INTRODUCTION



Interpreting “Fair Market Value”:

Zoning, land value, and the Possible Unanticipated Externalities of Oregon’s Ballot Measure 37

By

Chelsea R. Byers

Jesse Cathey

Robert Wyman

Presented to the Department of Economics, University of Oregon,

in partial fulfillment of requirements for honors in Economics.

under the supervision of

Professor Bill Harbaugh

________________________________

June 2005

Introduction

Oregon is known for its progressive land and zoning regulations. The motivations for these regulations are easy to see—a simple comparison with other areas of the United States shows much that distinguishes the beautiful state of Oregon from the rest. Our state is home to some of the most pristine farmlands, beautiful cities and well-used recreation areas in the country. Citizens have not wanted to see these natural and man-made amenities ruined by development, and thus have historically supported strict land-use and zoning regulations.

Recently Oregonians have supported a potential softening and partial rollback of the state’s previously established regulations, with the passage of Oregon Ballot Measure 37.

The goal for measure 37 is to compensate individuals whose property values have declined due to the implementation of zoning laws from the state level and below. According to the measure if you or your family has maintained ownership of a parcel of land before a zoning law was enacted, they are entitled to some form of relief. Injured parties can appeal to the infringing legislative body following the claims process laid out by the measure. The governing body has 180 days to respond to the claim, or the zoning measure is automatically repealed. Uniquely, the governing body can decide to lift the ordinance, or pay the claimants the difference in their land values. Zoning laws that cannot be repealed are those that protect the public health and safety.

The measure has been one of the most controversial state land-use initiatives in recent memory. Critics decry that land-use in Oregon—and all of the smart urban planning and environmental protections that Oregon land-use regulation policies sought to ensure—will never be the same. Advocates disagree sharply, saying that Measure 37 was a triumph for the rights of individual property owners, and an appropriate check on out-of-control state and local government regulations.

Economic analyses that aim to explain land-use policies can be beneficial to policymakers, the public, and advocates on both sides of the issues. It is this sort of analysis that our project will attempt to provide. In beginning our investigation, therefore, it is important to recognize that land-use regulations that attempt to minimize externalities and maximize public welfare can have both costs and benefits.

Perhaps the quintessential example illustrating the need for smart land-use planning is that of a feed lot abutting a residential area. The potential spillover effects of feedlot properties are obvious—the noise, manure, dirt and busy agricultural activities that accompany feedlots are resoundingly undesirable for residential neighbors, even though agriculture is an important part of a community’s economic activity.

A hypothetical scenario where a feedlot could simply be inserted into a residential area allows us to easily infer potential resulting affects on neighboring properties. Economic theory dictates that the negative externalities described above should lower the willingness to pay of consumers interested in purchasing the properties abutting the feedlot. Effectively, property values should decrease as prices are lowered until buyers are willing to purchase neighboring properties despite the feedlot’s negative externalities.

The feedlot example underscores the need for smart urban planning by illustrating what could happen in the absence of zoning regulations. Such regulations are put in place to minimize or eliminate unwanted externalities from uncomplimentary uses of neighboring or public properties. In this light, land-use planning can be seen as value maximizing: the values of many properties increase as zoning can make urban areas more desirable for consumers, industrial areas more dense and efficient, and so forth. Zoning regulations, however, effectively restrict the ways in which a property owner can use or develop his or her own land. Conflict often arises when landowners feel that these restrictions place an inordinate burden on their specific holdings, and that they are being forced to comply with regulations that may marginally help the community at the expense of their property ownership rights. Oregonians thus face a dilemma: should the government regulate land use if benefits to society exist, but subsequently devalue certain properties though regulation?

Though the above question is important and worth asking, our goal as economists is not to answer this normative question. Instead, we aim to provide a neutral economic analysis to the people of Oregon based on principles of cost-benefit analysis. At best, we hope to help give Oregonians different tools to interpret the Measure’s “fair market value clause” and aid individuals at the local and state level in making informed economic interpretations of the measure.

Oregon’s Ballot Measure 37

Just compensation as stipulated by Measure 37 is somewhat related to the classic issue of takings cases. The Fifth Amendment to the Constitution dictates that “private land shall not be taken for public use without just compensation.” While this seems to say that when a government entity takes private land for public use it owes a monetary award, in practice the word “takings” is open to interpretation given specific judicial determinations on a case by case basis. Nevertheless, in a traditional takings case there is substantial legal precedent for estimating the magnitude of a claim award. With property zoning issues that initiatives like Measure 37 seek to address, however, the high courts have remained reluctant to provide us with any clear cut means in which to test cases. As we will show, Measure 37 claims are by nature somewhat more ambiguous than usual takings cases, and that ambiguity may create extra difficulty when interpreting the Measure’s “fair market value” clause.

In a traditional takings case, a taking has occurred when the government takes 100 percent ownership of property. Once 100 percent of the land has been taken, the government is constitutionally required to compensate the owner for the fair market price of the land. Typically this is determined by an appraisal. Under Measure 37, Oregon’s voters now mandate that the state should compensate owners not only if their land is taken due to a government seizure, but also if it is devalued through a zoning ordinance. The governing body enacting an ordinance that diminishes a property’s value can now be sued and required to compensate the owner for an amount equal to the decrease in the value of the property. The governing body then can choose to lift the restriction or reimburse the owner the monetary damages.

In the few months since its implementation, there have been about 400 Measure 37 claims filed in many of Oregon’s counties. More are expected once the legislature defines rules and procedures for compensation. Many of the claims so far have dealt with agricultural land. For example, one claimant in Yamhill County is filing for five million dollars in compensation against land use regulation OAR 660-33-135 sections 6 and 7[1], which is believed to have reduced fair market value of the land in question. This statute limits the construction of dwellings on land zoned as high-value farmland. Only farms on this sort of land that produce more than $80,000 in revenue are exempted from these construction limits. This claimant would like to build dwellings on his high-value farmland, but he does not meet the exemption. The claimant is one example of the many people who have been denied the right to build on their property and would like the right to do so which was restricted sometime in the past.

While almost every claimant is seeking compensation into the hundreds of thousands of dollars, there are claims of only a few thousand dollars on the small end of the range. The newspapers usually write about the proposed subdivisions or golf courses, but other examples of claims include those who wish only to build one more home on their property. Varying interests and situations across Oregon have dictated a wide array of possible claims. One can only imagine how much work will be needed to investigate and determine the validity on each of the hundreds of claims made. The following demonstrates just some of processes that must be used to determine a final judgment on a given claim.

The text of Measure 37 stipulates that the State must respond to all claims within 180 days of the file date. If the regulation cited in the claim remains intact after this period, the claimant has a “cause of action for compensation under this act in the circuit court in which the real property is located.”[2] Therefore, the State has this period to decide whether to rule the claim invalid, waive the cited regulations, or pay the compensation as stated in the claim.

The measure was enacted in early December of 2004, meaning that the earliest claims are now due for decisions from various counties of Oregon. One of the earliest decisions has been made public by the Oregon Department of Land Conservation and Development.[3] This particular claim involves a piece of property zoned exclusive farm use (EFU) located in a rural section of Central Oregon near Bend. The following is a walk-through of the claim determination process and a demonstration of its arbitrary nature.

