogeneity in the model. Further, the ideal randomization would have included a random and sudden implementation of riparian ordinances in the county where only certain properties were forced to create and maintain riparian buffers on their property. Measuring sales values before and after such an event would allow for truly identifying the effect of the setback on sales value. Of course, no such natural experiment exists, and we believe our regression discontinuity model is as close to such a random prescription as possible. Additionally, the data likely suffers from multiple sources of measurement error. Data was obtained by Jackson County assessors over a period of 23 years. It is unlikely that over this time none of the assessors made errors in measuring acreage, square feet, or number of beds and bathrooms. Equally unlikely is the chance that data entry over this time period was 100 percent accurate. For example, the data included cannot take into account additions of improvements made to houses after the most recent sale that will certainly have an effect on property value. A third source of measurement error may have spurred from the authors themselves when creating the data set. Some properties selected from the GIS analysis are likely to not actually be affected by riparian ordinances. There are tax lots within the Jackson County that were grandfathered in when the ordinances first went into effect, and therefore do not actually have to maintain a riparian buffer. However, no data on what properties were grandfathered in was available, so all tax lots on applicable waterways in Jacksonville, Phoenix, Talent, and Unincorporated Jackson County are treated as being under the effect of a riparian ordinance. Implications of measurement error and erroneous selection as a result of an 21

22 unobserved grandfathering variable imply that our models estimates may have a systematic bias; measurement error can bias standard errors in the upward direction while selection bias can cause biased estimates in either direction. As a result, the ultimate effect of these biases is of an unknown direction and magnitude depending on the direction and magnitude of each source of error. VII. Results To determine whether or not riparian ordinances have had a significant impact on property values within Jackson County, Oregon, we run two models each with a series of restrictions and specifications. We start with the null hypothesis that riparian ordinances do not have a significant effect on residential property values. 7.1 Model One Results: Base Hedonic Property Evaluation Model The first regressions use sales values from residential properties located in Jacksonville, Phoenix, Shady Cove, Talent, and Unincorporated Jackson County located within 75 feet of the Rogue River, 50 feet of a Class I, or 25 feet of a Class II stream. (See appendix for maps of tax lots used in these regressions). Our model is as follows: (1) ln(saleprice! ) = β! + β! RegTalent + β! RegPheonix + β! RegJVille + β! RegUninc84 + β! RegUninc A(h! ) + L(! c)+ε! 22

23 Table 3 in the appendix displays the results from these regressions. The first column represents the results of the model ran without any imposed restrictions. An R- squared value of suggests that the model accounts for 49.5% of the variation in sales prices among riparian properties within the study area. Though this is not a particularly high explanatory value, it is in the range to be expected without restrictions placed upon house or lot size. The variables of interest (RegPheonix, RegJVille, Regtalent, RegUnin84, and RegUninc2004) are all slightly negative but insignificant with confidence intervals that include zero suggesting that none of the riparian regulations have had a significant effect on residential property values within the county. The other variables in the model appear with expected coefficients and significance. Lnsqft, number of baths, lnassessedland and vacantland are all significant and positive as expected, and the year of sales values capture the downward trend in home prices that we expected to see after the real estate bubble burst at the end of The coefficient on the vacantland variable of 3.28 is worth discussing here. This suggests that potential buyers are willing to pay more 300% more or 3.28 times more for vacant land with riparian buffers relative to similar riparian properties with existing structures on them. We hypothesize the intuition behind these results is that vacant properties will face lower or no costs of structure modification to abide by a riparian mandate. Our findings are in accordance with Colby and Wishart (2002) who find riparian ordinances generated a proportionally greater premium for underdeveloped lands. Unexpectedly, the unrestricted model finds an insignificant coefficient for the variable ln_acreage. In attempt to explain this unusual result we hypothesize that inclusion of a few very 23

