Economic Evaluation of Connecticut Lakes With Alternative Water Quality Levels

By Kara J. Fishman, Robert L. Leonard and Farhed A. Shah University of Connecticut College of Agriculture and Natural Resources Department of Agricultural and Resource Economics

Connecticut Department of Environmental Protection

Bureau of Water Management Lakes Management Program

79 Elm St., Hartford, CT 06106-5127 Arthur J. Rocque, Jr. Commissioner

1998

Pursuant to Section 314 of the Federal Clean Water Act of 1987

Table of Contents

Introduction 3 II. Methodology and Literature Review 4 Valuation Questions 9 IV. Study Areas and Data Collection 13 V. Survey Results 17 VI. Possible Bias in Data 32 VII. Aggregate Benefit Estimation and Analysis. 33 VIII. Impact of Changes in Water Quality Levels on the Incidence of Real Estate Taxes Collected by the Town 50 Conclusion 52 References 57 Glossary. 59

List of Tables 18 Table 1. Mean Values by lake and Weighted Average. Table 2. Determinants of EstLrnated Property Value with Alternative Water Quality Categories 20 Determinants of Estimated Property Value with Pooled Data 21 Table 3. 24 Table 4. Public Site Mean Values by Lake and Weighted Average Table 5. Determinants of Willingness to Pay at Public Sites with Alternative Water Quality Categnfies 26 Table 6. Determinants of Willingness to Pay at Public Sites by Lake, with Data Pooled Across Water Quality Categories 27 Table 7. Determinants of Willingness to Pay at Public Sites with Data Pooled Across lakes and Water Quality Categories 31 Table 8. Comparison of Property Owner’s Estimated Values to Sale Prices and Assessed Values for Improved Waterfront Properties 35 Table 9. Estimated Aggregate Values of Residential Waterfront Property with Alternative Water Quality Categories 35 Table 10. Calculation of the Total Willingness to Pay for Public Site Users of Bashan Lake 38 Table 11. Calculation of the Total Willingness to Pay for Public Site Users of Coventry Lake 39 Table 12. Calculation of the Total Willingness to Pay for Public Site Users of Crystal Lake 41 Table 13. Calculation of the Total Willingness to Pay for Public Site Users of Highland Lake 42 Table 14. Recreational Use Value of Public Sites with Alternative Water Quality Categories 44 Table 15. Estimated Annual Willingness to Pay for Residential Waterfront Properties with Alternative Water Quality Categories 47 Table 16. Total Annual Willingness to Pay by Waterfront Property Owners and Public Site Users 47 Table 17. Annual Loss in Social Welfare by a Shift From Water Quality A to Alternative Water Quality Categories 48 1 Table 18. Comparison of Percentage Changes in Value for Waterfront Property Owners and Public Site Users 49 Table 19. Excess Taxation and Potential Impacts on Average Property Value 53 Table 20. Annual Tax Revenues from Waterfront Properties with Alternative Water Quality Categories 54

2 I. Introduction

Recreational activities by the general public, such as boating, fishing, jet-skiing, sun- bathing and other beach uses, may contribute importantly to the valuation of lake water quality changes. However, lakes in suburban areas are often characterized by a relatively dense level of residential development in the immediate vicinity. This suggests that much of the impact of any change in the water quality of such lakes may accrue to adjoining properties. Not surprisingly, a property value approach has frequently been used to value water quality changes in cases of this type. In the absence of historical information on water quality variation for a particular lake, researchers often rely on cross-sectional data involving several lakes with differing characteristics. This paper presents an alternative approach in which a modified form of the contingent valuation method is used to capture the potential impact of water quality changes on property values for a given lake. The potential impact of such changes on recreational values accruing to public site users may be estimated using the conventional contingent valuation method. We argue that, if the valuation questions are phrased carefully, this exercise may yield information that can provide an interesting analysis of the overall use value of water quality changes in suburban lakes. We applied our proposed valuation approach to a sample of lakes in Connecticut. Each lake in the sample has significant public access as well as a highly developed waterfront. During the fall of 1995 and the winter of 1995-96 we surveyed waterfront property owners at these lakes and asked them to value their properties under four different scenarios regarding lake water quality. In the summer of 1995 public site users at the lakes were also surveyed with a comparable vfluation question. This paper provides a discussion and statistical analysis of the responses from the two surveys. Aside from the academic interest that the novelty of our approach is likely to generate, we believe that the results reported in this paper may be of value to state regulators and policy makers. Connecticut lakes with public access, such as the ones we have studied, currently receive public support for water quality protection and restoration. In an 3 era of budgetary limitations, our effort may prove worthwhile by leading to a better targeting of public spending and programs.

II. Methodology and Literature Review

Following standard practice, the economic value of a lake water quality change may be divided into use and non-use value components (Freeman, 1993). The use component includes recreational and amenity values that result from the direct consumption of the water resource.

Examples are swimming, fishing and the aesthetics one enjoys by living on or visiting the lake. The non-nse component of value is the enjoyment or "good feeling" one receives from knowing a resource is protected from pollution, that a natural habitat is maintained, and that generations to follow will enjoy the resource. Non-use value is often referred to as existence value and bequest value. The surveys administered to waterfront property owners and public site users in this study only asked about use value. Property price information reflects use value because housing is a marketed good. Implicit in the purchase price of a residence is the value of the bundle of benefits one enjoys by living at that specific location. Individuals who choose to live close to a lake have easier and more frequent access to it than persons who do not live near the lake. The use values at public sites were estimated with a contingent valuation survey. A water quality induced change in property prices is likely to capture the entire change in use value only if there is no public access to the lake, there are no impediments to buying and selling of property (i.e., zero transaction costs), and there are no property taxes. One would expect that property prices are higher for waterfront properties than non- waterfront properties, assuming that the water quality is relatively clean. This idea is the basis for attempting to measure the value of a lake by using property price data (e.g., Darling, 1973). The value of water quality changes at a lake may be inferred using the same idea. However, properties are not uniform in the various additional attributes which influence price (such as lot (1966), Griliches (1971), and Rosen (1974), hedonic price models have been developed to infer 4 the implicit price of water quality while:controlling for other attributes. A hedonic price function, which is an equilibrium relationship between property values and various other attributes of relevance, is first estimated. The partial derivative of this function with respect to water quality gives the implicit price of water quality. However, this approach is valid only if water quality changes are marginal. It should also be noted that the validity of the hedonic price approach has been questioned regardless of the extent of water quality changes (Maler, 1977, Kanemoto, 1988). The objections raised are conceptual as well as empirical in nature. Nonetheless, these critiques have failed to prevent applied researchers from continuing to employ the hedonic property value method. Most applications of the hedonic price method to water quality are based on cross-section data from a relatively large number of lakes (or other bodies of water) with different levels of water quality (e.g., David, 1968; Young and Teti, 1984; Michael, Boyle, and Bouchard, 1996). If one wanted to focus on a particular lake, time-series data involving significant variation in perceived water quality would be needed. We found only two studies that included both time- series and cross-section data (Falcke, 1982; Michael, Boyle, Bouchard, 1996). In this paper, we develop a straightforward alternative approach that may be used when neither type of data is available. Basically, we ask property-owners at given lakes to provide us their estimate of the market value of their property and how it would change with each hypothetical change in water

In order to further examine the implications of our approach, consider the following simplified scenario. Let there be an absence of taxes and a fixed supply of residential properties around a lake. Assume that the demand for properties is a function of property price, P, characteristics of the properties, ~ socioeconomic variables Y I and lake water quality, W. Figure 1 illustrates this function for two values of W and given values of "~and ~ The equilibrium value of P decreases with a lowering of water quality from W1 to W2, and if the demand curves are linear in P and W, then the area pipZ AB captures the change in social welfare This simple partial equilibrium framework seems tO underlie the very basic applications of the property value method in which the differential P~P: conveys all of the useful information and the demand curve need not be estimated. Typically, the researcher collects data on property prices and then performs an aggregation exercise using a formula such as ASW = :~ PPiNi, where

ASW is change in social welfare, ’i’ is a characteristic category (e.g., location or size), N~ is the number of properties in that category and PE is the average property price differential in the category.

Figure 1.

N = No. of Properties

A practical application of this concept should also account for the influence of tax rates and policies on the market price of residential property. A buyer’s total willingness to pay includes both the purchase price and assumption of responsibility for the annual property tax. The authority to levy an annual property tax is essentially the same as an equity interest in real 6 estate and should not be ignored when estimating the differential in social welfare associated with alternative levels of environmental quality (Niskanen and Hanke, 1977). Failure to account for property taxation will result in an underestimate of the differential in social value. Meanwhile, several provisions in the Federal income tax code favor home ownership and thereby inflate market value relative to actual willingness to pay. Failure to account for these favorable income tax provisions will result in an overestimate of the differential in social value associated with alternative levels of water quality (Freeman, 1993; Sonstelie and Pormey, 1980). A more specific discussion of the tax implications is included in a subsequent section on aggregate benefit estimation and analysis. If water quality changes are not observed, and property owners are asked to estimate the impacts of hypothetical water quality changes on their property, then the estimated property price differential PP’ may be biased. In other words, PP’ = PP~ + ei. where PP~ is the actual price

differential, e, is the error term and E(e) 0. Note that by interviewing property owners with regard to hypothetical changes in water quality, one would be using a variation of the contingent valuation method (Mitchell and Carson, 1989). However, by asking respondents to estimate changes in market value, rather than maximum willingness to pay (or minimum willingness to accept compensation) for a water quality change, one obtains a data set that is appropriate for use in a property value study. Furthermore, this form of the valuation question forces respondents to think in terms of a market good (i.e., a house) with which they have experience and may, therefore, provide answers that are less subject to information bias. The approach has the added advantage that public site users at the lake may be surveyed with a very similar contingent valuation question. For the sake of consistency in comparison, this question could be formulated in terms of a hypothetical user fee at different levels of water quality. As with any contingent Valuation survey, the alternatives evaluated should be relevant and understandable to both respondents and policy makers. Fortunately, water quality can be defined in relation to use categories without specification of the underlying technical parameters. A 7 water quality ladder is sometimes employed for this purpose. For example, Mitchell and Carson (1989) and Smith and Desvousges (!986) used a water quality ladder with the following categories: A. Sa~e to drink. B. Safe for swimming. C. Gamefish like bass can live in it. D. Okay for boating. E. Too polluted for any of these uses. Respondents were shown a visual aid with these categories at points along a ladder with rungs numbered from "0" for the worst possible water quality to "I0" for the best possible water quality. The category defmitiom included pictograms illustrating a nesting of uses with each category including the uses in lower water quality categories. In another example, Green and Tunstall (1991) defined the following water quality categories. Good enough:

A. For water birds (e.g. swans, coots, ducks, etc.) to use the water. B. To support many fish, including trout, and dragonflies; and to allow many different types of plants to grow both in the water and on the edge.

C. To be safe for children to paddle or swim. While the authors refer to use of a "... categorical scale of perceived water quality ...," the three categories appear to have been presented to respondents as independent options without an explicit nesting of uses. However, many respondents may have assumed category "C" to be inclusive of categories "A" and "B."

Concern about water quality at Connecticut lakes is generally with regard to swimming, fishing, boating and aesthetics. Suitability for one of these uses does not necessarily imply suitability for the other. Restrictions on swimming may be due to bacterial contamination, algal blooms and excessive weed growth, while restrictions on eating of fish are usually due to 8 contaminates that accumulate over several stages of the food chain. Lakes with substantial recreational use are not sources of public water supply. Thus, the concern is primarily about the independent and combined loss of water suitability for swimming, eating of fish caught, boating and aesthetics. To our knowledge, there are no examples of the use of the contingent valuation method to value water quality changes of New England lakes. A repeated discrete choice modelI was developed to estimate recreational swimming benefits of New Hampshire lakes (Needelman and Kealy, 1995). Young and Teti (1984) used a property value method to estimate the damage of water pollution on waterfront property values on the St. Albans Bay, lake Champlain, Vermont. Willis, Foster and Sewall (1983) used a property value method to determine if pollution

abatement activities on the Housatonic River in Massachusetts and Vermont affected the value of homes on or near the river. Michael, Boyle and Bouchard (1996) also used a hadonic property price model to estimate the effect of changes in water clarity on the market value of waterfront properties on Maine lakes. A recent study estimated the welfare effects of toxic contamination in freshwater fish in New York using a repeated discrete choice model (Montgomery and

Needelman, 1997).