The owners of the property near Bend claimed that the EFU zone designation limited their ability to subdivide their 9.75 acre lot into eight lots comprising 1.2 acres each. In fact, this is true; a regulation in 1993 declared that new lots in EFU areas must be at least 80 acres in area. Simply, the claimants wish to change the zone from low density to high density. In their original claim, the owners ask for compensation of $950,000 from the State. This number was arrived upon by taking the current property’s value of $325,000 and then estimating the value each of the eight subdivided lots would have if they existed. The claim states that each vacant lot would be worth $150,000, along with a value of $225,000 for the lot currently occupied by the claimant’s house. Thus, the subdivision, if it were to exist, would be worth $1,275,000. Finally, the claim takes this value and subtracts the current property value for a grand total of $950,000. According to the report, these figures are estimated using Deschutes County Tax Assessor information and appraised values of similar properties.[4]

To judge the validity of this claim, the State must first verify that ownership of this property predates the EFU regulation. In fact, it was found that the claimant’s family has owned this tract since 1957, which predates the EFU standard. However, the validity of the dollar amount claimed was found questionable. The report states that “Without an appraisal based on the seven vacant lots and one lot with an existing dwelling, it is not possible to substantiate the specific dollar amount the claimants demand for compensation.”[5] The report effectively states that the claimed compensation is arbitrary, and that the exact value of a subdivision, if it existed where the property in question in question is located, cannot easily be determined. Furthermore, “because the claim does not provide a specific explanation for how the specified restrictions reduce the fair market value of the property, a specific amount of compensation cannot be determined.”[6] Therefore, compensation is not a reasonable way to settle the claim due to lack of information. Based on this ruling, the final determination on this claim is that the land value most likely was decreased due to the EFU regulation. However, the claimants, it was ruled, are only allowed the waiver of land use regulations since they acquired the property in 1999. While Measure 37 allows compensation for regulations enacted since a family member owned the land, waivers can only apply to regulations passed after the claimant acquired the land. Since EFU zoning was passed in 1993, the claimants were denied the ability to subdivide their land.

These rulings help to demonstrate that the State, as well as claimants, is having trouble determining how to interpret and calculate the “fair market value” clause of Measure 37. The goal of this paper is to develop an empirical model that can make the desired amount of compensation less arbitrary in future claims. With this model, the authors hope to show the affect on the value for any given property if, for example, the zone designation was changed from low to high density. For the claim above, our model could have given an estimate of value if the property happened to be subdivided. In theory, this method is much less arbitrary than searching through old Tax Assessor files and appraisals of comparable properties.

Literature Review

Before furthering our discussion of Measure 37 zoning issues we will clarify some of the different land-use issues that policies concerned with urban sprawl and the efficient use of public goods traditionally seek to address. Historically, urban growth boundaries and land-use zoning have been the principal policy tools of land-use regulators. Urban grown boundaries typically use urban service boundaries to control the land area available for development, and sometimes also limit the number of individual building permits to lessen burdens on their city’s infrastructure.[7]

Zoning, the second policy tool of land use regulators, is at the forefront of Oregon’s Measure 37 debate. According to a standard Urban Economics Text by O’Sullivan, zoning falls under three general categories: nuisance zoning (reducing exposure to pollution), fiscal zoning (excluding households that might impose a fiscal burden), and design zoning (arranging activities within the metropolitan area to improve aesthetics or circulation).

Analyses of other states’ land use and zoning laws are also an important base of literature for understanding Measure 37 property rights issues. In recent years, Florida has been a hotbed of controversy surrounding property rights and the regulation of land. Early in the 1990’s Florida passed a law that is similar to that of Measure 37 in a number of ways. Many academic and government sources can help paint a picture regarding Florida’s land use laws, the reasons they were enacted, and their effects on the state. According to Vargas (1996)[8], the estimated transfer of wealth to land owners due to “inordinate burden” under these laws is 28 to 50 billion dollars. Like Oregon, there are no state funds explicitly set aside for claims, so it is expected that much of the expense is passed onto taxpayers.

This is an important area of interest because Oregon has yet to think of a “best” way to measure the reductions in land value in order to distribute compensation. In the case of Florida, land use entities—those who enforce the land use laws—must provide the claimant with a “ripeness decision” that details whether the claimant’s proposed use of land is legal under the current laws, and what uses are allowable. If the claimant decides that the ruled allowable uses impinge upon his or her proposed use of the property, or its value, the case is taken to court to determine what the proper compensation amount should be. In theory, this method of filing claims should eliminate cases where no compensation is required and cut down on costly court time.

Vargas asserts that Florida’s land use law was enacted in order to give landowners compensation for the “fair” value of property values taken from property owners when their land is “inordinately burdened.” The definition of “inordinately burdened” is somewhat arbitrary, but is based on a landowner’s loss of use of part of their property, loss of the right to exclude others, and the loss of possible investment value the land may have had. Also, the law is intended to “signal to local governments to use greater caution and common sense when enacting new land use regulations,” and among other things, “facilitate decentralized decision-making.” According to this professor, the law was made without the intent of redistributing much income; however, there is nothing in the current scheme of the law to “satisfactorily mete out fairness.” At a time when state funding is lacking, these land use laws cannot promise that compensation will not come from taxpayer’s pockets. Therefore, the law does not protect the interests of the many against the one landowner. Compensating one landowner will likely cost much more than the benefits it will bring to society as a whole. Like current problems in Oregon, Vargas finds fault with the lack of proper and full definitions of when land value has been taken from the owner.[9]

Florida Attorney David Powell says that proper compensation is decided on a case-by-case basis and is determined by a judge who is solely responsible for determining whether the property has been “inordinately burdened” and ultimately decides on compensation by measuring the difference in the fair market value before and after the regulation.[10] Appraisers on both sides of the argument decide on a land value. If the government is forced to pay compensation, it then owns the land, due to the logic that by placing a certain regulation on it, it was in effect taking the land. Oregon, on the other hand, allows property owners to retain rights to their land regardless of compensation.

Another stipulation of Florida’s law is that a claimant can only receive compensation if they prove a given land-use regulation violates their “investment-backed expectations.”[11] This restriction takes into account the fact that when people buy property, they buy that property with certain expectations of how it can be used in the present and the future. Therefore, if a regulation is enacted which prevents these expectations from being realized, the owner can make a claim for compensation. Similarly, if it is ruled that an owner could not have expected to use their property in a certain way, they cannot issue a claim. For example, a rural landowner cannot make a claim because they are unable to build a subdivision or golf course. Without a developed infrastructure built, including things such as adequate roads and utilities, there is no way this owner reasonably expected to build such a subdivision when he acquired the land. The costs of developing such infrastructure would be too great for one landowner to overcome. Measure 37, on the other hand, does not include such a stipulation. This means that rural landowners could potentially claim for the ability to build a subdivision or receive the desired compensation, no matter the infrastructure available in their area. There is no safeguard from those who claim that they have been prevented from building a subdivision, yet have no intention to do so due to the overpowering capital costs it would take to implement.

Texas’s law, however, is not quite as restrictive as Measure 37 in terms of how much land value can be reduced before compensation is owed. The statute only considers distributing compensation for regulations that take more than 25% of a property’s value. Again, the value of the land is measured before and after a certain regulation was put in place. Currently, the Oregon legislature is proposing a bill that would adopt this policy. Presumably, it has resulted in fewer claims, and Oregon is looking for limitations on how many claims are brought into the courts.

Although Oregon’s new law is similar to both Florida and Texas’s policies, it goes above and beyond the other laws in ways that will surely increase the number of claims and compensable rulings. The laws in Florida and Texas act proactively, only affecting land use that is enacted after the law’s passing. In Oregon, the law works both proactively and retroactively.[12] This means that anyone who has owned a tract of land since the beginning of Oregon’s land use laws can file a claim. Most likely, this signals a larger base of possible claims for Oregon than any other state. Also, there is a requirement in all other states that the claimant provide notification to neighbors that there may be a waiver of regulations in their area. This is sensible considering the fact that negative externalities may occur when a neighbor decides to develop his land. Oregon, however, has no requirements for public notification of possible waivers.