24 large (acreage) properties in the unrestricted model may be outliers that skew the results. Further analysis proved this hypothesis to be true, and consequently a second model that drops all tax lots larger than four acres is included in the second column of Table 3. The results indicate that by dropping 163 observations of properties larger than four acres, the explanatory power of the model increases to.538. The coefficient on ln_acreage is now positive and significant as we would hope to see, assuming that it is reasonable to expect a higher willingness to pay for larger properties. The coefficient of on ln_acreage can be interpreted as indicating doubling the lot size, in acres, for properties smaller than four acres leads to a 16.7% premium in sales value. Additionally, the coefficient on the variable vacantland is now a more reasonable 2.38, implying a 238% increase in willingness to pay for vacant riparian property relative to already developed properties. Interestingly, the coefficient on the 1998 Talent riparian ordinance is now negative (-.376) and significant at the 10% level implying that the imposition of the mandatory riparian buffer led to a 37.6% decrease in willingness to pay for riparian property within the city. This result is not convincing however, and should be taken with caution as the number of sales observations in Talent included in this model is only 37 observations, too small of a sample to hold much weight. Coefficients on all other regulation variables remains insignificant however, leading us to retain the null hypothesis that riparian regulations have had no measurable effect on property values for within Jackson County. 24

25 7.2 Model Two: Regression Discontinuity Comparing Sales Values in Shady Cove to Unincorporated Jackson County within a Three Mile Radius The second model compared residential properties in unincorporated Jackson County that were immediately outside of Shady Cove to those residential properties within Shady Cove s city limits and is as follows: (2) ln(saleprice) = β! + β! Unin84 + β! Uninc A(h! ) + L(c) + T(y! ) + I(y! ) + D(a! ) + B(d! ) + ε! The final version of the second model limited properties to only those less than four acres in size and within three miles of Shady Cove s city limits (see Appendix regarding model specification). The results for this model, as shown in Table 4 in the appendix, indicate both riparian ordinances (1984 and 2004) did not significantly affect sales values. The final model also included interaction terms for properties sold both in unincorporated Jackson County and for each year a property was sold. These interaction terms (named u_yr(year)) capture any trends in sales values for properties in an the unincorporated Jackson County area relative to properties with Shady Cove city limits. Results indicate several of these interaction terms are significant, suggesting a negative trend in sales values within the last decade for properties located in unincorporated Jackson County. Although the unincorporated ordinances have no significant effects on all sales values, we wanted to check for any heterogeneous effects in the data. That is, we wanted to determine whether certain types of properties in unincorporated Jackson County might be affected 25

26 differently by the ordinances. Specifically, the properties of interest were larger properties, vacant properties, and properties with higher quality homes. We begin by analyzing the effects of riparian ordinances on large properties relative to small properties. An above average sized property is defined in here as a property with an acreage that is greater than the mean acreage (7.18) acres of the data set. Table 5 (Appendix) shows riparian ordinances do not significantly affect property values for above average sized properties relative to smaller properties. We then compared any effects riparian ordinances might have on vacant properties relative to properties with structures. Initial hypotheses were that the effect of limiting land use on vacant property may be greater than to Analysis indicates the sales values of vacant properties are not significantly affected by the riparian ordinances relative to properties with structures (See Table 6 in the Appendix). Finally, we wanted to determine the effects of the riparian ordinances on properties with higher quality homes. A higher quality home is defined as any home with square feet greater than the mean (from the data collected). Table 7 (Appendix) shows the 1984 riparian ordinance had no significant effect on sales values for higher quality homes relative to lower quality homes. However, the riparian ordinance change in 2004 positively affected property values with higher quality homes at the 10 percent significance level. Upon further inspection, the effect of the 2004 riparian ordinance on higher quality homes would be the sum of the interaction term reguninc2004largesqft and the ordinance variable itself: thus, the total effect would be , or a 41% increase in property values relative to lower quality homes. However, an F- test was used to determine whether this coefficient was statistically significant. 26