III. Valuation Questions

The valuation questiom asked of waterfrom property owners and public site users were similar in the description of changes in water quality. The changes in water quality were described as changes in the fitness of the water for contact and noncontact recreation.2 The

t A repeated discrete choice model is a permutation of the hedonic and travel cost models. Recreational users are surveyed on their recreationa! activities and choice of site. A site is seen as an assemblage of many property characteristics. The marginal value of each characteristic is its bedonic price. The value of a site is a function of the values held by survey respondents for each of these characteristics, and the Ixavel costs incurred to get to the site. 2 Contact recreation is when the activity involves getting wet in the lake and possibly ingesting the lake water. Swimming is a water contact sport, and fishing usually is not a water contact sport. Boating can be categorized as contact or noncontact recreation, depending un the activities undertaken while on board. Examples of noncontact recreation are having a picnic on the beach, walking the dog, and jogging around the lake. 9 survey did not ask for nonuse values. By explicitly making housing a marketed good, the scope of the question in the waterfront property owner survey was narrowed so as to exclude nonuse values. Similarly, public site users were asked the valuation question in terms of a user fee. The survey question is repeated on the following page.

10 The following question was included in the surveys mailed to waterfront property owners at three lakes in Connecticut, Bashan, Crystal, and Highland? As we mentioned earlier, this survey is collecting information on the values waterfront property owners have for recreation and water quality at Connecticut public lakes. In an effort to determine these values, we are asking you to estimate the market value of your property with existing and alternative levels of water quality. All waterfront property owners on Lake are potential sellers of their property. In answering the questions which follow, please assume that you are willing but not anxious to sell. Also, keep in mind the fact that market values are determined by the interaction between buyers and sellers.

Water quality at Lake is now considered to be appropriate for swimming, fishing, eating of fish caught, and boating. These water quality- related uses are shown as available in Category A. Hypothetical losses of current uses are indicated in Categories B, C and D.

We first ask you to estimate the current market value of your property given the current water quality, Category A. We then ask you for a percentage loss in market value for your property given changes in water quality. Please do not skip this section due to uncertainty about the market value of your property. We are particularly interested in how changes in water quality affect your estimates of market value. In the spaces indicated below, please write your estimate of the market value of your property at Lake with existing and alternative levels of water quality.

3 In a test survey of waterfront property owners on a fourth lake, Coventry Lake, the valuation question positioned the property owner as a potential buyer. A substantial number respondents declared a $0 willingness to pay when faced with hypothetical reductions in water quality. In retrospect, these responses are not surprising because we asked property owners how much they would pay if they were in the market for a house. Logically, if a lake were polluted, many buyers would not be in the market for a lakefront house. These respondents had been asked to assume an uuaeeeptable choice set. Willingness to pay responses are typically solicited for non-marketed goods. There are no substitutes for clean air, an endangered species, or world peace. Housing, on the other hand, is a marketed good that is not fixed in supply; there are ample substitutes in most housing markets. The survey was revised and the new format used for the Bashan, Crystal, and Highland surveys. 11 WATER QUALITY CATEGORY A WATER QUALITY CATEGORY B [Existing] Swimmahle Fishable (edible) Fishable (edible) Boatable Boatable

$ Current Market Value % Loss from Category A

WATER QUALITY CATEGORY C WATER QUALITY CATEGORY D

Swimmable Fishable~ Fishable~ Boatable Boatable

% Loss from Category A % Loss from Category A

The following question was asked of public site users at four lakes in Connecticut, Bashan, Coventry, Crystal and Highland: Suppose {lake name}, along with other Connecticut lakes, became available only to households that paid an annual fee for the use of the lake. This fee would be in addition to entrance and parking fees you may already be paying. What is the maximum amount your household would pay, per year, for the right to use {lake name}? In answering the question, please consider you household’s income and expenses.

If the respondent answered with a positive figure, the following statement was read: One of the two major purposes of this survey is to find out people’s value of water quality. There are three general recreational activities at public lakes: swimming, fishing, and boating. Currently at {lake name} you can swim, fish, and eat the fish you catch.

At this point the interviewer would show the respondent a water quality table similar to the one shown to the waterfront property owners. The survey continues:

For Category A you stated a willingness-to-pay of $ {fill in with WTP answer for Category A.} In Category B, water quality in {lake name}, and in other 12 Connecticut Lakes, deteriorates to the point where swimming is no longer advisable but you could fish, eat the fish you catch and boat. What is the maximum amount your household would pay, per year, for the right to use {lake name}?

The interviewer then wrote the response on the table, and continued showing it to the respondent as similar payment questions were asked for Categories C and D. By seeing the table as it was filled in respondents were able to visualize the values for different water qualities as they relate to one another.4 If the respondent answered with a $0 bid or would not answer the question, they were asked why they said $0. The interviewer would check off one or more of the following:

Why did you say $0? Objects to charges Already pays for park through taxes Cannot afford charges. Other

IV. Study Areas and Data Collection The four Connecticut lakes included in this study have several common characteristics, many of which are also shared with other Connecticut lakes of similar size. The lakes have a wide variety of recreational uses at privately owned waterfront properties and at public sites. Water quality in all four lakes is considered safe for swimming, boating, fishing and eating of fish caught. However, maintenance of this quality has involved a continuing effort to control non-point sources of pollution and has included installation of sanitary sewers in the more

4 The visual aid of the table on a clipboard may be an implied value cue, a form of systemic error. According to Freeman (1993), when an individual is faced with an unfamiliar choice problem and does not have clear preferences, he might seek clues regarding the "correct" choice or value from the information supplied as part of the scenario specification, thus biasing the answer. The implied value cue appears as individuals rank their preferences for water quality based on their recreational activities. The order of the boxes on the table did not appear to influanee the respondant’s preferences. Water quality C was not necessarily worse than water quality B, particularly if the household’s primary activity at the lake was swimming. 13 densely populated areas. All of the lakes have residential developmem in waterfront areas at a density of four to eight times that in the rest of the town. Beyond the lake area, these towns have a rural/suburban level of development. Two of the lakes, Crystal and Higtdand, are managed by the state to be trout fisheries. Bashan Lake in East Haddam covers 276 acres and is a 45 minute drive southeast of Hartford, Connecticut’s capital. The lake is classified as oligotrophic. A few of the homes use water from the lake for non-drinking, residential uses. The only public access is a remote and unimproved state boat launch with limited parking. There are no sanitary sewers around the lake. More than 50% of the homes are second or seasonal homes. The waterfront properties include Wildwood a thirty-home seasonal development on 145 acres of cooperatively owned land. Excluding the cooperative, there are 134 properties with direct water frontage and 26 with deeded rights-of-way to the lake. Coventry Lake in Coventry covers 378 acres and is a 30 minute drive east of Hartford. The lakeis classified as early mesotrophic. The lakefront was developed in the 1920s - 1940s with summer cottage colonies. Conversion of the summer cottages into year-round residences began in earnest after World War II. The lots are small, often 5,000 - 10,000 square feet, and the density of residences is 2 - 8 per acre. There are about 200 waterfront properties. Approximately ten lake associations collect taxes from waterfront and non-waterfront property owners in their districts. The taxes go to maintain local roads and any association facilities such as a private beach. If an association has its own beach, non-waterfront property owners have deeded access to the lake over the beach. About 20% of the homes in the lake area are connected to public sanitary sewer and public or community water. Sewer installation occurred in the early 1990s, prior to the study period. Coventry lake has three public access points: a state boat launch that can accommodate 40 cars with trailers; Patriot’s Park, a beach developed under the Land and Water Conservation Fund, is open to all residents and non- residents; and Lisicke Beach, for town residents only. There is a parking fee at the boat launch on summer weekends for all users. There is a parking fee at Patriot’s Park on summer weekends only for non-residents. Crystal Lake in Ellington and Stafford covers 200 acres and is a 30 minute drive northeast of Hartford. This lake is classified as mesotrophic. There is a state boat launch with limited parking, and a beach for town residents only. Sanitary sewers were installed around

14 the entire lake in 1991. Most homes are oWner-occupied year-round. There are approximately 82 properties with direct water frontage. Highland Lake in Winchester covers 444 acres and is located in the rural northwestern part of the state. The trophic classification of the lake water is oligotrophlc. There is a state boat launch with ample parking and two small town beaches for town residents only. Parking at the launch is free of charge, except on weekends. The town is actively involved in lake issues through the Highland Lake Commission. Sanitary sewers were installed around the entire lake in 1994. There are approximately 540 land parcels in the lake area of which more than one-half have direct water frontage.

Data Collection from Waterfront Property Owners. In the fall of 1995 and the winter of 1995- 96 mail surveys were sent to waterfront property owners and property owners with deeded righls-of-way to the lake? Only those properties with a house were included in the study. Mailing addresses were obtained from the tax assessor. Surveys were sent from the University of Connecticut and included a cover letter from the first selectman or town manager. The initial mailing was followed by a combination "thank you" and "reminder" letter with an offer of a replacement questionnaire upon request. The surveys were anonymous and did not ask for information about the size or physical characteristics of the house. Test surveys had indicated that property owners were hesitant to answer questions that could identify their property. The survey included questions about use of the property, lot size and features, recreational activities, and household income level. The initial mailing and subsequent reminder to 699 property owners at the four lakes resulted in 340 replies with 237 sufficiently complete for use in the analysis. The reply rate was 48.6 percent, while the effective response rate was only 33.9 percent. This difference is due to use of only those responses that provided estimated property value for each of the four water quality categories.

~ Assessors’ records and conversations with Realtors indicated a substantial price premium for waterfront locations. In the absanee of either lake frontage or a deeded fight-of-way to the lake there is no clear evidence of a relationship between property value and distance from the lake. This finding coincides with the results of a bedohic study done hivolvhig the Honsatonic River in Massachusetts and Vermont where distance from the fiver was found not to be significant in predicting property values (Willis, Foster and Sewall, 1983). However, in other studies distance from die waterbody was a significant determinant of property value (Darling, 1973; Brown and Pollakowski, 1977; Lansford and Jones, 1995). 15 Missing values for independent variables were set at the mean or median value for the corresponding lake. Thirty-one missing values for lot size were set at the mean value. Twenty-one missing values for income and a few missing values for other independent variables were set at the median level, A proxy for lot size at the Wildwood cooperative was generated by multiplying the house site lake frontage by an assumed lot depth of 200 feet.

Data Collection for Public Site Users. A public site user is defined as a person who arrives at a public access point by car, motorcycle, bicycle, or foot. Frequent lake users often live close to the lake. Their travel costs are nominal. Other public site users include town residents and nonresidents. Interviews were conducted in the summer of 1995. Individuals at the public sites were approached randomly. Most of the time all of the users at a site could be surveyed over a 5 - 10 hour period. The user population at a public access point is generally orderly. For instance, one person/household will launch a boat and is surveyed. They go off to boat, and 10 minutes later, another boater comes in. Similarly, on a beach, a person/household arrives, sets up, and stays for several hours. By working their way from one end of the beach to the other, an interviewer can survey almost all households on the beach. New arrivals are surveyed in the next round, and so on. The sampling was without replacement in that respondents returning at a later date were not interviewed a second time. Many of the survey questions were the same asked of the waterfront property owners. Other questions included frequency of visits, the number of household members using the lake, the length of visit, driving distance to lake, and substitute sites. The respondents were asked how much, if anything, they were currently paying to use the access point. The bulk of respondents with an affirmative answer to this question paid for a seasonal pass or parking. Other people mentioned their fishing license and boat registration fees, and a federal tax of 10% on hunting and fishing equipment, which are not related to use of a particular lake and are not included in our survey results.