On the subject of zoning and its externalities not only on any given property, but the neighboring properties as well, there has been emerging research. Netusil (2004) determined that there are statistically significant effects on land value resulting from zoning, most notably, environmental zoning. This author discusses two types of effects from zoning in which one increases land value, and the other decreases it. The “development effect” decreases land value by limiting an owner’s ability to subdivide their property, or adding structures such as patios. This limits the demand for a certain property since it limits the possible uses of that land. The “amenity effect” increases land value in environmental zones due to proximity to natural areas. Netusil theorizes that zoning creates scarcity and preserves the natural beauty of the area but does have a “negligible effect on values” due to the development effect. Ultimately, this study finds that “amenities” located within a given distance of a property have different effects on property values.[13] Homeowners demand tree canopies, streams, and parks but prefer not to have them immediately on or near their properties. Thus, people prefer their neighboring properties to have many natural amenities so they can benefit from the positive externalities, yet want the freedom to develop their own property as they please.

This literature review has indicated that Measure 37, as it is currently written, is flawed and arbitrary. Policy-makers could take a number of cues from other states and clear up much of the language and ambiguity of the measure. Also, based on the results of past studies, it has been found that the zoning on any given property can affect not only the value of that property, but the value of its neighboring lots. With these points in mind, this study hopes to show that arbitrary changes in zoning resulting from Measure 37 claims can cause negative externalities on surrounding properties. The authors hope to continue an open dialogue concerning the best and fairest ways to handle Measure 37 claims in the future.

Hypotheses:

Using economic analytic tools, we hypothesize that there are several logical ways of analyzing “fair market value”, depending on the economic criteria one uses to define that term. Further, we believe that these analytic considerations can lead to very different calculations of the magnitude of “fair market value,” and that therefore there may be several different but valid economic determinations of a given property value according to the text of Measure 37. Two general methods of determining “fair market value” are partial and general equilibrium. Partial equilibrium entails estimating the property value by excluding the effects on the neighboring properties. General equilibrium, on the other hand, estimates the value of a given piece of land accounting for the effects of its own change in value on surrounding properties.

If this hypothesis is true, then it is arguable that while “fair market value” is tacitly interpreted and discussed with reference to the hypothesized partial equilibrium—price of a property, interpreting Measure 37 property values in these exclusively partial equilibrium terms may ignore economic spillover effects, and the fact that a property’s own value is a function of its location in relation to other properties.

We theorize that as a property moves along the spectrum of zoning from low-density (R1) to high-density (R4), its value will increase. The value of a residential property with regards to zoning, however, is not only a function of its own zoning, but also that of the zoning of its neighboring properties. While we guess that the effect of low-density and medium-density neighbors on an R1 property may be insignificant, we suspect that the land value of an R1-zoned site will decrease if it is close to high-density housing. Thus, while a Measure 37 claimant requesting a zoning change from R1-R4 may see their own property value increase if they are able to sell their formerly R1 property as land for R4 high-density housing, there will also be a calculable negative spillover effect on the property values of it’s neighbors.

Before we continue, it should be noted that zoning was enacted as solution to the effects of rapidly growing urban areas. Various externalities occur when certain land uses go on relatively close to each other. For instance, when households are located near industrial or commercial zones, they may find the associated noise, pollution and activity annoying. Zoning areas of a city specific to land uses attempts to mitigate the conflicts that arise when opposing uses are in competition with each other.

This type of change would be represented by a claimant’s desire to divide their land into many properties, with the intention of building a housing subdivision. Similarly, if a claimant desired to turn their property into an apartment complex, their property value will increase due to a higher potential income. In theory, the greater the number of tenants living per square foot of land, the greater possible income the owner could extract in rent. Such an apartment building or subdivision would be even more valuable if it were located in a high income (i.e. R1) residential neighborhood, because potential tenants would be drawn to that neighborhood’s better schools and parks.

Conversely, it can be inferred that the property values in a high income area will decrease if an apartment building is suddenly built. Home buyers would opt for the neighborhoods and properties far away from a potentially noisy and congested apartment area, thereby decreasing demand for land near this development. Therefore, we should see a positive effect for a property owner who increases the allowable density on their land, and a decrease in property values due to negative externalities on the neighboring lands.

We further theorize that for a given residential-zoned property, a change to commercial or industrial zoning will have a positive land-value effect on a single property changed from residential to either commercial or industrial, but that negative effects will be forced on the properties in close proximity to that changed property. Again, we should notice a positive value effect on an owner’s property due to increased potential income from a commercial or industrial site. However, if this site were located in a residential or agricultural area—which is entirely possible given the current text of Measure 37—there should be a negative effect on surrounding property. New buyers will again opt for land far away from a potentially noisy and dirty area, thereby decreasing the demand for land near the commercial or industrial establishment.

These spillover effects are not explicitly considered in the language of Measure 37, but should be calculable by economic methods. The measure only allows relief from land devaluations due to land regulations enacted after an owner acquired their land, but does not consider those owners whose land has been potentially devalued from Measure 37 as a regulation in-and-of itself. Using models that incorporate and estimate zoning changes of properties within any given property in the Eugene area, we believe that these spillovers can be empirically verified. Further, acknowledging monetary damages other neighboring citizens may suffer in the form of decreased property values may be an integral part of reasonably calculating the “fair market value” of a M37 claim.

Data:

Obtaining and securing a viable data source was tricky. We secured statistical property data from RLID (Regional Land Information Database of Lane County) using the UO’s Geographic Information System (GIS) lab. The City of Eugene Planning and Zoning department declined to allow subsidized access to the information system. However, the data for 2003 was obtainable through the Economics Department at the University of Oregon, who had purchased it for use in projects from prior years.

RLID is said to be the most comprehensive information database available, containing extensive statistical data on all properties in Lane County, and specifically in the Eugene metro area.[14] This data contains specific land information that was central to a study such as this. The database has been maintained for over thirty years by the cities of Eugene and Springfield, the Eugene Water and Electric Board (EWEB), and the Lane Council of Governments. Among other things, the data gives information about many different property characteristics, latitude and longitude values, land area, property tax value, and geographic characteristics for all properties—and for some, even includes information about the most recent selling price and lien holders. This data makes possible many different applications, for example the ability to access any property given a unique identifier termed “maplot ID”. The RLID data is very extensive. However, the data we used was the tax lot id, price and housing characteristics. The larger data set set goes into detail greater than was necessary for a study such as ours.

The RLID data system contains up to date and current statistical property information. The frequencies with which certain portions of data are updated vary. Any changes in site addresses, tax lots and boundary information, map layers, and certain types of maps are updated on a weekly basis. The County Assessor files are integrated into the system on a monthly basis while the Deeds and Records indices are updated on a daily basis. Current and accurate information on properties such as those listed above is vital to any statistical property analysis. Without this information, it would be nearly impossible to extract information pertinent to drawing an analysis such as this.

Detailed property searches on RLID are also possible. One could potentially locate a given property and its characteristics by searching with the owner name, site address, Lane County Assessor account number, or the assessor map lot identification number. This feature is pertinent to our regression analysis of the properties so that we point out the effects of a change in zoning for certain properties.

Central to our study is the idea that land prices change over time with respect to the zoning of one’s property and the property surrounding the lot in question. RLID contains a history of all zoning based on a preceding application, CELAND. Zoning ordinances enacted previously to 1999 are not available through this application because CELAND was not Y2K compatible. Even though we do not use the specific zoning over time, it is worthwhile to account for this because it would be plausible to use in future analyses. For example, it would be interesting for us to question whether the changes in property values over time are correlated with the changes in zoning.