27 The results of the F- test are posted below: The coefficient lies just outside of the 10 percent significance level, implying the 2004 riparian ordinance does not significantly affect higher quality homes more relative to lower quality homes. VIII. Conclusion The goal of implementing riparian buffers is to improve environmental conditions and habitats for local ecosystems. Property owners face costs associated with the implementation of riparian buffers, as well as benefits from improved water quality and other positively valued features. The first model introduced in this paper, which uses properties located in Shady Cove, Jacksonville, Phoenix, Talent, and unincorporated Jackson County, suggests that none of the city ordinances significantly affected property values. The second model, which used a discontinuity model between properties within Shady Cove and properties immediately outside of Shady Cove in unincorporated Jackson County, also found that property values were not significantly affected by the 1984 or 2004 riparian ordinances. In looking for heterogeneous effects in our second model, we find that vacant properties, higher quality homes, and above average acreage are not significantly affected by riparian ordinances relative to properties with structures, lower quality homes, and smaller sized properties. 27

28 Although this project uses sales data a very good indicator of how much people value or disvalue riparian buffers, there still exist a few adjustments that could be made to the model. Specifically, future projects with the time and resources could include information pertaining 50 and 100 year flood plains to control for any effects that might make properties in Shady Cove different than properties immediately outside of the city limits. Furthermore, more data concerning the quality of the houses on properties might reveal a heterogeneous effect with respect to higher quality houses. VIIII. Appendix A.1 Model Specification The second model, which compares properties in unincorporated Jackson County to properties in Shady Cove, underwent many changes before a final specification was selected. The properties in unincorporated Jackson County had a larger range of size (in acres). There was concern that the few large properties were outliers in our data set. Initially, the model had no limitation on acreage. As shown in Table 8 (further down in the appendix), the R 2 came in at The model was then modified to limit the size of properties to less than 10 acres. Only 80 properties were dropped from the model, and the R 2 increased to Furthermore, the number of baths becomes statistically significant in sales values, which was an anticipated result. Finally, the model was modified to limit the size or properties to less than 4 acres. Another 80 properties were dropped from the model, and the R 2 increased to The benefit of using properties immediately outside of Shady Cove in unincorporated Jackson County is that they share many similarities to properties within Shady Cove, including 28

29 very similar distances to amenities. Sales data was collected for properties within 1.5 miles, 3 miles, and 5 miles of the Shady Cove city limit. The goal was to use properties in unincorporated Jackson County immediately next to the city boundary are otherwise identical to properties in Shady Cove, and yet collect enough sales data to run relevant regressions. Table 9 displays the results of the model using different boundary radii. In the following models, only properties less than 4 acres were included. The third column of Table 9 limited our model to include unincorporated properties within 1.5 miles of Shady Cove. This model produced a high R 2 of however, there were only 169 properties included in the model. To expand the model, we include properties within 3 miles of Shady Cove. The number of observations increases to 438, and maintained a relatively high R 2 of Finally, properties within 5 miles of Shady Cove were included in the model. The number of observations only increased by 53 properties, and the R 2 decreased to Regardless of the property size or location with respect to the city limit, the riparian ordinances of 1984 and 2004 did not significantly affect property values. 29

30 A.2 Variables Lists Description of Variables Variable ln_salevalue ln_acres ln_sqft ln_assessland beds baths yrblt dis_priv dis_pub dis_pub_elem dis_hospital aboveavgacreag e1 LargeSqft vacantland jacksonville phoenix shady_cove talent unincorporated within1_5mil within3mil within5mil onrogue sold1980- sold2013 u_yr(1980)- u_yr(2013) Definition Natural log of sale price (in US$) Natural log of total acreage Natural log of total square feet Natural Log of assessed land Total number of bedrooms Total number of bathrooms Year home was built (0 if property is vacant) Distance of Property to nearest Private School Distance of Property to nearest Public School Distance of Property to nearest Public Elementary School Distance of Property to nearest Hospital Dummy variable indicating whether property is greater than 1.01 Acres (=1 if yes, =0 otherwise) Dummy variable indicating whether residence is greater than 1459 square feet (=1 if yes, =0 otherwise) Dummy variable indicating whether property is has no built structure (=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Jacksonville(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Phoenix(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Shady Cove(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Talent(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Unicorporated Jackson County(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Unicorporated Jackson County and within 1.5 Mile Radius of Shady Cove(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Unicorporated Jackson County and within 3 Mile Radius of Shady Cove(=1 if yes, =0 otherwise) Dummy variable indicating whether property is within Unicorporated Jackson County and within 5 Mile Radius of Shady Cove(=1 if yes, =0 otherwise) Dummy variable indicating whether property is located on the Rogue River Unicorporated Jackson County(=1 if yes, =0 otherwise) Dummy variables representing year of sale (=1 if year sold is same as year, =0 otherwise) Interaction Terms generated by sold(year)*unincorporated 30