16 Vo Survey Results

A. Results of Waterfront Property Owners Survey Mean values of variables by lake and averages weighted by Sample size are shown in Table 1. Property value estimates at each lake indicate a greater concern about loss of swimming than loss of fish edibility. The three-lake weighted averages of estimated property value are: $183,535 with swimmable water quality and edible fish; $118,452 without swimmable water quality; $148,303 without edible fish; and $105,510 with neither swimmable water quality nor edible fish. The relationships between estimated property value and water quality category are consistent among lakes. This consistency is especially apparent when attention is focused on percentage losses in value associated with shifts from water quality Category A (current water quality) to each of the other water quality categories. These percentage relationships are almost identical for Bashan and Highland, which are also similar in other respects. Differences between Crystal and the other two appear to be related to differences in household income and type and use of the property. Crystal has the lowest current average property value along with highest percentage of year-round residential use by the property owner (75 % at Crystal as compared to 33% at Bashan and 36% at Highland). This otherwise surprising combination is consistent with the relative income distributions. As shown in Table 1, Crystal respondents reported a substantially higher percentage of households with an annual income of less than $40,000 and a much smaller percentage with income above $100,000. The percentage decline in estimated property value from a change in water quality from Category A (current water quality) to Category D (swimming and eating of fish caught are not advisable) is the smallest at Crystal. This is not surprising in that a decline in water quality seems likely to have a relatively smaller impact on the value of a year-round residence than on the value of a seasonal property. However, the percentage decrease in value due to a shift from water quality Category A to Category C (eating of fish caught is not advisable) is the largest at Crystal. This concern about fish edibility is consistent with the relatively high percentage of households who reported fishing at Crystal and may be related to the lower income level. All replies sufficiently complete for use in the statistical analysis reported some engagement in swimming-related activities, defined in the survey to include swimming,

17 Table 1. Mean Values by Lake and Weighted Average

Weighted Variable Bashan C~sml ~ . average Property Value by Water Quality:

A - Swimmable and Fishable $178,880 $168,500 $189,060 $183,535 (edible) B - ~w~’:’.~’r.a~le and Fishable 115,410 115,160 120,580 118,452 (edible) C - Swimmable and Fishable 146,530 132,560 152,680 148,303 (~) D - S’:’.’:r~,~,~b!e and Fishable 102,110 102,310 107,780 105,510

Percentage Loss in Value by a Shift From Water Quality A to: Water Quality B 35.48 31.66 36.22 35.46 Water Quality C 18.08 21.33 19.24 19.20 Water Quality D 42.92 39.28 42.99 42.51 Year-round Residence (%) 32.81 75.00 36.17 40.51

Lot Size (1000 sq. ft.) 36.17 25.39 25.39 28.30 Dock (%) 73.44 90.62 91.49 86.50 Sandy Beach (%) 34.38 40.62 23.40 28.69 Activities: Swimming &/or sunbathing (%) 100.00 100.00 100.00 100.00 Boating (%) 89.06 96.88 89.36 90.29 Fishing (%) 73.44 90.62 77.30 78.06 General Outdoor Activities (%) 87.50 84.38 90.78 89.03 Winter Activities (%) 45.31 78.12 51.06 53.16 Income Category (%) A - Less than $20,000 1.57 6.24 4.96 4.22 B - $20,000 - $39,999 15.62 34.38 8.51 13.92 C - $40,000 - $59,999 25.00 25.00 22.70 23.63 D - $60,000 - $79,999 23.44 18.75 27.66 25.32 E - $80,000 - $99,999 15.62 9.38 9.93 11.39 F - More than $100,000 18.75 6.25 26.24 21.52

Sample Size 64 32 141 237

18 sunbathing and swimming lessons. The prevalence of swimming-related activities over fishing is consistent with the estimated property values with alternative water quality categories. However, the universality of swimming precludes the use of swimming in statistical analysis of the data. Estimated relationships between property value and related independent variables are reported in Tables 2 and 3. Table 2, with data for all three lakes, shows the results of a separate linear regression for each water quality category. Estimated property value with each water quality category is a separate dependent variable. The dependent variable is in thousands of dollars in all cases. L-LOTSIZE is the natural log of lot size in thousands of square feet. All of the other independent variables are "0,1" dummies with "1, for presence of the characteristic. CRYSTAL and HIGHLAND are included to account for property value variation among the three lakes. Variables regarding the respondent’s property and household correspond to those in Table 1 except for the absence of swimming-related, general outdoor activities, and winter activities. Property values are more closely related to type of residential use and to lot size and features than to recreational activities. As shown in Table 2, the coefficients for Y-R- RESIDENCE and L-LOTSIZE are positive and significantly different from zero at the 0.01 level for each of the four water quality categories. The coefficients for DOCK and S-BEACH in Table 2 are of particular interest. As might be expected, the coefficients of these variables are smaller in the absence of swimmable water quality. The coefficients for DOCK are 49.551 and 47.012 with water qualities A (current water quality) and C (eating of fish caught is not advisable) as compared to 31.812 and 30.111 with water qualities B (swimming is not advisable) and D (swimming and eating of fish caught are not advisable). The two larger values are significant at the 0.01 level, while the two smaller values are significant at the 0.05 level. In a similar manner, the coefficients for S- BEACH are 27.212 and 28.230 with water qualities A and C as compared to 14.602 and 15.427 with water qualities B and D. The larger two values are significant at the 0.01 level, while the smaller two values are significant only at a 0.10 level. In summary, the survey

19 Table 2. Determinants of Estimated Property Value with Alternative Water Quality Categories

Water Quality Category A B C D Variable Coefficient Coefficient Coefficient Coefficient (T-Ratio) (T-Ratio) (T-Ratio) (T-Ratio) CRYSTAL -22.117 -9.647 -21.757 -9.885 (-1.422) (-0.703) (- 1.478) (-0.740) HIGHLAND 11.840 8.436 7.508 6.820 (1.087) (0.878) (0.728) (0.729) Y-R-RESIDENCE 55.191 50.481 49.833 49.386 (5.915)** (6.136)** (5.644)** (6.165)** L-LOTSIZE 24.503 19.769 21.164 16.021 (5.005)** (4.579)** (4.568)** (3.811)** DOCK 49.551 31.812 47.012 30.111 (3.517)** (2.560)* (3.526)** (2.489)* S-BEACH 27.212 14.602 28.230 15.427 (2.803)** (1.706) (3.073)** (1.851) BOATING 11.113 -5.265 -6.188 -11.860 (0.687) (-0.369) (-0.404) (-0.854) FISHING 2.511 - 1.831 -4.025 -3.259 (0.229) (-0.190) (-0.388) (-0.346) INCOME-B -15.751 -0.939 -1.967 -2.242 (-0.644) (-0.044) (-0.085) (-0.107) INCOME-C -2.105 7.547 10.712 5.641 (-0.092) (0.373) (0.494) (0.287)

INCOME-D 4.956 8.589 15.812 9.886 (0.217) (0.427) (0.732) (0.505) INCOME-E 17.228 24.159 30.676 17.288 (0.689) (1.095) (1.296) (0.805) INCOME-F 55.708 47.333 69.233 45.573 (2.385)* (2.298)* (3.133)** (2.273)* CONSTANT 13.163 -3.492 2.497 5.295 (0.404) (-0.121) (0.081) (0.189)

SAMPLE SIZE 237 237 237 237 ADJUSTED R2 0.332 0.265 0.317 0.244

* Significant at the 0.05 level. ** Significant at the 0.01 level. 20 Table 3. Determinants of Estimated Property Value with Pooled Data

Individual Lakes Three Lakes Combined

Bashan Crystal Highland Model A Model B Coefficient Variable Coefficient Coefficient Coefficient Coefficient (T-Ratio) (T-Ratio) (T-Ratio) (T-Ratio) (T-Ratio)

omitted CRYSTAL N.A. N.A. N.A. -15.851 (-2.235)*

Omitted HIGHLAND N.A. N.A. N, A, 8.651 (1.741) -65.1)81 WQ-B -63.472 -53.344 -68,476 ,65.081 (-5.077)** (-5.533)** (-10.230)** (-11.710)** (-11.630)** -35.227 WQ-C .32.348 -35.938 -36.372 -35.227 (.2.587)* (-3.728)** (-5,432)** (-6.341)** (,6.297)** -78.024 WQ-D -76.770 -66.187 -81.279 -78,ff24 (-6.141)** (-6,865)** (-12.140)** (-14.040)** (-13.950)** 48.014 Y-R-RESIDENCE 70.307 52.09~ 50.063 51,223 (6.293)** (4.882)** (9.684)** (12.040)** (11.610)** 19.646 L-LOTSIZE 27.458 6.700 16.264 20.364 (4.965)** (1,454) (5.967)** (9.123)** (9,077)** 42.552 DOCK 34.529 44.796 33.488 39,621 (2,918)** (1.772) (3.608)** (6,167)** (6.751)** 19.340 S-BEACH 8.793 -1,003 28.841 21.368 (0.910) (-0.126) (4,955)** (4.828)** (4.373)** -7.864 BOATING 31.743 36.458 -4,846 -3.050 (1.629) (1.109) (-0.571) (-0.413) (-1.075) -3.225 FISHING -9.446 32.269 .4.425 -1,651 (..0.797) (1,854) (-0.737) (-0,330) (-0.643) 0.711 5,962 -5,225 -10,935 INCOME-B -80.357 (-0,986) (-1.905) (0,047) (0.430) (-0.469) 5.448 4.793 INCOME-(: ,68.720 44.866 0,333 (-1.659) (2.861)** (0.028) (0.521) (0.457) 15.390 13.327 9.811 10.703 INCOME-D -68.957 (I.ff23) (-1.657) (0.937) (1.143) (0.943) 21,433 ‘66.744 22,416 37.373 22.338 INCOME-E (1.874) (-1.524) (0.754) (2.832)** (1.958) 54,462 56.632 INCOME-F 24,260 51.547 39.275 (0.554) (2.589)* (3.352)** (5.114)** (5.296)** 69.558 -13.301 80,046 48,949 59,253 CONSTANT (3.208)** (4,098)** (1.752) (-0.379) (4.3~0)** 128 564 948 948 SAMPLE SEE 256 0.411 0.403 ADJUSTED R2 0.490 0,554 0.394

* Significant at the 0.05 level. ** Significant at the 0,01 level. 21 results indicate that the contributions of a dock and sandy beach to property value are dependent upon the existence of swimmable water quality.6 The signs of the coefficients for BOATING and FISHING in Table 2 are of some interest. The coefficients for BOATING and FISHING are positive with water quality A and negative with the other water quality categories. These signs are as might be expected; however, none of these coefficients is statistically significant at the 0.05 level. The relation- ship between income and property value is generally weak except for the positive effect of INCOME-F, which indicates a household income of $100,000 or more. Comments on the surveys indicate a substantial number of retired owners and long term ownership of seasonal properties. Thus, the relationship between current income and value of lake front property may be less than would be expected for more typical residential property. Table 3 shows the results of linear regressions that pool the four property value estimates by each respondent. The water quality effect is measured through the use of dummy variables WQ-B, WQ-C and WQ-D. Coefficients for these water quality variables are estimated differences, in thousands of dollars, between property value with water quality A and the alternative water qualifies. Variables regarding the respondent’s property and household are the same as Table 2. Table 3 shows estimates for each lake and for two models with the three lakes combined. The independent variables in Model A include CRYSTAL and HIGHLAND to account for property value variation among the three lakes. Model A is of interest if attention is focused only on the three lakes in the survey. Omission of CRYSTAL and HIGHLAND in Model B results in a more logical basis for generalizing results to other lakes with similar property and household characteristics. The coefficients of the water quality variables (WQ-B, WQ-C and WQ-D) are significantly different from zero at the 0.01 level in all cases except for the coefficient of WQ- C (eating.of fish caught is not advisable) at Bashan, which is significant at the 0.05 level. Coefficients of the water quality variables for individual lakes are the same as the respective differences between mean property values in Table 1. However, the coefficients of the water quality variables for the three lakes combined are not exactly the same as the respective differ-

6 In this context the "contributions" of a dock and sandy beach to property value are of a statistical nature. The incidence of these features is probably associated with favorable shore characteristics and may be correlated with house size and quality.

22 ences in the weighted average property valOes in Table 1. This divergence is due to variation among lakes in property and household characteristics. The coefficients of WQ-B, WQ-C and WQ-D are the same in models A and B. This common set of coefficients results from the fact that CRYSTAL and HIGHLAND are both orthogonal to WQ-B, WQ:C and WQ-D. However, omission of CRYSTAL and HIGHLAND in Model B does change the coefficients of the non- orthogonal property and household related independent variables (Greene, 1993, p. 247). Pooling of estimated property values across water quality category averages the results obtained with a separate regression for each water quality category. Averages across water quality categories of the coefficients for independent variables in Table 2 are equal to the corresponding coefficients in Model A, Table 3. This averaging effect is also inherent in results for Model B and for the individual lakes.