RLID records an in-depth deeds and records index that acts as an indexed search tool of documents recorded in Lane County, and the corresponding document images. This portion of RLID is updated each evening, and serves as a record of all recent transactions involving real property transactions and events. This information and document imagery contains the following: Grantor or grantee name, instrument or document number, instrument type and year, along with map and tax lot number. This is pertinent to this study because it keeps a keen record of all properties, and their recent selling values. Further data on land within the Eugene-Springfield area was accessible at the University of Oregon GIS computer lab. This Geographic Information System contained information from the Lane Council of Governments, along with United States Census data, which was maintained on a less frequent basis than RLID. This system not only contained the property tax value, but also detailed geographic information and specific housing characteristics which were excluded from RLID.

In order to utilize both data sources, the pertinent information in each data group needed to be merged and have certain features extracted that were central to our research. First and foremost, the data needed to be condensed into one data set in order to run regressions. Each observation in the different data groups shared a common map lot id, giving us a way to match properties across the different databases..

In order to retrieve a good, combined data source, we needed to integrate the GIS and RLID data sets. Each combined observation was transferred to Microsoft Excel. We found that the number of observations exceeded Excel’s capacity, so once the data was converted into Excel there were two groups for each data type. First and foremost, each property was assigned a dummy variable of 1 for its zoning classification. Secondly, each observation was assigned its own unique identification number, and these identifiers were ordered as ascending integers in our database so that each observation had a distinct and easily referenced identifier for our research.

Our next step in Excel was to put the data through macros. The program calculated how many and which properties were positioned within 1/4 and 1/8 mile away from a specific observation for each individual property. This portion of the research took the longest; we needed the program to calculate how far all 80,001 properties were from each of the 80,000 other properties. Incidentally this was very interesting as we blew a circuit in the GIS lab, perplexing other economic students and the evening maintenance crew. The different zones within the given distances from any given observation were also calculated.

One of the crucial aspects of our analysis was determining every property’s proximity to every other property in the database. In determining this distance, we used x-y coordinate data from the databases. These coordinates correspond to a map layout of Eugene metro area properties and were measured explicitly in feet, and thus are not arbitrary. There is one coordinate set for every property, presumably pertaining to the centermost point of the given property. To find the distance between properties we used a standard geometric formula to calculate the distance between two sets of x-y values.

For the sake of simplicity, we will call the selected property “property x”. Determining property x’s relationship to all other properties in the database was necessary so that we could determine several aspects of property x’s situation:

1) The number of other properties within quarter mile and eighth mile radii of the center of property x, respectively.

2) The zoning of the properties identified within the quarter and eighth mile circular observation zones.

3) The zoning composition of neighboring properties within the eighth and quarter mile circles, in percentage terms.

Diagrams for the calculated regions within the eighth and quarter mile conditions are provided in diagrams A1 and A2 in the appendix, the purple area being the observed region.

One potential problem with using the eighth and quarter mile measures together is that one area is included within the other. The circle comprised of the eighth mile radius is nested within the quarter-mile radius circle, therefore the first circle would study the aspects of zoning within an eighth of a mile of x and the second would study the aspects of zoning within both an eighth and a quarter mile of x. To be able to observe the independent effects of properties within a quarter of a mile but not within an eighth of a mile, we created a third “donut” variable which counts the properties within a quarter mile but excludes those only found within the eighth mile region. A graphical representation of the observed quarter mile donut area can be found in A3.

Thus, we observed zoning factors in three regions with regard to the center of property x: the region within an eighth mile radius, the region within a quarter mile radius, and the region within a quarter-mile “donut” of property x, where quarter mile donut=1/4 mile radius-1/8 mile radius.

Another potential problem with considering zoning factors in a set, circular, geometric region with regard to property x is that determining the percent zoning composition of a property’s neighbors within the exact circular boundary entailed a level of complicated calculation and special data interpretation that could not be supported by our available resources. Consequently, in considering the zoning of properties within the eighth and quarter mile spheres of influence we counted properties “within” these zones as properties whose given x-y identifying coordinates fell within either radius.

For example, the diagram in A4 shows a visual representation of the properties within the eighth mile circle of x. Property x is represented by the cyan blue region, and uncolored space on the chart is land that is not considered by this model. There are two properties that we considered within an eighth mile radius, besides property x: the rose colored property zoned R2 and the tangerine property zoned Ag. While the R2 property clearly extends beyond the eighth mile border, its entire area is counted as being “within” the eighth-mile region because at least a portion of it is within x’s eighth mile radius. A percentage calculation for the percent of R2 properties within an eighth mile of property x, then, would be arrived at by dividing the total area of the R2 property by the total areas of the three properties R2, Ag and R4 added together. Again, the diagram shows that this makes for a “sphere of influence” calculation for the eighth mile proximity, rather than an exact geometric breakdown of the zoning composition strictly within that eighth mile circle.

The logic behind the quarter mile sphere of influence with regard to property x is essentially the same, and a diagram is provided in A5. Six properties are considered within the quarter mile criteria: two R4 properties, one Ag, one R2, an I1 industrial property and a C1 commercial property. Again, the percent composition of a given type of zoning within a quarter mile of property x would be arrived at by dividing the represented area of the zoning type in question by the sum of the total area of all the properties cited to lie within the borders of the quarter mile region.

Diagram A6 represents this methodology in the context of the “quarter-mile donut” sphere. It is important to notice that while the orange Ag property would be included in both the eighth and quarter mile sphere criteria, it is left out of the calculations in the donut model. The method of calculating the percentage composition of types of zoning within the given, considered area remains the same, but only those properties that at least partially overlap into the region between the eighth and quarter mile circle are included in the calculation.

Results:

Regression sets one and two represent two preliminary but explanatory types of modeling for potential spillover effects of zoning. The regressions below are meant to be illustrative examples of the modeling potential of the database we have created. Now that we have developed a database of proximity relationships between properties in Eugene and Springfield, there is much larger scope for creating complex models that accurately isolate and describe some of the spillover effects of zoning that may otherwise go tacitly unconsidered by initiatives such as Ballot Measure 37. With the most accurate modeling possible as our final goal, the following are some of our initial findings. Table I below describes the variables used in the regressions that follow. Furthermore, summary statistics for each variable are in Table A7 in the appendix. Table A8 includes important statistics regarding the residential property dummy variables.

Table I:

Descriptions of Used Variables

|Variable |Definition |

|Landval |Land value, in dollars |

|Log(landval) |Log transformation of landval |

|Acres |Number of acres present in a single property |

|Log(acres) |Log transformation of acres |

|Percentr1a8th |Of the known zoning types within an eighth-mile radius of a given property, the percentage of that |

| |known area zoned r1a. |

|“good” zoning |For our preliminary calculations, the “good” zoning variable represented residential areas, equal to|

| |r1a+r2+r3+r4 |

|Percentgdonut4th |Of the known zoning types within a quarter-mile donut radius of a given property, the percentage of |

| |that known area zoned “good”. |

|Percentgood8th |Of the known zoning types within an eighth-mile radius of a given property, the percentage of that |

| |known area zoned “good”. |

|R1a, R2, R3, R4 Dummy |All of the residential zoning types in our dataset. R1a represents the lowest density housing, |

|Variables |comprising of suburban-type homes. As the number increases, so does the density. R4 represents the|

| |highest density housing such as an apartment building. |

|Accountedarea8th |A dummy which takes a value of 1 if, given the expected geometric square acreage within an |

| |eighth-mile radius of any given observation, the acreage calculated within the eighth-mile proximity|

| |of a particular observation sums to between 70% and 130% of a the expected acreage within the |

| |eighth-mile circle, and is zero otherwise. |

|Accountedarea4th |A dummy which takes a value of 1 if, given the expected geometric square acreage within a |

| |quarter-mile radius of any given observation, the acreage calculated within the quarter-mile |

| |proximity of a particular observation sums to between 70% and 130% of a the expected acreage within |

| |the eighth-mile circle, and is zero otherwise. |

|Excludez |A dummy variable which takes a value of 1 if the zoning of an observation is Ag, Bk, CI, GO, H, H/, |

| |HD, HI, LD, LM, MD, MR, MU, NC, PL, QM, SD, SH and takes a value of zero otherwise. |

|Excludea |A dummy variable that takes a value of 1 if the accountedarea8th or the accountedarea4th variable is|

| |equal to zero, and takes a value of zero otherwise. |

|Excludev |A dummy variable that takes a value of one if a given land value is less than or greater than a |

| |given set of property value parameters, and takes a value of zero otherwise. |

|Excluder |A dummy variable that takes a value of 1 if the sum of r1a+r2+r3+r4 is equal to zero, and takes a |

| |value of zero otherwise. |

Data Restrictions:

For the regressions below we restricted the excludev variable so that our sample included properties whose values range from 0 to $197,820.00. We justified these exclusions because of visible outliers and anomalies in our data set, including property values of zero that were most likely records of inherited property transactions, and extremely expensive properties (in some cases, upwards of 3 million dollars) that could potentially have skewed our sample observations.