31 regjville regjacksonville VacantLand regjacksonvillea boveavgacreage 1 regjacksonvilllel argesqft regphoenix regphoenixvaca ntland regphoenixabov eavgacreage1 regphoenixlarge Sqft regtalent regtalentvacant Land regtalentabove AvgAcreage1 regtalentlarges qft reguninc84 reguninc84vaca ntland reguninc84abov eavgacreage1 reguninc84large Sqft reguninc2004 reguninc2004va cantland reguninc2004ab oveavgacreage1 reguninc2004lar gesqft Dummy variable representing being located within Jacksonville and sold after the implementation of the Riparian Ordinance (=1 If yes, =0 otherwise) Interaction term generated by regjacksonville*vacantland Interaction term generated by regjacksonville*aboveaverageacreage Interaction term generated by regjacksonville*largesqft Dummy variable representing being located within Phoenix and sold after the implementation of the Riparian Ordinance (=1 If yes, =0 otherwise) Interaction term generated by regphoenix*vacantland Interaction term generated by regphoenix*aboveaverageacreage Interaction term generated by regphoenix*largesqft Dummy variable representing being located within Talent and sold after the implementation of the Riparian Ordinance (=1 If yes, =0 otherwise) Interaction term generated by regtalent*vacantland Interaction term generated by regtalent*aboveaverageacreage Interaction term generated by regjacksonville*largesqft Interaction term generated by reguninc84*vacantland Dummy variable representing being located within Unincorporated Jackson County and sold after the implementation of the Riparian Ordinance (=1 If yes, =0 otherwise) Interaction term generated by reguninc84*aboveaverageacreage Interaction term generated by reguninc84*largesqft Dummy variable representing being located within Unincorporated Jackson County and sold after the implementation of the Riparian Ordinance (=1 If yes, =0 otherwise) Interaction term generated by regjuninc2004*vacantland Interaction term generated by regjacksonville*aboveaverageacreage Interaction term generated by reguninc2004*largesqft 31

32 A.3 Tables Table 2: Descriptive Statistics Variable Mean Std. Dev. Min Max salevalue acreage beds baths yearblt vacantland assesland assesimp residential shady_cove talent jacksonville pheonix unicorporated onrogue onclass sold sold sold sold sold sold sold

33 sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold sold

34 sold N

35 Table 3: Regression with Shady Cove, Talent, Phoenix, Jacksonville, and Unincorporated Jackson County (1) (2) VARIABLES Without Restrictions Less Than 4 Acres beds (0.0330) (0.0356) baths ** (0.0499) (0.0501) ln_sqft 0.495*** 0.409*** (0.0869) (0.0837) yearblt -1.01e e-05 (4.13e-05) (4.10e-05) ln_acreage * (0.0356) (0.0937) ln_landvalue 0.108*** (0.0335) (0.0337) shady_cove (0.116) (0.110) phoenix (0.760) (0.665) talent * (0.209) (0.198) jacksonville (0.184) (0.179) vacantland 3.284*** 2.718*** (0.607) (0.586) regunincrogue (0.151) (0.154) reguninc1984class (0.133) (0.132) regtalent (0.725) (0.633) regphoenix (0.757) (0.658) regjacksonville (0.343) (0.301) sold (0.555) (0.504) sold (0.569) (0.521) sold (0.582) (0.508) sold (0.593) (0.536) 35