B. Results of Public Site Users Survey Mean values of variables by lake and averages weighted by sample size are shown in Table 4. Willingness to pay is the sum of existing annual user fees and stated annual willing- ness to pay in addition to existing fees. The four-lake weighted averages of willingness to pay for annual permits are $72.02 with swimmable water quality and edible fish; $28.78 without swimmable water quality; $42.13 without edible fish; and $15.24 with neither swimmable water quality nor edible fish. The willingness to pay estimates at each lake indicate a greater concern about loss of swimming than loss of fish edibility. With the e~ception of Bashan, there is greater than 50% drop in willingness to pay when the water is not swimmable. Reductions in willingness to pay from loss of fish edibility range from 37.77% - 46.55%. At Bashan the decrease in willingness to pay due to loss of fish edibility is almost as large as the decrease in willingness to pay due to loss of swimming. Bashan differs from the other lakes by the absence of a public beach and predominance of fishing over other public site recreational activities. The willingness to pay measures for each water quality at Highland are approximately twice those at Bashan. This difference is not attributed to income since the percentage of respondents in the three upper income categories is higher at Bashan than at Highland. The socio-economic data shows that Highland users are twice as likely as Bashan users to be town residents, make four times as many visits to the lake each year, and have 50% fewer substitute lake sites. Highland has the highest percentage of winter users and jetskiers. Winter users 23 Table 4. Public Site Mean Values by Lake and Weighted Average

Variable Bashan Coventr~ ~ Hi~ Weighted Average Willingness to Pay by Water Quality: (in dollars per year) A - Swimmable and Fishable $39.27 $79.66 $63.02 $93.03 $72.02 (edible) B - 8wimmabie and Fishable (edible) 21.80 27.88 21.97 39.63 28.78 C - Swimmable and Fishable (edible) 23.24 50.37 38.13 49.72 42.13 D - ,F~fimmabte and Fishable (edible) 10.43 13.52 10.90 23.44 15.24 Percentage Loss in Value by a Shift from Water Quality A to: Water Quality B 44.49 65.00 65.14 57.40 60.04 Water Quality C 40.82 36.77 39.50 46.55 41.50 Water Quality D 73.44 83.03 82.70 74.80 78.84

Town resident (%) 22.78 56.03 66.02 50.40 50.59 Travel time (minutes one way) 30.98 12.24 14.34 19.11 18.28 Visits per year 9.92 36.59 32.63 37.08 30.79 Hours per visit 4.82 3.39 3.45 3.51 3.71 Household size 2.34 3.12 2.69 2.47 2.68 User fees ($ per year) 0.34 12.76 17.90 10.94 11.15 Activities: Swimming and/or sunbathing (%) 29.11 80.17 61.16 64.00 61.23 Boating (%) 31.65 26.72 30.10 37.60 31.68 Jetskiing (%) 2.53 2.59 0.97 5.60 3.07 Fishing (%) 75.95 18.!0 45.63 34.40 40.42 General Outdoor Activities (%) 5.06 35.34 19.42 23.20 22.22 Winter Activities (%) 1.27 11.21 8.74 15.20 9.93 Alternative Sites: lakes (0 to 3) 1.95 0.90 1.07 1.01 1.17 Rivers (0 or 1) 0.41 0.13 0.16 0.18 0.20 Ocean (0 or 1) O. 13 0.23 0.29 O. 18 0.21 Income Category (%) A - Less than $20,000 7.59 14.66 6.80 13.60 11.11 B - $20,000 - $39,999 18.99 20.69 26.21 29.60 24.35 C - $40,000- $59,999 39.24 38.79 41.75 32.80 37.82 D - $60,000 - $79,999 21.52 17.24 12.62 15.20 16.31 E - More than $80,000 12.66 8.62 12.62 8.80 10.40

Sample Size 79 116 103 125 423

24 interviewed at public sites during the summer are essentially year-round users. Jetski owners can spend thousands of dollars t~ purchase a jetski. Their access to public lakes is very limited, however, and they are willing to pay an amount higher than other lake users to maintain that access. Highland public site users are apparently willing to pay more because they use the lake more. The willingness to pay estimates are the lowest for Bashan. Since more than 75 % of the Bashan public site users are not town residents, their travel costs are likely to be higher. Thus, if an annual fee were imposed they could choose to recreate elsewhere. Also correlated with the relatively low willingness to pay estimates at Bashan is the high percentage of fisherman, 75.95%, versus the 18.10% - 46.63% range for the other three lakes. At Coventry, Crystal and Highland the primary activity was swimming and/or sunbathing, at 80.17%, 61.16% and 64.00% respectively. The percent of public site users engaged in boating is similar across lakes, at 26.72% - 37.60%. Estimated relationships between willingness to pay and related independent variables are reported in Tables 5 and 6. Table 5 shows the results of a separate linear regression for each water quality category with the data pooled across lakes. The estimated willingness to pay with each water quality category is a separate dependent variable. The dependent variable is in dollars per year in all cases. TRAVEL-T, VISITS-Y, HOURS-V, and HOUSEHOLD-S are actual numbers indicated by respondents. The LAKES variable is 0, 1, 2 or 3 depending on the number of substitute lakes a respondent indicated. The number 3 was entered if the respondent mentioned more than three substitute lakes. All of the other ind.ependent variables are "0,1" dummies with "1" for presence of the characteristic. COVENTRY, CRYSTAL and HIGHLAND are included to account for willingness to pay variation among the four lakes. Bashan Lake and Income Category A are the base independent variables against which the alternative variables are measured. Variables .regarding the respondent’s household characteristics and recreational activities correspond to those in Table 4 except for the absence of user fees. In Table 5, the annual willingness to pay for each water quality category is more closely related to the recreational activity than to income, frequency of use, or household characteristics. JETSKIING is significant at the 0.01 level for water quality categories A (current water quality), B (swimming is not advisable) and C (eating of fish caught is not advisable), and is significant at the 0.05 level for water quality category D

25 Table 5. Determinants of Willingness to :Pay at Public Sites with Alternative Water Quality Categories

Water Quality Category A B D

Variable Coefficient (T-Ratio) Coefficient (T-Ratio) (T-Ratio) Coefficient (T-Ratio) COVENTRY 19.56 (1.00) 21.60 (1.90) 15.19 (1.06) 7.64 (0.89)

CRYSTAL 1.99 (0.11) 4.94 (0.45) 2.13 (0.16) 0.41 (0.05)

HIGHLAND 26.50 (1.46) 22.17 (2.09)* 5.89 (0.44) 9.29 (1.16)

TOWN-R 27.42 (1.87) 9.92 (1.16) 12,55 (1.17) 6.40 (0.99)

TRAVEL-T -0.05 (-0.10) 0.00 (0.01) 0.08 (0.22) 0.09 (0.39)

VISITS-Y 0.15 (1.14) 0.13 (1.77) 0.12 (1.25) 0.05 (0.89)

HOURS-V 3.83 (1.71) 0.76 (0.58) 1.26 (0.77) o.8~ (0.83)

HOUSEHOLD-S 2.75 (0.67) 0.31 (0.13) -1.61 (-0.54) -1.82 (-1.Ol)

SWIMMING -2.58 (-0.17) -20.01 (-2.25)* 1.25 (0.11) -lO.3O (-1.53)

BOATING 41.47 (3.21)** 16.12 (2.14)* 18,20 (1.92) 13.56 (2.38)*

IETSKIIHG 88.87 (2.62)** 56.56 (2.86)** 116.58 (4.71)** 38.54 (2.58)*

HSHING -10.75 (-0.74) !9.16 (2.26)* -6.39 (-0.60) 2.42 (0.38)

GENERAL-0 -2.68 (-0.18) -2.83 (-0.33) -11.49 (-1.08) -4.69 (-0.73)

WINTER 79.82 (3.68)** 24.86 (1.96) 67.13 (4.24)** 35.60 (3.72)** LAKES -6.81 (-1.45) -3.39 (-1.24) -3.72 (-1.09) -2.74 (-1.32)

RIVERS -6.11 (-0.41) 8.51 (0.97) -8.39 (-0.77) -1.80 (-0.27)

OCEAN -24.01 (-1.65) -1.80 (-0.21) -3.67 (-0.35) -5.55 (-0.87)

INCOME-B 22.38 (1.10) 12.78 (1.08) 6.50 (0.44) 8.05 (0.90)

INCOME-C 31.01 (1.61) 10.04 (0.89) 10.65 (0.76) 8.88 (1.04)

INCOME-D 21.24 (0.96) 9.85 (0.77) 8.79 (0.54) 5.73 (0.59)

INCOME-E 135.93 (5.56)** 43.35 (3.04)** 75.05 (4.20)** t0.92 (2.86)**

CONSTANT -18.41 (-0.53) -12.30 (-0.61) 4.29 (0.17) ,1.66 (-0.11)

SAMPLE SIZE 480 480 480 480 ADJUSTED R~ 0.208 0.123 0.183 0.105

* Significant at the 0.05 level. ** Significant at the 0.01 level.

26 Table 6. Determinants of Willingness to Pay at Public Sites by Lake, with Data Pooled Across Water Quality Categories

Bashan Coventry Crystal Highland

Variable Coefficient Coefficient Coefficient if-Ratio) if-Ratio)

WQ-B -17.47 -51.79 -53.40 (-4.05)** (-5.40)** (-6,74)** (-3.74)** WQ-C -16.02 -29.29 -24,89 -43.31 (-3.71)** (-3.06)** (-4,09)** (-3.03)** WQ-D -28.84 -66.15 -52,12 -69.60 (-6.68)** (-6.90)** (-8.56)** (-4.87)** TOWN-R -1.64 -3.53 12.12 42.94 (-0.5O) (-O.33) (I .48) (2.45)* TRAVEL-T -0.04 -0.44 0.12 0.54 (-0.35) (-0.8O) (0.34) (1.15) VISITS-Y 0.16 0.04 0.62 -0.15 (0.~) (0.64) (10.46)** (-1.18) HOURS-V 1 A0 0.94 2.58 5.67 (2.22)* (0.95) (1.46) (1,57) HOUSEH-S 1.84 1.21 -6.75 -5.63 (1.18) (O.5O) (-3.46)** (-1.36) SWIMlVl]NG 7.90 -43.44 16.14 -9.57 (I .81) (-3.73)** (2.31)* (-0.68) BOATING 9.14 7.37 17.25 158.61 (1.74) (0.79) 0.29)** (4.29)** IETSKIING -13.19 305.19 2.01 11.18 (-1.25) (11.74)** (O.O8) (O.44) FISHING -1.30 -27.80 6.33 -0.14 (-0.23) (-2.56)* (1.12) (-0.Ol) GENERAL-0 -26.73 -17.05 -3.84 10.36 (-3.30)** (-2.09)* (-0.61) (0.77) WINTER 63.33 44.92 15.52 88.17 (3.88)** (3,14)** (1.73) (4.78)** -4.19 0.88 1.84 -15.97 (-3,33)** (0.27) (0.93) (-2.94)** RIVERS 9.53 -45.68 -12.46 -2.70 (2.76)** (-3.69)** (-1.72) (-0.18) -30.34 OCEAN -0.68 13.63 -9.82 (-0.11) (i .~o) (-1.82) (-1.90) 24.56 INCOME-B -2.73 22.62 4,98 (-o,39) (1.81) (0.51) (1.41) INCOME-C 4.41 9.82 1.92 44.76 (0.71) (0.89) (0.20) (2.56)* 6.30 INCOME-D 12.00 31.88 14.37 (1.75) (2.35)* (1.34) (0.31) INCOME-E 5.11 58.31 30.80 138.95 (0.70) (3.67)** (2.70)** (5.64)** 24.92 92.58 18.97 15.29 CONSTANT (0,48) (2.31)* (4.24)** (I .26)

316 464 412 500 SAMPLE SIZE 0.478 0.212 ADJUSTED R2 0.297 0.426

* Significant at the 0.05 level. ** Significant at the 0.01 level.