Regression 1a:

log(landval) = βlog(acres)+βpercentr1a8th

if excludez + excluder + excludea + excludev==0 & r1a

Source | SS df MS Number of obs = 23314

-------------+------------------------------ F( 2, 23311) = 3899.54

Model | 3982.30588 2 1991.15294 Prob > F = 0.0000

Residual | 11902.8855 23311 .510612395 R-squared = 0.2507

-------------+------------------------------ Adj R-squared = 0.2506

Total | 15885.1914 23313 .681387698 Root MSE = .71457

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

llandval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

lacres | .4874324 .0055716 87.48 0.000 .4765116 .4983531

percen~1a8th | .0820943 .0187668 4.37 0.000 .04531 .1188785

_cons | 10.90102 .017281 630.81 0.000 10.86715 10.93489

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

Regression 1a looks at the logged land value of R1a properties with our aforementioned exclusion criteria, using βlog(acres) and βpercentr1a8th as regressors. The coefficient for βlog(acres) is extremely significant, indicating that every one acre increase in βlog(acres) will increase land value by approximately 0.49%. The coefficient βpercentr1a8th is also strongly significant, indicating that for every one percent increase in the presence of R1a zoned land within an eighth mile of a property, land value will increase by approximately 8.20%.

Regression 1b:

log(landval)=βlog(acres)+ βpercentr1a8th

if excludez + excluder + excludea + excludev==0 & r2

Source | SS df MS Number of obs = 1223

-------------+------------------------------ F( 2, 1220) = 237.72

Model | 274.35706 2 137.17853 Prob > F = 0.0000

Residual | 704.015467 1220 .577061858 R-squared = 0.2804

-------------+------------------------------ Adj R-squared = 0.2792

Total | 978.372527 1222 .800632182 Root MSE = .75965

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

llandval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

lacres | .4611639 .0211741 21.78 0.000 .4196221 .5027057

percen~1a8th | .0907883 .061713 1.47 0.142 -.0302871 .2118637

_cons | 10.80998 .0583185 185.36 0.000 10.69556 10.92439

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

Regression 1b looks at the logged land value of R2 properties with our aforementioned exclusion criteria, again using βlog(acres) and βpercentr1a8th as regressors. The coefficient for βlog(acres) is still extremely significant, and according to the model, every one acre increase in βlog(acres) will increase land value by approximately 0.46%. The coefficient for βpercentr1a8th, however, is not significant in this regression—indicating that an increase in R1a property within an eighth-mile area radius of an R2 property does not visibly and substantially increase the original property’s value.

Regression 1c:

log(landval)=βlog(acres)+ βpercentr1a8th

if excludez + excluder + excludea + excludev==0 & r3

Source | SS df MS Number of obs = 491

-------------+------------------------------ F( 2, 488) = 90.05

Model | 87.5068456 2 43.7534228 Prob > F = 0.0000

Residual | 237.101905 488 .48586456 R-squared = 0.2696

-------------+------------------------------ Adj R-squared = 0.2666

Total | 324.608751 490 .662466838 Root MSE = .69704

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

llandval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

lacres | .4642008 .0372945 12.45 0.000 .3909233 .5374784

percen~1a8th | -.2314051 .1217568 -1.90 0.058 -.4706373 .0078271

_cons | 11.2226 .1106828 101.39 0.000 11.00513 11.44007

Regression 1c takes the logged land value of R3 properties with our aforementioned exclusion criteria, again using βlog(acres) and βpercentr1a8th as regressors. The coefficient for βlog(acres) is still very significant, demonstrating that for every one acre increase in βlog(acres), land value will increase by approximately 0.46%. The coefficient for βpercentr1a8th, however, is not significant in this regression—indicating that an increase in R1a property within an eighth-mile area radius of an R3 property does not visibly and substantially increase the original property’s value.

Regression 1d:

log(landval)=βlog(acres)+ βpercentr1a8th

if excludez + excluder + excludea + excludev==0 & r4

Source | SS df MS Number of obs = 130

-------------+------------------------------ F( 2, 127) = 20.47

Model | 11.0012873 2 5.50064364 Prob > F = 0.0000

Residual | 34.1197175 127 .268659193 R-squared = 0.2438

-------------+------------------------------ Adj R-squared = 0.2319

Total | 45.1210048 129 .349775231 Root MSE = .51832

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

llandval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

lacres | .6373253 .1080705 5.90 0.000 .4234733 .8511773

percen~1a8th | -.8022532 .1978377 -4.06 0.000 -1.193738 -.4107681

_cons | 12.09272 .2769033 43.67 0.000 11.54478 12.64067

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

Regression 1d uses the logged land value of R4 properties with our aforementioned exclusion criteria, using βlog(acres) and βpercentr1a8th as regressors. The coefficient for βlog(acres) is still very significant, indicating that according to the model every one acre increase in βlog(acres) will increase land value by approximately 0.63%. The coefficient for βpercentr1a8th, interestingly enough, is both significant and negative in this regression. The model asserts that for every one percent increase in the presence of an R1a zoned piece of land within an eighth mile of the observed property, land value should decrease by approximately 80.0%.

Regressions 2a-2d:

Data Restrictions:

For these regressions we told Stata to restrict the excludev dummy so that it excluded observations from the populations whose landval variable was above the 95th percentile and below the 5th percentile. We implemented this reasoning that a somewhat narrower band of observations may lead to a higher R-squared value, and even more representative results.

Regression 2a:

Landval=βacres+βpercentgdonut4th+βpercentgood8th

if r1a==1 & excludez + excludea + excludev + excluder==0

Source | SS df MS Number of obs = 22086

-------------+------------------------------ F( 3, 22082) = 98.55

Model | 4.6304e+10 3 1.5435e+10 Prob > F = 0.0000

Residual | 3.4583e+12 22082 156611430 R-squared = 0.0132

-------------+------------------------------ Adj R-squared = 0.0131

Total | 3.5046e+12 22085 158686767 Root MSE = 12514

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 1045.38 65.42361 15.98 0.000 917.145 1173.615

percentgdo~h | 4053.092 545.0221 7.44 0.000 2984.81 5121.375

percen~od8th | -1398.311 585.5651 -2.39 0.017 -2546.061 -250.5618

_cons | 29272.32 448.9491 65.20 0.000 28392.35 30152.29

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

The above regression uses the land value of the R1a properties that meet our criteria for inclusion in the model, with βacres, βpercentgdonut4th, and βpercentgood8th as regressors. According to the statistically significant coefficient on acres, every one acre increase to a given property should increase the land value of that property by $1045.38. The statistically significant coefficient for βpercentgdonut4th tells us that for every one percent increase in the presence of land with “good” zoning within a quarter mile “donut” of a property, land value should increase by $4053.09. The zoning of land within an eighth mile radius also has a statistically significant effect, where the coefficient for βpercentgood8th implies that every additional one percent of land with “good” zoning should decrease land value by $1398.31. Although it is curious that residential property further away from a given low-density residential area is more valuable than the same property within very close proximity, the authors surmise that land value may increase if there is a small neighborhood store or shopping center nearby. This convenience effect may cause a small amount of commercial property close by to increase value, but a large amount of commercial property a little further away to have a negative effect on value.