36 sold (0.555) (0.501) sold (0.563) (0.493) sold (0.541) (0.484) sold (0.537) (0.475) sold (0.530) (0.473) sold (0.530) (0.473) sold (0.538) (0.506) sold (0.543) (0.480) sold (0.548) (0.495) sold (0.541) (0.482) sold (0.531) (0.470) sold * (0.528) (0.467) sold * (0.537) (0.479) sold (0.533) (0.472) sold ** (0.527) (0.465) sold ** (0.524) (0.460) sold * 1.088** (0.525) (0.461) sold * 1.134** (0.525) (0.461) sold ** 1.344*** (0.524) (0.461) sold ** 1.459*** (0.527) (0.465) sold ** 1.451*** (0.529) (0.467) sold ** 1.580*** (0.542) (0.478) sold ** 1.325*** (0.530) (0.467) 36

37 sold *** 1.743*** (0.548) (0.485) sold ** 1.265*** (0.539) (0.480) sold ** 1.523*** (0.535) (0.472) sold ** 1.410*** (0.535) (0.473) sold * 1.228*** (0.531) (0.468) sold (0.929) (0.811) onrogue (0.110) (0.103) dis_hosp -5.06e-06*** -5.20e-06*** (1.58e-06) (2.00e-06) dis_priv -1.16e e-07 (1.01e-06) (1.05e-06) dis_pub_elem 7.27e e-06 (5.60e-06) (6.58e-06) Constant 6.329*** 8.137*** (0.906) (0.858) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 37

38 Table 4: Final Regression with Shady Cove and Unincorporated Jackson County VARIABLES 3 Mile Boundary and Less Than 4 Acres beds (0.0423) baths 0.106* (0.0592) ln_sqft 0.334*** (0.0926) yearblt 1.59e-05 (4.39e-05) ln_acreage 0.214* (0.111) ln_landvalue (0.0405) unincorporated 1.550** (0.670) vacantland 2.119*** (0.649) regunincrogue (0.202) reguninc1984class (0.152) sold (0.520) sold (0.527) sold (0.529) sold (1.010) sold (0.923) sold (0.510) sold (0.875) sold (0.862) sold ** (0.855) sold * (1.003) sold

39 (0.510) sold (0.926) sold ** (0.878) sold ** (0.850) sold ** (0.835) sold *** (0.843) sold *** (0.914) sold ** (0.867) sold *** (0.827) sold *** (0.820) sold *** (0.829) sold *** (0.821) sold *** (0.819) sold *** (0.818) sold *** (0.836) sold *** (0.839) sold *** (0.830) sold *** (0.852) sold *** (0.859) sold *** (0.831) sold *** (0.851) sold *** (0.824) u_yr (0.975) u_yr

40 (0.841) u_yr (0.792) u_yr (0.753) u_yr * (0.737) u_yr * (0.915) u_yr (0.815) u_yr *** (0.866) u_yr (0.752) u_yr ** (0.720) u_yr ** (0.739) u_yr * (0.821) u_yr ** (0.762) u_yr * (0.724) u_yr *** (0.697) u_yr * (0.710) u_yr * (0.698) u_yr * (0.706) u_yr ** (0.705) u_yr *** (0.741) u_yr ** (0.772) u_yr * (0.741) u_yr ** (0.767) u_yr * (0.770) u_yr *** 40

41 (0.749) u_yr ** (0.751) u_yr ** (0.714) onrogue (0.118) dis_hosp -4.30e-06* (2.41e-06) dis_priv 8.28e-07 (1.44e-06) dis_pub_elem -1.17e-05 (9.89e-06) Constant 6.524*** (1.119) Observations 438 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 41

42 Table 5: Regression with Shady Cove and Unincorporated Jackson County (Properties with Above Average Acreage) (1) VARIABLES Interaction Terms: Above Average Acreage reguninc2004aboveavgacreage (0.283) reguninc1984aboveavgacreage ** (0.128) beds (0.0407) baths 0.108* (0.0565) ln_sqft 0.372*** (0.0906) yearblt 2.21e-05 (4.41e-05) ln_acreage (0.118) ln_landvalue (0.0371) unincorporated (0.774) vacantland 2.476*** (0.632) regunincrogue (0.230) reguninc1984class (0.165) Constant 7.534*** (1.011) Observations 491 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 42