27 (swimming and eating of fish caught are not advisable). As was evidenced in Table 4, jetskiiers, who have a limited lake choiceset, are willing to pay higher amounts than all other recreational users for access to a lake, regardless of water quality. The coefficient for SWIMMING is significant at the 0~05 level only for water quality B. As expected, the SWIMMING coefficient is negative for water qualities B and D. The main loss in use value for swimmers is felt when water quality deteriorates to the point where swimming is no longer advisable, Category B. Public site users who use the lake for winter activities are less sensitive to changes in water quality, basically because the water is frozen and water quality has a smaller influence on use. The WINTER coefficient is positive for all water qualities, and is significant at the 0.01 level for water qualities A, C and D. As might be expected, winter use has little influence on willingness to pay for swimmable water quality. A household’s income level was only significant in determining willingness to pay for the highest income group. The income coefficients for all water qualities are positive, as would be expected. Income Category E, the highest income group, is significantly different from Income Category A, the lowest income group, which indicates a household income of less than $20,000. Comments by respondents indicated that some high income households thought that the higher the fee the greater the exclusivity of recreational use. Several respondents mentioned user fees homeowners on private lakes pay for exclusive use of a lake and the maintenance of lake water quality. High income respondents may have motives of improving and/or maintaining the quality of the recreational experience by limiting public use. In Table 6 the.data is pooled across water quality for each lake. Water Quality A (current water quality) and Income Category A are the base case in determining willingness to pay. WQ-B (swimming is not advisable), WQ-C (eating of fish caught is not advisable) and WQ-D (swimming and eating of fish caught are not advisable) are significant independent variables at the 0.01 level for all lakes. The only income category that has a significant impact at the 0.01 level for three of the four lakes is INCOME-E. Independent variables significant at the 0.01 level for Bashan are GENERAL-O, WINTER, LAKES, and RIVERS. GENERAL-O is negatively correlated with willingness to pay because the public access point to Bashan is limited to water-related activities. At the other lakes one can walk around the lake, have a picnic or simply view the lake from one’s car. The boat launch at Bashan is situated in a small cove with essentially no view of the lake. The large positive coefficient for WINTER may

28 stem from the high percentage of public: site users that fish. The coefficient for LAKES is negative, indicating that if a household has substitute lake sites they would pay less for Bashan or choose to buy a permit at another lake instead of Bashan. The coefficient for RIVERS is positive, indicating that rivers are not a substitute for lakes. Since the majority of users at Bashan are fishermen, fishing on a lake is a different experience than fishing on a river. Significant at the 0.05 level at Bashan is the length of time per visit, HOURS-V. As shown in Table 4, the number of visits per year is the lowest at Bashan, but the hours per visit is the highest. The more time a person plans to spend at Bashan, the more he is willing to pay. Note that none of the income categories are significant in determining willingness to pay. At Coventry, the significant variables at the 0.01 level are SWIMMING, JETSKIING, WINTER, RIVERS and INCOME-E. The negative coefficients for SWIMMING, FISHING and GENERAL-O are relative to willingness to pay responses given by all respondents including jetskiers. Jetskiers have a very high willingness to pay and weight the overall average upward. The reasons for the significance of WINTER and RIVERS are the same as for Bashan. INCOME-D is significant at the 0.05 level. Coventry is the only lake for which INCOME-D has a significant impact on willingness to pay. Independent variables significant in determining willingness to pay at Crystal are VISITS-Y, HOUSEH-S, BOATING, INCOME-E, and SWIMMING. All are significant at the 0.01 level except SWIMMING, which is significant at the 0.01 level. The SWIMMING coefficient is positive. Different from Coventry, town residents must purchase a permit to use the town beach. Swimming is the primary activity at the town beach. The permit not only limits use of the town beach to residents, but also limits the use to residents who can afford the $35 or $45 annual fee. The beach at Crystal opens at 11 a.m. or noon7 and closes at 7 p.m., versus the town beaches at Coventry and Highland that are open all the time. There is actually a locked gate and fence surrounding Crystal’s beach, so access after hours is not possible. HOUSEH-S is negatively correlated with willingness to pay only at Crystal. A reason for this unique feature is not apparent from the data on household size and income. Independent variables significant in determining willingness to pay at Highland are BOATING, WINTER, LAKES, INCOME-E, INCOME-C and TOWN-R. The coefficients are significant at the 0.01 level for the ftrst three and at the 0.05 level for the latter two.

The opening time depends on whether or not there are swimming lessons offered in the morning. Table 7 shows the results of pooling the :data across water qualities and lakes. Each survey provided four willingness to pay responses, corresponding to the four water quality levels. Each response is treated as a separate observation, so the sample size is four tunes the 423 size in Table 4. In Model A there are dummy variables for individual lakes and water ...... quality levels. In Model B there is a dummy variable for water quality levels, but not for individual lakes. The water quality variables WQ-B, WQ-C AND WQ-D, BOATING, JETSKIING, WINTER and INCOME-E are significant at the 0.01 level in both models. TOWN-R and LAKES are significant at the 0.01 level in Model A and the 0.01 level in Model B. VISITS-Y is significant at the 0.05 level for both models. The COVENTRY and HIGHLAND coefficients are significantly different from BASHAN (the base case). This result corresponds to the willingness to pay estimates in Table 4 in that Coventry and Highland have the highest willingness to pay measures for all water qualities. The closeness of the R2s in both models (0.200 for Model A and 0.197 for Model B) indicates that individual lake char racteristies have a small impact on willingness to pay. Of the items included in the survey, the water quality level has the largest impact on willingness to pay. The Adjusted R2s for the regressions reported in Tables 5, 6 and 7 are low. This indicates that the survey did not capture independent variables that explain a significant part of a respondent’s willingness to pay response. Individual preferences and motivations beyond the surveyed household characteristics and recreational activities are captured in the error term. For instance, if a person is a member of an environmental organization, or contributes money to one, the willingness to pay is likely to be higher than if the person does not. The level of concern with environmental issues is most likely correlated with political patty preferences. Other influences on willingness to pay responses not captured in the survey may include one’s opinion as to the current water quality and how one thinks his annual fee will be used. Regarding current water quality, the survey stated that current water quality at the lake was suitable for swimming, fishing, eating of fish caught and boating. Some respondents did not think that the current water quality was appropriate for these activities. Regarding use of the fee, some respondents indicated they thought the fee would be used to improve water quality

30 Table 7. Determinants of Willingness to Pay at Public Sites with Data Pooled Across Lakes and Water Quality Categories

Model A Model B

Variable Coefficient Coefficiem (T-Ratio) (T-Ratio)

COVENTRY 16.00 omitted (2.25)* CRYSTAL 2.37 omitted (0.35) H/GHLAND 15.96 omitted (2.41)* WQ-B .43.24 .43.24 (-7.62)** (-7.61)** WQ-C -29.88 -29.88 (-5.27)** (-5.26)** WQ-D -56.78 -56.78 (-10.01)** (-9.99)** TOWN-R 14.07 11.83 (2.64)** (2.25)* TRAVEL-T 0.03 -0.02 (0.16) (-0.12) VISITS-Y 0. I 1 0.12 (2.39)* (2.60)* HOURS-V 1.67 1.61 (2.04)* (1.97) HOUSEH-S -0.09 -0.16 (-0.06) (4).11) SWIMMING -7.91 -6.22 (- 1.42) (-1.12) BOATING 22.34 22.63 (4.74)** (4.82)** JETSKIING 75.13 79.46 (6.09)** (6.47)** HSHING 1.11 (0.21) (-0.44) GENERAL-0 -5.42 -3.82 (-L02) (-0.72) WINTER 51.85 53.79 (6.56)** (6.83)** LAKES .4.16 .4.60 (-2.44)* (-2.70)** RIVERS -1.95 -3.06 (-0.36) (-0.57) OCEAN -8.76 -10.23 (-1.66) (-1.94) INCOME-B 12.43 II.12 (1.68) (i .51) INCOME-C 15.14 13.22 (2,16)* (1.88) INCOME-D 11.40 9.98 (1.42) (1.24) INCOME-E 71.31 68.11 (8.00)** (7.68)** CONSTANT 25.46 39.56 (1.94) (3.45)**

SAMPLE SIZE 1692 1692 ADJUSTED R~ 0.200 0.197

31 rather than maintaining the current level. There is:likely to be a correlation between a person’s opinion on current water quality and the intended use of the fee money.

VI. Possible Bias in Data

As mentioned in Section II, there is often some element of bias in a contingent valuation survey. Some of the possible biases in the two surveys done for this study are discussed below.

Possible Bias in the Survey of Waterfront Property Owners. Hypothetical and strategic biases may be present in our results. Hypothetical bias is present when we ask a question about changes in water quality that have not occurred. There is no true value against which to compare the hypothetical value because water quality is not a marketed good. A unique set of strategic biases may also be present. Property owners could have upward and downward strategic biases in estimating their property values. First, if an owner thinks the results will be used by public officials to allocate funds to clean up the lake, he may have an upward strategic bias for Category A (current water quality) and a downward strategic bias for lower water quality categories. Second, if an owner thinks the results of the survey, when made public, will positively influence the prices paid by buyers, he may have an upward strategic bias. Third, if an owner thinks the tax assessor will use the information to increase his assessment, he may have a downward strategic bias. However, there is no evidence that respondents underestimated the value of their homes for assessment purposes. The owner’s estimated property values with the existing water quality may be somewhat high relative to actual market values and town assessors’ values. Table 8 compares the respondent’s estimated values to sale prices and assessed values for improved waterfront properties at the three lakes. Sales data were collected for the years 1990 through 1996. Prices and assessed values were adjusted to a 1996 basis. The ratios of average values estimated by property owners to average adjusted sale prices are 1.0705 at Bashan, 1.1688 at Crystal, and 1.0489 at Highland. These ratios do not clearly indicate unrealistic estimates by the owners. The number of sales varied from ten at Crystal to 24 at Highland, which has the smallest ratio of estimated values to sale prices. Moreover, we do not know the extent to which the respondent and sales data pertain to the 32 same property. Meanwhile, the average assessed values of the properties actually sold adjusted to a 100 percent basis and then adjusted to 1996 prices are substantially lower than comparable sale prices. While the number of sales is small, this data does suggest that caution is advisable in an attempt to infer market values from assessed values.

Possible Bias in the Survey of Public Site Users. In addition to hypothetical bias, discussed above, information and strategic biases may also be present in these results. Respondents have familiarity with existing recreational opportunities; however, they may not be able to imagine a situation where all the lakes in the state are polluted and fees are charged to use all of them. Upward strategic bias may be present for persons who interpret the payment as a way to clean up the lake and/or protect the water quality. Downward strategic bias may be present in the responses of users, especially those of fishermen who think the results of the survey will be used to impose another fee on people who want to fish in public lakes. There was no evidence of substitute site bias. Public site users who do not live near the lake and visit 5 - 10 times per season have alternative sites. A respondent who only recreates at lakes had to determine which lakes he wanted to buy a pass for, given his set of preferences and budget constraint. If the lake where he is being interviewed does not make the chosen list of lakes, he will have a legitimate $0 response. Public site users who do live near the lake, and visit frequently, if not daily, have a different set of substitute sites. As with any on-site survey, individuals who visit more frequently have a higher probability of inclusion in the sample. The average number of visits per year by households in the sample is almost certain to be larger than the average number of visits per year by all households that visit at least once per year. The consequences of this upward bias in our estimate of the annual average number of visits per household will be discussed in the following section regarding estimation of aggregate benefits.

VII. Aggregate Benefit Estimation and Analysis

The values estimated in the waterfront property owner and public site user surveys are for individual households. Estimates of total property value and total willingness to pay by public site users are measured by aggregating the individual values over the appropriate

33 populations. The aggregation procedures used for the two types of population impacted are reported below. The annual loss in social welfare from a decline in water quality is estimated to be the sum of the decrease in annual willingness to pay by waterfront property owners and public site users. Estimates of annual willingness to pay for waterfront property are based on property values and interest and tax rates with procedures reported in the subsection on social welfare.

A. Waterfront Property Owners As with all survey data, an expansion of results to a population requires an assumption that the respondents are representative of the larger group. With this assumption the aggregate values of residential waterfront property at each lake with alternative water quality categories are estimated by multiplying mean property values by the number of residential waterfront properties (including those with deeded rights-of-way to the lake). Results are shown in Table 9. With our method and data the estimated aggregate property value at Bashan is $39.5 million at the existing water quality and would decline to $22.6 million with a loss of swimming and fish edibility. The comparable values are $17.4 million and $10.5 million at Crystal and $70.9 million and $40.4 million at Highland.