Regression 2b:

Landval=βacres+βpercentgdonut4th+βpercentgood8th

if r2==1 & excludez + excludea + excludev + excluder==0

Source | SS df MS Number of obs = 1149

-------------+------------------------------ F( 3, 1145) = 15.66

Model | 8.1638e+09 3 2.7213e+09 Prob > F = 0.0000

Residual | 1.9897e+11 1145 173771541 R-squared = 0.0394

-------------+------------------------------ Adj R-squared = 0.0369

Total | 2.0713e+11 1148 180428761 Root MSE = 13182

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 2843.721 429.1931 6.63 0.000 2001.628 3685.814

percentgdo~h | 5359.596 2537.09 2.11 0.035 381.7299 10337.46

percen~od8th | -385.8447 2668.601 -0.14 0.885 -5621.742 4850.052

_cons | 25366.77 1701.942 14.90 0.000 22027.5 28706.05

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

Similarly, the above regression uses the land value of R2 properties that meet our criteria for inclusion in the model, with βacres, βpercentgdonut4th, and βpercentgood8th as regressors. According to the statistically significant coefficient on acres, every one acre increase to a property should increase the land value of that property by $2843.72. The statistically significant coefficient for βpercentgdonut4th tells us that every one percent increase in the presence of land with “good” zoning within a quarter mile “donut” of a property, land value should increase by $5359.60. The zoning of land within an eighth mile radius does not have a statistically significant effect on land value in this regression.

Regression 2c:

Landval=βacres+βpercentgdonut4th+βpercentgood8th

if r3==1 & excludez + excludea + excludev + excluder==0

Source | SS df MS Number of obs = 468

-------------+------------------------------ F( 3, 464) = 5.57

Model | 3.0180e+09 3 1.0060e+09 Prob > F = 0.0009

Residual | 8.3847e+10 464 180704571 R-squared = 0.0347

-------------+------------------------------ Adj R-squared = 0.0285

Total | 8.6865e+10 467 186006200 Root MSE = 13443

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 1060.584 383.9328 2.76 0.006 306.1218 1815.047

percentgdo~h | 866.2767 4078.231 0.21 0.832 -7147.812 8880.366

percen~od8th | -7555.56 4459.636 -1.69 0.091 -16319.14 1208.024

_cons | 39545.11 3490.522 11.33 0.000 32685.92 46404.3

Here we use the land value of R3 properties that meet our criteria for inclusion in the model, with βacres, βpercentgdonut4th, and βpercentgood8th as regressors. According to the statistically significant coefficient on acres, every one acre increase to a property should increase the land value of that R3 property by $1060.58. In this case, the quarter mile donut area does not reveal substantial evidence that land value is altered by R3 parcels being in close proximity to “good” zoning. However, βpercentgood8th is significant at the 10% level. This indicates that having residential property near this high-density housing has a positive affect on land value, presumably opposed to the negative affect that commercial or industrial property would exert if it were near an R3 property.

Regression 2d:

Landval=βacres+βpercentgdonut4th+βpercentgood8th

if r4==1 & excludez + excludea+ excludev + excluder==0

Source | SS df MS Number of obs = 126

-------------+------------------------------ F( 3, 122) = 7.67

Model | 2.7318e+09 3 910600108 Prob > F = 0.0001

Residual | 1.4490e+10 122 118769800 R-squared = 0.1586

-------------+------------------------------ Adj R-squared = 0.1379

Total | 1.7222e+10 125 137773727 Root MSE = 10898

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 41575.18 9392.908 4.43 0.000 22980.98 60169.38

percentgdo~h | 2843.396 6758.773 0.42 0.675 -10536.27 16223.06

percen~od8th | 1495.196 6271.912 0.24 0.812 -10920.68 13911.07

_cons | 21525.06 6074.233 3.54 0.001 9500.505 33549.61

Now we use the land value of R4 properties that meet our criteria for inclusion in the model, with βacres, βpercentgdonut4th, and βpercentgood8th as regressors. According to the statistically significant coefficient on acres, every one acre increase to a property should increase the land value of that R4 property by $41,575.18. In this case, neither the quarter mile donut area nor the eighth mile radius area reveal substantial evidence that land value is altered by R4 parcels being in close proximity to “good” zoning.

Regressions 3a-3d:

Regression 3a

Landval=βacres+βr1a+βr2 +βr3

if excludez + excluder + excludea + excludev ==0

Source | SS df MS Number of obs = 25741

-------------+------------------------------ F( 4, 25736) = 491.80

Model | 1.1929e+13 4 2.9821e+12 Prob > F = 0.0000

Residual | 1.5605e+14 25736 6.0637e+09 R-squared = 0.0710

-------------+------------------------------ Adj R-squared = 0.0709

Total | 1.6798e+14 25740 6.5262e+09 Root MSE = 77870

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 12869.87 290.2941 44.33 0.000 12300.87 13438.86

r1a | -2393.129 6771.267 -0.35 0.724 -15665.19 10878.93

r2 | -6074.208 7088.884 -0.86 0.392 -19968.82 7820.403

r3 | -899.9654 7580.395 -0.12 0.905 -15757.96 13958.03

_cons | 33667.48 6752.501 4.99 0.000 20432.2 46902.76

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

This regression contains all of the residential properties as dummy variables. R4 is left out of the regression. Since all of the included dummies have negative coefficients, it appears that R4 is the most valuable of all the residential zoning types. For instance, a property of the same size will be worth $2393.13 less if it is zoned R1a instead of R4. Unfortunately, only the coefficient on acres is significant at the 10% level.

Regression 3b

Landval=βacres+βr2+βr3 +βr4

if excludez + excluder + excludea + excludev ==0

Source | SS df MS Number of obs = 25741

-------------+------------------------------ F( 4, 25736) = 491.80

Model | 1.1929e+13 4 2.9821e+12 Prob > F = 0.0000

Residual | 1.5605e+14 25736 6.0637e+09 R-squared = 0.0710

-------------+------------------------------ Adj R-squared = 0.0709

Total | 1.6798e+14 25740 6.5262e+09 Root MSE = 77870

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 12869.87 290.2941 44.33 0.000 12300.87 13438.86

r2 | -3681.079 2215.636 -1.66 0.097 -8023.85 661.6916

r3 | 1493.163 3481.55 0.43 0.668 -5330.871 8317.197

r4 | 2393.129 6771.267 0.35 0.724 -10878.93 15665.19

_cons | 31274.35 520.914 60.04 0.000 30253.33 32295.37

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

This regression includes all residential dummies except R1a. This regression states that a property of the same size is worth more if it is zoned R4 or R3 rather than R1a. However, if it were zoned R2, it would be worth $3681.08 less than if it were zoned R1a. Only the coefficients on acres and R2 are significant at the 10% level.