43 Table 6: Regression with Shady Cove and Unincorporated Jackson County (Properties with Vacant Land) (1) VARIABLES Interaction Terms: Vacant Land reguninc2004vacantland (0.364) reguninc1984vacantland (0.193) beds (0.0424) baths 0.105* (0.0592) ln_sqft 0.336*** (0.0927) yearblt 8.65e-06 (4.47e-05) ln_acreage 0.222* (0.113) ln_landvalue (0.0420) unincorporated 1.540** (0.670) vacantland 2.179*** (0.653) regunincrogue (0.220) reguninc1984class (0.157) Constant 6.588*** (1.126) Observations 438 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 43

44 Table 7: Regression with Shady Cove and Unincorporated Jackson County (Properties with Higher Quality Homes) (1) VARIABLES Interaction Terms: Above Average Square Feet reguninc2004largesqft 0.656** (0.284) reguninc1984largesqft (0.145) beds (0.0380) baths 0.114** (0.0551) ln_sqft 0.353*** (0.0920) yearblt 7.31e-06 (4.62e-05) ln_acreage (0.117) ln_landvalue * (0.0437) unincorporated (0.590) regunincrogue (0.226) reguninc1984class (0.168) Constant 7.354*** (1.093) Observations 350 R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 44

45 Table 8: Regression with Shady Cove and Unincorporated Jackson County With Different Acreage Limitations VARIABLES (1) (2) (3) No Limitation on Less Than 10 Acres Less Than 4 Acres Acreage beds (0.0367) (0.0372) (0.0406) baths * 0.106* (0.0551) (0.0535) (0.0563) ln_sqft 0.453*** 0.419*** 0.370*** (0.0948) (0.0883) (0.0911) yearblt -1.46e e e-06 (4.47e-05) (4.25e-05) (4.40e-05) ln_acreage * (0.0380) (0.0638) (0.104) ln_landvalue 0.138*** (0.0365) (0.0356) (0.0372) unincorporated 1.410* (0.789) (0.734) (0.778) vacantland 2.997*** 2.841*** 2.449*** (0.660) (0.616) (0.635) regunincrogue (0.175) (0.176) (0.184) reguninc1984class (0.148) (0.139) (0.146) onrogue (0.121) (0.111) (0.113) dis_hosp -4.83e-06*** -4.51e-06** -4.79e-06** (1.78e-06) (1.93e-06) (2.26e-06) dis_priv -1.11e e e-07 (1.26e-06) (1.24e-06) (1.34e-06) dis_pub_elem 7.60e e e-06 (6.06e-06) (6.13e-06) (7.10e-06) Constant 4.867*** 6.254*** 7.499*** (1.227) (1.161) (1.016) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 45

46 Table 9: Regression with Shady Cove and Unincorporated Jackson County Within Different Distances to City Boundary (1) (2) (3) VARIABLES Within 5 Miles of Shady Cove Within 3 Miles of Shady Cove Within 1.5 Miles of Shady Cove beds (0.0406) (0.0423) (0.0792) baths 0.106* 0.106* 0.248** (0.0563) (0.0592) (0.116) ln_sqft 0.370*** 0.334*** (0.0911) (0.0926) (0.139) yearblt 8.93e e ** (4.40e-05) (4.39e-05) ( ) ln_acreage 0.194* 0.214* (0.104) (0.111) (0.184) ln_landvalue (0.0372) (0.0405) (0.0754) unincorporated ** (0.778) (0.670) vacantland 2.449*** 2.119*** (0.635) (0.649) (0.994) regunincrogue (0.184) (0.202) (0.309) reguninc1984class (0.146) (0.152) (0.774) Constant 7.499*** 6.524*** 9.426*** (1.016) (1.119) (1.787) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 46

47 47

48 48

49 49