B. Public Site Users The results from the public site users, as presented in Table 4, are for a single household. In order to estimate the total value for all recreational users of a lake we needed to estimate the number of households using the lake. Information regarding number of visits and number of households was obtained from several sources. First, the surveyor made a periodic count of cars in the parking lots of public sites throughout the day. The difficulty in using these numbers is that the sample size is small. We spoke with the DEP site managers for the four boat launches, and obtained town beach user counts from the recreation departments in Coventry, Ellington and Winchester. The estimates of the number of households using public sites annually appear in Tables 10, 11, 12 and 13. The number of households visiting each boat launch was estimated by dividing the DEP estimate of the annual number of visits by the average number of visits per household per year

34 Table 8. Comparison of Property Owner’s Estimated Values to Sale Prices and Assessed Values for Improved Waterfront Properties

Average Values (1996 $)* Bashan** ~ ~ (1) Property Owner’s $178,883 $168,500 $189,056 Estimated Market Value (n =) 64 32 141

(2) Actual Sale Price of $167,106 $144,159 $180,243 Properties Sold (n =) 16 10 24

(3) Assessors’ 100% value $139,649 $128,558 $144,394 of Properties Sold (n =) 16 10 24

Ratios (1) / (2) 1.0705 1.1688 1.0489 (2) / (3) 1.1966 1.1214 1.2483 (1) / (3) 1.2809 1.3107 1.3093 * Housing prices indexed for 1990 - 1995 using Connecticut Attorney’s Title Insurance Co. data; 1995 - 1996 prices indexed using University of Connecticut, center for Real Estate and Urban Economic Studies data. ** Co-opera, ely owned property at Wildwood is not included in either sale prices or assessed values.

Table 9. Estimated Aggregate Values of Residential Waterfront Property with Alternative Water Quality Categories

Bashan C~stal H.H_ig_hland Number of Properties 221 10._~3 37_~5 Property Value by Water (in millions of dollars) Quality: A - Swimmable and $39.5 $17.4 $70.9 Fishable (edible) B - Sv:Lnxr.-~b!e and $25.5 $11.9 $45.2 Fishable (edible) C - Swimmable and $32.4 $13.7 $57.3 Fishable (edibte) D - gwinmaWale and $22.6 $10.5 $40.4 Fishable (edible)

35 of the respondents surveyed at that site.s For the public beaches we used several indicators to estimate the number of user households, including the number of season passes sold, the number of nonresident passes sold, and beach attendance counts. Details of the calculations appear in footnotes accompanying the tables. The total number of households using public sites annually is multiplied by the average willingness to pay figure for each water quality level, which are listed by lake in Table 4. The estimates of the number of households using public sites annually and aggregate use values appear in Tables 10, 11, 12 and 13, and are summarized in Table 14.

C. Social Welfare The annual cost of owning a waterfront property includes interest paid and/or earnings foregone on invested capital, property taxes and maintenance costs. Some of these costs are generally offset by reductions in Federal income tax through the deduction of property tax payments and interest expense from taxable income. We have no basis for a precise estimate of outstanding debt and interest costs or of the opportunity cost of capital to the owners. Likewise, we have no basis for estimating the income tax reductions associated with home mortgage interest deductions. A conservative estimate of annual capital cost can be obtained by using the average yield on a 5 year U.S. Treasury note in the summer of 1995. The average yield was 5.93% in June 1995, 6.28% in July 1995 and 6.24% in August 1995.9 We use an average of the three months, 6.15%. Annual property taxes are directly related to the property values and local tax rates. An estimate of the Federal income tax advantage of the property tax deduction is based on an extrapolation to all households of the situation judged to be most consistent with levels of household income reported in the survey. The most common situation appears to be a household that itemizes deductions and is in the 28% marginal tax bracket. While not precise, this approach seems more reasonable than available alternatives.

s The aggregate annual use values for public site users in Tables 10, 11, 12 and 13 may be underestimated due to sampling bias. The probability of interviewing a frequent household user was greater than the probability of interviewing a household that visited only once or twice per year. This creates an upward bias in the estimated number of visits per household per year and a downward bias in our estimate of the number of households using public sites. 9 Appraiser News, published by the Appraisal institute, Chicago IL, September 1995, October 1995, November 1995.

36 Annual maintenance costs would presumably be the same with each water quality category. Thus, subsequent estimates of annual willingness to pay for waterfront property will be exclusive of maintenance costs.

37 Table 10. Calculation of the Total Willingness to Pay for Public Site Users of Bashan Lake

Calculation of Number of Households Using Boat Launch Source: Don Goss, DEP Site Manager for Bashan Lake boat launch

Avg. # Vehicles per Day Time Period Using Boat Launch Total Visits a) April - October (weekdays) 13 1,781 b) April - October (weekends & holidays) 20 1,180 c) Jan - February (daily) 20 1ASO Total Household Visits 4,141 divided by average # visits per HH per year 9.92 = Total Number of Households Using Boat Launch Annually 417

Total Number of Households Using Public Sites Annually 417 [

Aggregate Annual Use Value for Public Site Users: Water Quality A 417 x $39.27 $16,376 Water Quality B 417 x $21.80 $9,091 Water Quality C 417 x $23.24 $9,691 Water Quality D 417 x $10.43 $4,349

38 Table 11. Calculation of the Total Willingness to Pay for Public Site Users of Coventry Lake

Calculation of Number of Households Using Boat Launch Source: Bruce Elliot, Site Manager for Coventry Lake boat launch Mark Paquette, Director of Recreation for the Town of Coventry

Avg. # Vehicles per Day Time Period Using Boat Launch Total Visits a) April 15 - October 15 (weekdays) 20 2,740 b) April 15 - October 15 (weekends & holidays) 30 1,770 Total Household Visits to Boat Launch 4,510 divided by average # visits per HH/yr for HHs interviewed at Boat Launch 40.55 = Total Number of Households Using Boat Launch Annually 111

Calculation of Number of Households Using Town Beaches Resident: Population of Coventry1° 10,905 x Percent of Population That Swims at a Beach11 13% x Percent of Population that Recreates in Home Town 48% 680 divided by Avg. HH Size of Public Beach Respondents 3.40 = Gross No. of Resident HHs Using Public Beaches Annually 200 Nonresident: No. Daily Passes Sold at Patriot’s Park 377 divided by 4 visits per HH/yr. 12 = 94 + No. Season Passes Sold to Patriot’s Park 117 = Gross No. of Nonresident HHs Using Public Beaches Annually 211

~o State of Connecticut, Department of Public Health, Estimated Populations in CT as of 7/1/95. ~ State of Connecticut, Department of Environmental Protection, Statewide Comprehensive Outdoor Recreation Plan, 1993-1998, Final Draft, September !993, Chapter V "Outdoor Recreation Demand Survey," page 7. The SCORP was prepared on recreational activities statewide. The percent of population who swim at a beach is a statewide figure. ~ Non-resident households entering the park on weekends and holidays Memorial Day through Labor Day pay a $5.00 fee to park. A utility maximizing household would plan the summer’s recreation schedule out before summer arrived. If the utility maximizing number of visits to Patriot’s Park was 4, the family would be indifferent to paying $20 for a season pass or a $5 parking fee for each of 4 visits. If the family planned to visit more than 4 times per summer, it would purchase a season’s pass. If the family did not plan out the recreation schedule for the summer, and ended up visiting 8 - 10 times during the summer when die $5 fee is charged, it would pay more than double what it would have paid with planning. The next year they might not buy a season pass because their annual number of visits varies from year to year; or they might buy a season pass if they plan to visit the park at least the same number of times as the current year. 39 Table 11. Calculation of the Total Willingness to Pay for Public Site Users of Coventry Lake (continued)

Tota! Number of Ho~eho!ds Using Tow B~¢h~ ~u~!y i 4!!

Total Number of Households Using Public Sites Annually 522 [

Aggregate Annual Use Value for Public Site Users: Water Quality A 522 x $79.66 $41,583 Water Quality B 522 x $27.88 $14,553 Water Quality C 522 x $50.37 $26,293 Water Quality D 522 x $13.52 $7,057

4O Table 12. Calculation of the Total Willingness to Pay for Public Site Users of Crystal Lake

Calculation of Number of Households Using Boat Launch Source: Jim Kane, Site Manager for Crystal Lake boat launch Bob Tedford, Director of Recreation for the Town of Ellington, and his staff

Avg. # Vehicles per Day Time Period Using Boat Launch Total Visits a) April - November 45 10,080 b) January - February 20 z z_b_ 80 Total Household Visits to Boat Launch 11,260 divided by average # visits per HH per year for HHs interviewed at boat launch 31.89 = Total Number of Households Using Boat Launch Annually 353

Calculation of Number of Households Using Town Beach No. Family Passes to Town Beach Sold 232 + No. Individual Passes to Town Beach Sold 24 + No. Senior Citizen Passes Issued to Town Beach 61 + No. Family Passes to Town Beach Sold to Stafford Residents 44 + [No. Adult Daily Passes Sold to Town Beach, (1,317)t3 divided by average no. visits per HH per year, (9)] 146 = Total No. I-His Using Town Beach Annually 507 I To~ Number of Households Using Public Sites Annually 860

Aggregate Annual Use Value for Public Site Users: Water Quality A 860 x $63.02 $54,197 Water Quality B 860 x $21.97 $18,894 Water Quality C 860 x $38.13 $32,792 Water Quality D 860 x $10.90 $9,374

~3 Daily entry fee was $2 per adult and $1 per child. The break-even point for an adult and 2 children is 8.75 visits, rounded to 9 visits. 41 Table 13. Calculation of the Total Willingness to Pay for Public Site Users of Highland Lake

Calculation of Number of Households Using the Boat Launch Source: Richard Miska, Site Manager for Highland Lake boat launch John Bennett, Director of Recreation for the Town of Winchester

Avg. # Vehicles per Day Time Period Using Boat Launch Total Visits a) May 1 - September 1 (weekdays) 35 2,695 b) May 1 - September 1 (weekends & holidays) 55 1,925 c) January - February 15 885 d) April 2O 600 Household Visits to Boat Launch 6,480 divided by average # visits per HH per year for HHs interviewed at boat launch 37.08 = Total Number of Households Using Boat Launch Annually 175

Calculation of Number of Households Using Town Beaches Total Average Daily Beach Attendance at Reisha Beach14 5,594 divided by average HH size for HHs interviewed at Reisha Beach 2.58 divided by average # visits per HH per year 46.70 = Total Number of Households Us’rag Reisha Beach

Total Attendance (based on average daily attendance for either a 30 day month (July & August) or a 15 day month Oune & September). Jtme Average Daily Attendance = 56.25 people (counts taken for 4 days in June) July Average Daily Attendunee = 69.13 people (counts taken for 15 days in July) August Average Daily Attendance = 61.06 people (counts taken for 16 days in August) September Average Daily Attendunee (no counts taken in September - use June Average)

June = 56.25 x 15 = 844 July = 69.13 x 30 = 2,074 August = 61.06 x 30 = 1,832 September = 56.25 x 15 8.~_~ 5,594

42 Table 13. Calculation of the Total Willingness to Pay for Public Site Users of Highland Lake (continued)

Total of Average Daily Beach Attendance at Holland Beach15 10,930 divided by the average HH size of HHs interviewed at Holland Beach 1.77 divided by average # visits per HH per year 35.54 = Total Number of Households Using Holland Beach 174

Total Nnmher of Households Using Public Sites Annually 395~6I

Aggregate Annual Use Value for Public Site Users: Water Quality A 395 x $93.03 $36,747 Water Quality B 395 x $39.63 $15,654 Water Quality C 395 x $49.72 $19,639 Water Quality D 395 x $23.44 $9,259

15 Total Attendance (based on average daily attendance for either a 30 day month (July & Augus0 or a 15 day month (June & September). June Average Daily Attendance = (assumed to be 80% of July’s count, based on similar change for Reisha Beach - no counts taken in June) July Average Daily Attendance = 148.80 people (counts taken for !0 days in July) August Average Daily Attendance = 96.48 people (counts taken for 21 days in August) September Average Daily Attendance (assumed to be 80% of July’s count, based on similar change for Reisha Beach - no counts taken in September) June = 119.04 x 15 = 1,786 July = 148.80 x 30 = 4,464 August = 96.48 x 30 = 2,894 119.04 x 15 ~ September = 10,930