Regression 3c

Landval=βacres+βr1a+βr3 +βr4

if excludez + excluder + excludea + excludev ==0

Source | SS df MS Number of obs = 25741

-------------+------------------------------ F( 4, 25736) = 491.80

Model | 1.1929e+13 4 2.9821e+12 Prob > F = 0.0000

Residual | 1.5605e+14 25736 6.0637e+09 R-squared = 0.0710

-------------+------------------------------ Adj R-squared = 0.0709

Total | 1.6798e+14 25740 6.5262e+09 Root MSE = 77870

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 12869.87 290.2941 44.33 0.000 12300.87 13438.86

r3 | 5174.242 4064.508 1.27 0.203 -2792.422 13140.91

r1a | 3681.079 2215.636 1.66 0.097 -661.6916 8023.85

r4 | 6074.208 7088.884 0.86 0.392 -7820.403 19968.82

_cons | 27593.27 2162.602 12.76 0.000 23354.45 31832.09

This regression includes all of the residential dummies except R2. It states that every other residential property zone type is worth more than an R2 zoned property. Here, the coefficients on acres and R1a are significant at the 10% level.

Regression 3d

Landval=βacres+βr1a+βr2 +βr4

if excludez + excluder + excludea + excludev ==0

Source | SS df MS Number of obs = 25741

-------------+------------------------------ F( 4, 25736) = 491.80

Model | 1.1929e+13 4 2.9821e+12 Prob > F = 0.0000

Residual | 1.5605e+14 25736 6.0637e+09 R-squared = 0.0710

-------------+------------------------------ Adj R-squared = 0.0709

Total | 1.6798e+14 25740 6.5262e+09 Root MSE = 77870

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

landval | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

acres | 12869.87 290.2941 44.33 0.000 12300.87 13438.86

r1a | -1493.163 3481.55 -0.43 0.668 -8317.197 5330.871

r2 | -5174.242 4064.508 -1.27 0.203 -13140.91 2792.422

r4 | 899.9654 7580.395 0.12 0.905 -13958.03 15757.96

_cons | 32767.52 3447.408 9.50 0.000 26010.4 39524.63

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

This regression contains all of the residential dummies except for R3. It states that a property of the same size is worth less if it is zoned R1a or R2. However, if it is zoned R4, it is worth $899.97 more. Unfortunately, only the coefficient on acres is significant at the 10% level.

Table II: Summary Stats by Zone

|Sampled Zoning |Percentage of 1/8 |Percentage of 1/4 |Mean Property |Mean Property |

|Type |Mile accounted for|Mile accounted for|Value |Size (acres) |

| |by 1 acre |by 1 acre | | |

|R1a |3.18 |0.796 |$43,586.81 |0.65 |

|R2 |3.18 |0.796 |$43,539.32 |0.71 |

|R3 |3.18 |0.796 |$58,168.06 |0.55 |

|R4 |3.18 |0.796 |$52,428.39 |0.75 |

Table III: Calculated Benefits (3a-3d)

|Zone Type |Benefit R1a Zoning |Benefit R2 Zoning |Benefit R3 Zoning |Benefit R4 Zoning |

|R1a |- |-3681.08 |1493.16 |2393.13 |

|R2 |3681.80 |- |5174.24 |6074.21 |

|R3 |-1493.16 |-5174.24 |- |899.97 |

|R4 |-2393.13 |-6074.12 |-899.97 |- |

Table IV: Calculated Total 1/8 Mile Radius Costs (1a-1d)

|Zone Type |Total Spillover Effect of One Additional % Zoned r1a in 1/8-mile Radius |

|R1a |3,573.44 |

|R2 |4814.44 |

|R3 |-13,426.74 |

|R4 |-41,708.36 |

Table V: Effect on Property Value, by Zone (2a-2d)

| |Sample Zoning |Avg. Effect of one additional |Additional 1% “good”, |Additional 1% “good”, 1/4|Est. Additional 1% |

| |Type |acre of same zone to original |1/8 Mile |Mile Donut |“good”, 1/4 Mile |

| | |property | | | |

|2b |R2 |**$2,019.04 |-$385.84 |**$5,359.60 |$4973.76 |

|2c |R3 |**$583.32 |*-$7,555.56 |$866.27 |-$6,689.29 |

|2d |R4 |**$31,181.25 |$1,495.20 |$2,843.40 |$4,338.59 |

Regression Summaries and Conclusions:

Tables III and IV give the results of calculated costs and benefits that result from zoning changes, given the coefficients in regression sets 3 and 1, respectively. Table IV provides the total spillover effects that would occur within an eighth-mile radius of a property with a given residential zoning, if R1a property made up one additional percent of the composition of its neighboring properties.

Benefits were calculated using the coefficients from the dummy variable zoning matrix in regressions 3a-3d, given in Table III. Because of the form of our model, these coefficients could be copied directly into the table, as they represent the dollar amount land value would increase or decrease if the zoning on a given property were changed.

To calculate the costs we used regression sets 1a-1d. Due to the log-log form of the model, the coefficients give the percent land value increase per an additional one percent increase in R1a zoning within a given R1a-R4 zoned property’s eighth mile radius. To translate the log-log model into the total dollar amount for which properties within the eighth-mile radius would be affected by this one percent R1a addition, we used the formula:

[(βpercentr1a8th / 100) · {(31.4) / (avg. Ri dummy property size)}] · [(avg. Ri dummy property value) · (avg. no. properties w/in 1/8-mile of given Ri)] = Total Spillover Cost for Ri, 1/8th-mile

Where Ri accounts for each type of residential zone. We divide βpercentr1a8th by 100 because in its form in the regressions, it is a proportion value (of the effect on land value) and we must change it to a percent value. We then multiply this by the number of acres in an 1/8th mile radius area (31.4) divided by the average property size to figure out how much of the 1/8th area is taken up by each property (on average). We then multiply this value by the total property value of all properties within the 1/8th mile radius (on average). The resulting answer gives us the dollar effect that Ri has on the land values of on average number of properties of an average size within the 1/8th mile radius.

These regressions give us an illustration of the positive benefit a property receives from acquiring another “acre” of the same zoning. This simulates any owner of a property immediately purchasing and adding an acre onto their land. They also tell us the way that a given property would be influenced by the spillover effects of additional R1a properties within the given eighth-mile radius. One of the most encouraging aspects of this regression set is that in general, for a given property of a particular zoning type, acquiring more acreage of the same zoning type appears to increase land value. Also, value seems to increase in proportion to higher allowable density on the expanded property area. For example, an additional acre of R1a added to an R1a property seems to increase property value less than an additional acre of R4 added to an R4 property. We might infer from this model that increasing the amount of high density zoning for any given owner’s property will increase land value much more than a similar area increase on the original property, this time zoned low density.

Empirically speaking, we were happy to notice that the net benefits as reported in Table III increase as we move from R1 to R4 (with the exception of R3). A priori, we expected this result because higher density zones should be more potentially income-generating and therefore be worth more. We are curious as to why the R3 results do not follow this trend; there could be problems with the R3 data, a need to increase the complexity of the model, a potential omitted variable bias, or more restrictions may be needed on the dataset to eliminate outliers. Our general inference regarding the value of additional acreage increasing as allowable density increases, however, must be tested in the future with an analysis which regresses the land value of every type of zoning against any additional acres of differing zoning types. Succinctly, land zoned higher density is apparently proportionally more valuable for the original observation to acquire, though again this conclusion requires further investigation.

These Regressions continue to uphold economic theories about how zoning spillover effects can affect the land values of neighboring properties. Although the results of our regression analysis are not conclusive, we can—maintaining certain assumptions—further notice that zoning effects land values of not only the re-zoned property, but that of neighboring properties. For simplicity, we omitted certain housing characteristics, along with other factors that influence consumer demand such as elevation for view, proximity to parks and recreational facilities, quality of a school district, crime rates, and shopping. Singling out the these fringe benefits of home ownership allowed us to concentrate primarily on the effects of zoning on land prices, and the effects of changing the zoning classification.