16 Not included in the total # of households figure are use of the lake by Camp Delaware and the town’s summer playground program. According to the town’s Director of Recreation, 8 - 10 campers per day are brought down to the lake by Camp Delaware, and there are 40 - 55 children per week that are brought to Reisba Beach once a week for 8 weeks. 43 Table 14. Recreational Use Value of Public Sites with Alternative Water Quality Categories

Annual Values Water Quality Bashan Coventr~ Crys~l ~ A $16,376 $41,583 $54,197 $36,747

B $9,091 $14,553 $18,894 $11,638

C $9,691 $26,293 $32,792 $19,639

D $4,349 $7,057 $9,374 $9,259 The annual willingness to pay for residential waterfront property with alternative water quality categories is estimated as follows:

A=V(i+0.720 where: A = annual willingness to pay V = aggregate property value i = interest rate of 0.0615 t = the 1995 tax rate times 0.70 to adjust for assessment at 70% of market value

This procedure incorporates the assumption that Federal income taxes are reduced by $0.28 for each $1.00 of local property tax. Estimated annual willingness to pay for residential waterfront properties with alternative water quality categories are shown in Table 15. The total annual willingness to pay by waterfront property owners and public site users for each water quality category are shown in Table 16. Estimates of the annual loss in social welfare from a shift from water quality A to each of the other water quality categories are shown in Table 17. Each loss is the sum of the corresponding decreases in willingness to pay by waterfront property owners and public site users. Care was taken to avoid double counting problems that may occur when more than one evaluation technique is used for the same lake. McConnell (1990) addresses this problem when hedonic property value and travel cost methods are used. The issue is the inclusion of avoided travel costs in waterfront property values in the form of location rent. We do not double count the capitalized location rent, or travel cost, inherent in the waterfront property owners’ values because a travel cost model is not used in the public site value estimation. A second issue is the population over which the benefits are aggregated. A few waterfront property owners use the state boat launches because their water frontage is rocky or steep. Their values as public site users are excluded from our sample because waterfront property owners were screened out by a question asked at the beginning of the survey. A comparison of the percentage loss in social welfare by public site users and waterfront property owners from a shift from water quality A to alternative water qualities is

45 shown in Table 18. The percentage loss is higher for the public site users than for the waterfront property owners for Bashan, Crystal and Highland.~7 The percentage loss in value is almost twice as high for public site users than waterfront property owners for all water quality changes at Crystal and Highland. The larger !oss for public site users is because they do not have a ’residual residential value’. For instance, with water quality level D, waterfront property owners suffer a loss in property value but still have use of their property as a residence. There is a residual value to a waterfront property owner. Public site users receive no such residual. The magnitude of the value estimates for waterfront property owners is much greater than the corresponding value estimates for public site users, indicating that the bulk of benefits and losses from changes in water quality accrue to waterfront property owners.

~7 The comparison cannot be made for Coventry because results of the survey of waterfront property owners were not usable. Refer to footnote #3.

46 Table 15. Estimated Annual Willingness to Pay for Residential Waterfront Properties with Alternative Water Quality Categories

Bashan C_C.C.C.C.C.C.C.C_~~ Number of Properties 221 10__~3 37_.j.5

Property Value by Water Quality: A $3,123,066 $1,379,762 $5,813,595 B $2,014,943 $942,988 $3,707,835 C $2,558,267 $1,085,468 $4,694,910 D $1,782,739 $837,765 $3,314,235

Table 16. Total Annual Willingness to Pay by Waterfront Property Owners and Public Site Users

Property Value by Water Bashan

A $3,139,442 $1,433,959 $5,850,342 B $2,024,034 $961,882 $3,719,473 C $2,567,958 $1,1189260 $4,714,549 D $1,787,088 $847,139 $3,323,494

47 Table 17. Annual Loss in Social Welfare by a Shift from Water Quality A to Alternative Water Quality Categories

Water Qu~ity~ve! Bash~. ~ i Highland B $1,115,408 $472,077 $2,130,869

C $571,434 $315,699 $1,135,793

D $1,352,354 $586,820 $2,526,848

48 Table 18. Comparison of Percentage Changes in Social Welfare for Waterfront Property Owners and Public Site Users

Shift from Water Quality A to: P~rcentage Loss in Value Bashan ~ Hi~ B - Waterfront Property Owners 35.48 31.66 36.22 B - Public Site Users 44.49 65.14 68.33

C - Waterfront Property Owners 18.08 21.33 19o24 C - Public Site Users 40°82 39.49 46.56

D - Waterfront Property Owners 42.92 39.28 42.99 D - Public Site Users 73.44 82.70 74.80

49 VIII. Impact of Changes in Water Quality Levels on the Incidence of Real Estate Taxes Collected by the Townis

Maintaining lake water quality is in the interest of all property owners in a town with wate~oni prop~i when property values decline the taxable real estate value in a town declines. When the deterioration of lake water quality causes a decrease in the taxable base, a municipality has two choices: increase the mill rate for all property owners, or decrease municipal expenditures and, by extension, services. Assuming the current level of municipal services to be the status quo, if the taxable base decreases as water quality deteriorates the tax rate must increase to raise the same amount of revenue. The fiscal result of a change in water quality includes both a short-term and a long-term redistribution of tax incidence within a municipality. There are two short-term impacts. First, as the market values of waterfront properties decline due to deteriorating water quality, the effective tax rate on waterfront properties will be higher than for non-waterfront properties until such time as assessed values are changed to reflect new market values. Second, waterfront property market values may be further reduced by the capitalization of the excess tax expense. To illustrate the short term effects, assume an assessed value of $100,000 for two homes in the same town, one on a lake and one not on a lake. Further assume that the assessed value is 100% of the market value. If the effective tax rate is 2.0% for both residential properties, each property owner pays $2,000 per year in real estate taxes. If there is deterioration in water quality, the waterfront property may no longer have a market value of $100,000. The decline in market value attributable only to the change in water quality is, say $20,000. The assessed value does not change in the short-run and the owner is now paying an effective tax rate of $2,0001580,000 or 2.5 %, while non-waterfront property owners continue to pay 2.0%. This inequity is the first short-term effect. Until the assessed value changes to reflect the new $80,000 market value. The property owner is paying $400 more per year in property taxes than the owner of a property away from the lake with a value holding steady at $80,000. The second short-term effect is a further reduction in market value due to the high effective tax rate (Gyourko and Tracy, 1989). Two or three years might pass before the decline in value is clearly established by property sales. An additional period of up to ten

~s In reading this section, the reader should note that Water Quality A is the existing water quality, Water Quality B is when swimming is no longer advisable, Water Quality C is when eating of fish is no longer advisable and

50 years could pass before the next revaluation. The previous example will be continued with an assumption that the two delays total eight years and the future values are discounted at the 6.15% annual rate previously used in estimating annual willingess to pay for waterfront property. The present value of eight annual payments of $400~ discounted at 6.15%, is $2,469. The $80,000 value would decrease to $77,531. In summary, a decline in waterfront property values initially increases the effective tax rate, a result that may further reduce the market value. Estimates of excess taxation and potential impacts on average property values at Bashan, Crystal and Highland are shown in Table 19. The estimates pertain to the average property values estimated by owners with each of the four water quality categories. The mill rates are for taxes due in 1995. The estimated excess tax is based on an assumption that a decrease in water quality changes market values immediately and changes assessed values after eight years. Future values are discounted at a 6.15% annual rate. As an example of the calculation, the owner of a waterfront property on Bashan with Water Quality A pays $178,880 x 0.01697 = $3,036 per year in taxes. Then the water quality level changes to Water Quality B the market value declines to $115,410. A property owner would pay $115,410 x 0.01697 = $1,959 in taxes if the assessed value were immediately revalued to reflect the change in water quality. Instead, the owner pays an excess tax of $3,056 - $1,959 = $1,077. The present value factor of a $1 payment annually, discounted at 6.15% for 8 years is 6.173. The present value of the excess tax payments is 6.173 x $1,077 = $6,648. The $115,410 value is adjusted down by $6,648 for a property value of $108,762. In the long-run, a revaluation should eventually equalize the assessment ratios and the effective tax rate. Potential losses in tax revenue from waterfront properties are estimated from the aggregate property values reported in Table 9 and property tax rates effective for the 1995 tax year.19 The annual tax revenue from waterfront properties with alternative water quality categories is shown in Table 20. The total potential annual losses in tax revenue due to a shift from Water Quality A to Water Quality B are estimated at $237,580 (Bashan), $96,250 (Crystal) and $467,226 (Highland). The potential annual losses due to a shift from Water Quality A to Water Quality C are estimated at $120,487 (Bashan), $64,750 (Crystal) and

Water Quality D is when both swimming and eating of fish caught are no longer advisable. ~9 The property value estimates are 100% of market value. The actual 1995 mill rates in each town have been multiplied by 0.70 since property is taxed at 70% of its market value, 51 $247,248 (Highland). A shift from Water Quality A to Water Quality D could result in potential tax revenue losses of $286~793 (Bashan), $120,750 (Crystal) and $554,490 (Highland). The 10SS of ~ revenue ould result ~ either comparable reductions in municipal expenditures or increases in the tax rate to recover the loss from all taxpayers. In the latter case, the increase in tax rate will depend on both the loss in value of waterfront property and the aggregate value of waterfront properties relative to the total assessed value of all taxable property.

Conclusion

This study provides a way to value changes in lake water quality by examining the potential impact on property values and recreational use values. The approach is particularly useful when data on actual variation in environmental quality is not available and a conventional property price approach is therefore not feasible. We used our methodology to value hypothetical changes in water quality at a sample of suburban lakes in Connecticut. The lakes have similar trophic classifications and recreational use patterns. All four have a state boat launch with developed access, and a higher density of residential development around the lake than in the rest of the town in which the lake is located. Based on responses received from mail surveys sent to waterfront property owners (and those with deeded rights-of-way to the lake), we found that property value estimates are more responsive to loss of swimming (and related activities) than to loss of fish edibility. This ranking is consistent across lakes, and even the magnitude of the responsiveness is remarkably comparable when viewed in percentage terms. For example, property value estimates go down by 40-43% at each lake if the water is no longer swimmable and the fish caught are not edible. Statistical regression analysis was performed to examine relationships between estimated property values and other variables. This was done in two ways. First, data from all the lakes was combined and a separate regression run for each water quality level. We found that property values are more closely related to incidence of year-round residence by the owner and to size and features of the lot than to recreational activities. The separate regressions for each water quality also indicate that the contribution of a dock and sandy beach to property value are substantially smaller in the absence of swimmable water quality.

52 Table 19. Excess Taxation and Potential Impacts on Average Property Values

Water Quality Category MillRate~° A B C D Bashan 16.97 Property Value Estimated by Owner $178,880 $115,410 $146,530 $102,110 Excess Tax per Year 0 $1,077 $549 $1,303 Present Value of Eight Excess Payments 0 $6,648 $3,389 $8,044 Adjusted Property Value $178,880 $108,762 $143,141 $94,066

17.50 Property Value Estimated by Owner $168,500 $115,160 $132,560 $102,310 Excess Tax per Year 0 $934 $629 $1,158 Present Value of Eight Excess Payments 0 $5,766 $3,883 $7,149 Adjusted Property Value $168,500 $109,394 $128,677 $95,161

Highland 18.18 Property Value Estimated by Owner 189,060 $120,580 $152,680 $107,780 Excess Tax per Year 0 $1,245 $661 $1,478 Present Value of Eight Excess Payments 0 $7,686 $4,080 $9,124 Adjusted Property Value $189,060 $112,894 $148,600 $98,656

~o The property value estimates are 100% of market value. The actual 1995 mill rates in each town have been multiplied by 0.70 since property is taxed at 70% of its market value. 53 Table 20. Annual Tax Revenues from Waterfront Properties with Alternative Water Quality Categories21

Water Quality Level Bashant ~ Highland*** A $670,315 $ 304,500 $1,288,962

B $432,735 $208,250 $821,736

C $549,828 $239,750 $1,041,714

D $383,522 $183,750 $734,472

2~ The property value estimates are 100% of market value. The actual 1995 mill rates in each town have been multiplied by 0.70 since property is taxed at 70% of its market value.