One factor that probably skewed our results may be the neighborhoods surrounding the University Oregon. Our regressions indicate that residential land is more valuable when closer in proximity to high density zoned land. These results are puzzling; we expect that residential land owners would not want to live near an apartment complex. In turn, this would reduce demand for these properties and therefore decrease land values. We postulated that this effect would take place due to the negative features of living near an apartment complex. These problems include excess noise and traffic that magnify neighbor-to-neighbor problems. However, we believe that residential housing near the University is highly demanded, and therefore is worth more. Moreover, these low density properties are located very near many apartments which house many students. Therefore, due to preferences, residential houses may be worth more when near apartment complexes. We also have to assume that consumer preferences still play a role on the land value, because this is also demand-based value. Our regressions do not do a very good job explaining this phenomenon, so we have to assume that we are omitting an important preference variable.

Take a second look at the campus neighborhoods, where students are willing to compromise their standard of living for a shorter commute to school. In the immediate vicinity of the campus there are larger concentrations of apartment buildings and further out, the concentration of apartment complexes decreases. Mixed into the decreased apartment density are residentially zoned lots. The land value of these properties is higher than what would be expected. The demand for housing near the University is much higher, and therefore, the higher willingness to pay for rent increases the land value. We assume that the land value is based on the stream of revenues it generates for land owners. Due to the left-over principle which states that the cost of land is high because the cost of the output (dwellings) is also high due to the high demand. Perhaps if this study were conducted in a large city rather than a small “college town,” we would find results that match our expectations.

Some of our results are arguably skewed because our analysis does not coincide completely with the hypothesis. According to the theory of Monocentricity[15], residential land near a city’s downtown area is worth more than residential land away from the city. However, these properties occupy lower income households due to the demand for lower transportation costs and less square footage. Acreage, newer homes, and less noise, on the other hand, drive high income households away from the city. Consequently, the properties near the city are smaller so that the low income own little of this high-value land. Therefore, if the data for high density residential properties includes the value of the land which an entire apartment building is built on, and if that building were counted as one property, this could skew our data since what is in reality a group of low-density housing will appear to be one high-density lot in our regressions. This could explain why Table III shows that R4 housing near R1a housing increases the value of the R1a property. This counteracts our hypothesis; only low density properties should increase the value of other low density properties, according to our expectations.

The University of Oregon makes up for a remarkably large composition of the city’s inhabitants. Students wishing to economize on their commuting costs live closer to the university and occupy smaller spaces in exchange for the decrease in costs. Even though the university campus is just blocks away from Eugene’s downtown area, we surmise that the land value near the campus is relatively more expensive.

To explain this effect, we might take a closer look at the evidence. Some of our results (Regression 1D) seem to support the theory that the price of land may increase near the city center. Regression 1D illustrates this possibility due because it indicates that low-density zoned property near high-density apartment buildings seems to reduce the value of these high-density structures. The authors may conclude that a greater increase in commercially zoned properties would result in a higher land value for these lots. This implies that it will cost low-income workers less to commute and that they have the convenience of close proximity to shopping areas. We would expect the demand for low-density, high-income residential land to decrease near the city center due to the nuisances that come with downtown living.

Possible Future Investigation:

The results presented from our investigation thus far are preliminary and derive from much simpler models than would likely be desirable for a full empirical analysis of the spillover effects public land-use policies like Oregon’s Ballot Measure 37. However, we believe that with slightly stronger and more specific modeling that follows the general formats our regressions have taken thus far, spillover effects are likely and ultimately calculable. Potential additions to our model could come in the form of controlling for elevation, geographic location and proximity to resources and desirable areas, or an estimation of land value as a function of property value as a whole, controlling for housing and other aesthetic characteristics provided by the GIS and RLID databases. In other words, land value might visibly be a function of characteristics like view and natural amenities, in addition to the potential influence of a given property’s zoning.

It may also be beneficial to run regressions in the future which incorporate the effects on commercial or industrial zoned properties. The authors believe that these regressions could contribute to adequately explaining whether living near the city center has a positive effect on high-density housing. This could tell us if the pros of living near the city are that one would be at work faster. However, for lower-income household, do these pros outweigh the larger crime rates, noise, pollution, lack of space?

It is possible that such a study could be of substantial use in Measure 37 claims, as an accurate regression of proximity zoning changes within a chosen area of a given observation could accurately (or at least somewhat satisfactorily) estimate the magnitude of the negative externalities that a Measure 37 claim could cause. Through this estimation officials could perhaps increase or decrease the magnitude of a “fair market value” determination based solely on appraisals to also reflect spillover land value costs in the context of that claimant’s larger community. Therefore the final compensable value that a claimant can receive could decrease significantly if it is determined by our model that substantial community negative externalities could occur. According to our preliminary models, this could occur if a claimant wished to change their low-density zoned residential land to high-density apartment buildings or subdivisions in a rural community. Our group is encouraged by the level of analysis possible from the databases we created for this project, and is very interested in deeper investigation of the spillover effects zoning may have on Measure 37 determinations of “fair market value” in the future.

[pic]

[pic]

[pic]

[pic][pic][pic]

Table A7: Regressed Variables

|Variable |Mean |Standard Deviation|Min |Max |

|acres |1.64 |11.25 |.0003 |647.25 |

|llandval |10.26 |.96 |2.30 |17.36 |

|lacres |-1.26 |1.26 |-8.11 |6.47 |

|percentr1a8th |.50 |.46 |0.00 |1.00 |

|percentgdonut4th |.51 |.42 |0.00 |1.00 |

|percentgood8th |.55 |.45 |0.00 |1.00 |

Table A8: Zone Dummy Variable Statistics (No Restrictions)

|Variable |Acres |Land value |Number of Properties within 4th Mile |

  |Mean |SD |Min |Max |Mean |SD |Min |Max |Mean |SD |Min |Max | |R1A |0.65 |5.20 |0.00 |409.87 |43586.81 |209453.70 |0.00 |34500000.00 |304.20 |95.98 |1.00 |544.00 | |R2 |0.71 |5.68 |0.00 |258.37 |43539.32 |153575.60 |0.00 |5296950.00 |292.80 |112.65 |2.00 |548.00 | |R3 |0.55 |2.49 |0.00 |47.28 |58168.06 |306569.60 |0.00 |9817070.00 |306.53 |101.63 |7.00 |526.00 | |R4 |0.75 |5.36 |0.01 |86.48 |52428.39 |88911.88 |0.00 |1073520.00 |313.15 |92.61 |2.00 |495.00 | |

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

[1] OAR statue 660-33-135 can be viewed at: .

[2] Ballot Measure 37, full text, Section (6).

[3] Department of Land Conservation and Development. Measure 37 Staff Report and Recommendation of May 3, 2005 found at .

[4] Ibid, page 5.

[5] Ibid, page 5.

[6] Ibid, page 6.

[7] O’Sullivan, Arthur. Urban Economics. New York: McGraw-Hill, 2003

[8] Vargas, Sylvia. “Florida’s Property Rights Act: A Political Quick Fix Results in a Mixed Bag of Tricks.” Florida State University Law Journal, 1996.

[9] Ibid

[10] Powell, David. “An Introduction to Florida’s Landmark Law Protecting Private Property Rights.” Hopping Green Sams & Smith, P.A. Tallahassee, Florida.

[11] Vargas, Sylvia. 1996.

[12] Barringer, Felicity. “Property Rights Law May Alter Oregon Landscape.” The New York Times, November 26, 2004.

[13] Netusil, Noelwah. “The Effect of Environmental Zoning and Amenities on Property Values: Portland, Oregon.” Forthcoming in Land Economics.

[14] Found at .

[15] O’Sullivan, Arthur. Urban Economics. New York: McGraw-Hill, 2003

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download