54 The second type of statistical regression analysis pooled the four property value estimates by each respondent and included dummy variables for each of the three hypothetical water quality categories. Regressions were run for each lake and for the three lakes combined. Estimated reductions in property value due to shifts from existing t6 hypothetical water quality categories are significantly different from zero at the 0.01 level in all cases except for the loss of fish edibility at Bashan, which is significant at the 0.05 level. The two model types produced generally consistent estimates regarding the effect of property and household characteristics on property value. To obtain a complete measure of recreational use value, public site users who are not waterfront property owners were surveyed. Based on these survey responses, we found that the greatest loss in value occurs when there is a loss of swimming at three of the four lakes. The exception is Bashan, where the change in value from loss of swimming is comparable to the change in value from loss of fish edibility. This is consistent with the high percentage of fishermen using the lake. Also consistent with the high percentage of fishermen using Bashan is the finding that rivers are not substitute fishing sites for the lake. The on-site survey data was pooled across lakes and a separate linear regression was run for each water quality. The annual willingness to pay was found to be more closely related to recreational activity than to income and frequency of use. Income was only significant in determining willingness to pay for the highest income group. Winter users of the lake and jet skiers generally have higher values than other users because of the limited number of substitute sites. The data was also pooled across lakes and water qualities and a linear regression run. Again, recreational use had the greatest influence on value. The independent variables for individual lakes were not significant. Town residency affected the magnitude of the willingness to pay responses. At Coventry, Crystal and Highland more than 50% of the public site users were town residents and visited the lake 30 - 40 times per year. The willingness to pay responses for these three lakes were higher than at Bashan where less than 25% of the public site users were town residents, and the average number of visits annually was 10. Coventry, Crystal and Highland have public beaches (for residents and/or non-residents) which provide access to the lake for swimming. There is no public beach on Bashan, and, hence, no access developed for swimmers. For this reason, swimming is not the primary activity for

55 most users of Bashan while fishing is. The publi6 use value of a recreational lake in suburban Connecticut is likely to increase when public access for swimming is developed. Estimates of the annual loss in social welfare from a shift from Water Quality A to each of the other water quality categories is the sum of the corresponding decrease in annual willingness to pay by waterfront property owners and public site users. The percentage loss in social welfare is largest when the ftsh are not edible and the water is not swimmable. The percentage loss in social welfare is smallest when the fish are not edible. At Bashan Lake, for example, the annual loss in social welfare is $571,434 when the ftsh are not edible and $1,352,354 when the fish are not edible and the water is not swimmable. A decline in waterfront property values from a decrease in water quality will initially increase the effective tax rate on these properties relative to non-waterfront property. Upon a general revaluation, the inequity is eliminated through a redistribution of the tax burden among all property owners in the town. Policy makers may be interested in knowing whether the results from our study are likely to carry over to other lakes. While our sample size is too small to permit any confident generalizations, some of the empirical regularities and consistencies are encouraging. We are inclined to think that they may well apply to other suburban lakes with sizes and residential patterns similar to the ones in our study. However, this hypothesis needs to be tested with data from additional lakes. Furthermore, a greater variety in the characteristics of the lakes selected for the study would permit more refined hypotheses to be formulated and tested.

56 X. REFERENCES

Brown, G.M., Jr. and H.O. Pollaskowski. "Economic Valuation of Shoreline," The Review of Economics and Statistics (August 1977): 272-278.

Cameron, T.A. and J. Englin. "Respondent Experience and Contingent Valuation of Environmental Goods," Journal of Environmental Economics and Management 33, (July 1997): 296-313. Darling, A. H. "Measuring Benefits Generated by Urban Water Parks," Land Economics 49 (February 1973): 22-34.

David, E.L. "Lakeshore Property Values: A Guide to Public Investment in Recreation," Water Resources Research 4 (August 1968): 697-707. Falcke, C. O. "Water Quality and Property Prices: An Econometric Analysis of Environmental Benefits," Commentationes Scientiarum Socialium 18 (1982): 1-187.

Freeman, A. M., HI. The Measurement of Environmental and Resource Values: Theory and Methods. Washington D.C.: Resources for the Future, 1993.

Green, C.H., and S.M. Tunstall. "The Evaluation of River Water Quality Improvements by the Contingent Valuation Method," Applied Economics 23 (1991): 1135-1146. Greene, W.H. Econometric Analysis, 2nd Edition. New York: Macmillan Publishing Co., 1993.

Griliches, A. ed. Price Indexes and Quality Change. Cambridge, Mass.: Harvard University Press, 1971. Gyourko, J. and J. S. Tracy. "On the Political Economy of Land Value Capitalization and Local Public Sector Rent-Seeking in a Tiebout Model," Journal of Urban Economics 26 (1989): 152-173. Kanemoto, Yoshitsugu. "Hedonic Prices and the Benefits of Public Projects," Econometrica 56 (July 1988): 981-989.

Kopp, R.J., and V.K. Smith, editors. Valuing Natural Assets, Washington D.C.: Resources for the Future, 1993. Lancaster, K.J. "A New Approach to Consumer Theory,~ Journal of Political Economy 74 (April 1966): 132-157.

Lansford, N.H., Jr. and L.L. Jones. "Recreational and Aesthetic Value of Water Using Hedonic Price Analysis," Journal of Agricultural and Resource Economics 20(2) (1995): 341-355.

57 M~ler, K., "A Note on the Use of Property Values in Estimating Marginal Willingness to Pay for Envirorunental Quality," Journal of Environmental Economics and Management 4 (December 1977): 355-369.

McConnell, K.E. "Double Counting in Hedonic and Travel Cost Models," Land Economics, 66 (2), (May 1990): i2i~27~

Michael, H. J., K. J. Boyle, and R. Bouchard. "Water Quality Affects Property Prices: A Case Study of Selected Maine Lakes," Miscellaneous Report 398, Maine Agricultural and Forest Experiment Station, University of Maine, February 1996.

Mitchell, R.C., and R.T. Carson. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington D. C.: Resources for the Future, 1989. Montgomery, M. and M. Needelman. "The Welfare Effects of Toxic Contamination in Freshwater Fish," Land Economics 73 (2) May 1997:211-23.

Needelman, M.S. and M.J. Kealy. "Recreational Swimming Benefits of New Hampshire Lake Water Quality Policies: An Application of a Repeated Discrete Choice Model," Agricultural and Resource Economics Review, April 1995: 78-87.

Niskanen W. A. and S.H. Hanke. "Land Prices Substantially Underestimate the Value of Environmental Quality," The Review of Economics and Statistics 59 (3) 1977: 375-377. Pearce, D.W., ed. The MIT Dictionary of Modern Economics, 4h edition, Cambridge: The MIT Press, 1992. Rosen, S. "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition," Journal of Political Economy 82 (January/February 1974): 34-55.

Smith, V.K. and William H. Desvousges. Measuring Water Quality Benefits, Boston: Kluwer Nijhoff Publishing, 1986, Chapter 4.

Sonstelie, J.C. and P. R. Portney. "Gross Rents and Market Values: Testing the Implications of Tiebout’s Hypothesis," Journal of Urban Economics 7 (1980): 102-118.

Willis, C.E., J.H. Foster and K. Sewall. "Valuation of Intangibles: The Case of Water Quality," Publication No. 133, Water Resources Research Center, University of Massachusetts at Amherst, January 1983.

Young, C.E., and F.A. Teti. "The Influence of Water Quality on the Value of Recreational Properties Adjacent to St. Albans Bay, Vermont." ERS Staff Report No. AGES 831116~ USDA, ERS, Natural Resource Economics Division, 1984.

58 XI. Glossary

Bias: Bias in survey responses due to the flaming of the contingent commodity or in the survey procedures (Smith and Desvousges, 1986). The different types of biases in contingent valuation are variously classified, depending on the economist. A list of the major biases includes strategic, information, hypothetical, starting point, payment vehicle, sampling, and interviewer. General definitions of the biases follow. Strategic bias may occur when the respondent has an ulterior motive. Information bias may occur when the respondent does not have enough or the correct information to give a reasoned answer. Hypothetical bias can occur because the contingent market does not exist and real budgetary decisions do not have to be made. Simulations may mitigate hypothetical bias. Starting point and payment vehicle biases are correlated with the question format. An example of sampling bias is when only the people who are interested in a topic complete and return a mail survey; the respondents self-select themselves. Interviewer bias may occur when the interviewer is not properly trained or in some manner influences the respondent’s answers. Capitalized Value: The market value that would be placed on an asset at its existing level of earnings and at current market rates of interest (Pearce, 1992). This value may be calculated by taking all future earnings from the asset and adding them up after diseotmting them to the current time period. The same procedure can be used to fred the present value of future expenses, such as excess taxes. Contingent Valuation: An approach used to estimate benefits of a non-marketed good or service. Individuals are given hypothetical choices and asked either willingness to pay or willingness to accept compensation questions. CV is a direct method of valuation. The travel cost and property value methods are indirect methods. Double Counting: The counting of a single element of benefit or cost more than once in a cost- benefit analysis (Pearce, 1992).

Hedonic Price: The implicit or marginal price of a characteristic of a commodity. There are a number of characteristics that determine the quality and value of a commodity. A part of the price of a marketed good can be associated with each characteristic. A consumer buys a bundle of these characteristics, for instance, when purchasing a house. The bundles are different for different houses, hence different sale prices. Hedonic price theory was developed for marketed goods, but has been used to value quality changes in environmental goods. Level of Significance: The level of significance is the probability of an incorrect rejection of a null hypothesis. In this report, t-ratios are used to test a hypothesis that the corresponding regression coefficient is not significantly different from zero. A t-ratio that is significant at the 0.05 level implie~ that the magnitude of the corresponding regression coefficient would occur by chance in only 5% of all possible samples; thus, we reject the null hypothesis of no relationship between that independent variable and the dependent variable.

Question Format: There are five general types of contingent valuation question formats: 1) Direct Question (i.e. how much are you willing to pay for....?); 2) Bidding Game or Iterative Bidding (i.e. would you pay $10? If yes, would you pay $20? If yes, would you pay $25, or if no, would you pay $157); 3) Payment Card (respondent selects a payment category from a 59 list of several; 4) Referendum (i.e. would you b~ willing to pay $100 more in taxes for a decrease in air pollution? Respondent votes yes or no; and 5) Ranking of Choices and/or Payment Categories.

R2: Also referred to as the coefficient of determination. A statistic which indicates the explanatory power of a regression equation: It is the percentage of variation in the dependent variable that is explained by the regression equation. The R2 statistic lies between zero and one. The closer it is to zero the lower the explanatory power of the equation. The closer it is to one, the higher the explanatory power. A R2 of 1.00 means the regression equation explains 100% of the variation in the survey responses. The dependent variables in our study are the market value estimates of properties given changes in water quality, and the willingness to pay an annual user fee given changes in water quality.

Reliability: Whether a {benefit} measure will give approximately the same result when applied to another, similar population and issue. Use Value: The value that individuals attach to the in situ consumption of the services of a natural resource (Kopp & Smith, 1993, p.341). Significance: [See Level of Significance]

Validity: Whether a {benefit} measure indeed measures what it is supposed to be measuring. (Cameron & Englin, 1997).

Willingness to Pay: Falcke (1982) defines willingness to pay as the maximum dollar amount a consumer is prepared to give up in order to either enjoy an activity or be protected from it. Willingness to pay is equivalent to a benefit. A respondent is willing to pay to use a lake at alternative water qualities if the fee is less than or equal to the utility he receives from recreating at the lake with a given level of water quality.

60 Fo~ more information or to contact the Lakes Management Program, call Charles Lee at (860) 424-3716.

Connecticut Department of Environmental Protection 79 Elm St. Hartford, CT 06106

The DEP is an equal opportunity/affirmative action employer, offering its services without regard to race, color, religion, national origin, age, sex, or disability. In conformance with the Americans with Disabilities Act, the DEP makes every effort to provide equally effective services for persons with disabilities. Individuals with disabilities needing auxiliary aids or services should call the human resource division at (860) 424-3006.

Printed on recycled paper." 25% post-consumer content.

DEP Website http://dep.state.ct.us