ii Property Owners’ Willingness To Pay For Restoring Impaired Lakes: A Survey In Two Watersheds of the Upper Mississippi River Basin This study was conducted on behalf of the Minnesota Pollution Control Agency, Sauk River Watershed District, and City of Lake Shore, Minnesota by Patrick G. Welle, Ph.D.* Professor of Economics and Environmental Studies Bemidji State University Bemidji, Minnesota and Jim Hodgson Upper Mississippi River Basin Coordinator North Central Regional Office Minnesota Pollution Control Agency Brainerd – Baxter, Minnesota October, 2008 wq-b4-01 iii Acknowledgements Assistance with the mail survey and interviews was provided by Laura Peschges, Bridget Sitzer, Emily Seely, Asher Kingery, Scott Gullicksrud and Brent Hennen. The format and certain sections of this report are adapted from three previous reports to the Legislative Commission on Minnesota Resources: one by the lead author “Multiple Benefits from Agriculture: A Survey of Public Values in Minnesota” (2001) and two by the lead author and two co-authors, Daniel Hagen and James Vincent: “Economic Benefits of Reducing Toxic Air Pollution: A Minnesota Study” (1992) and “Economic Benefits of Reducing Mercury Deposition in Minnesota” (1998). Additionally the authors would like to thank the following staff of the MPCA North Central Office staff: Bev Welle for her assistance in setting up the advisory committees, typing and copying the many drafts of the survey, Bonnie Finnerty for review and comments of the survey first as the Crow Wing County Planner and then as a staff member of the MPCA; Maggie Leach, Bob McCarron, and David Christopherson for their editorial and content comments; Greg Van Eeckhout, for assistance in the development of the description of the current conditions for both watersheds; and Pete Knutson for the GIS and mapping assistance. iv Table of Contents Section Page I. Executive Summary 1 II. Introduction 5 III. The Context for Policies to Reduce Nutrient Loads to Impaired Waters 8 IV. Watershed Management Practices to Provide Environmental and Economic Benefits 26 V. Methods for Inferring Economic Value 30 VI. An Overview of the Contingent-Valuation Method 33 VII. Survey Design 38 VIII. Survey Execution 49 IX. Empirical Results: Economic Values 52 X. Empirical Results: Management Options 78 XI. Summary and Conclusions 83 XII. References 88 Appendix A Mail-survey instruments and Relative Frequencies A-1 Appendix B Cover Letter and Follow-up Letters for Mail Survey B-1 Appendix C Estimation Techniques: Logistic and Censored Logistic Regression C-1 Appendix D Selected Statistical Tables, Including Recreational Interviews D-1 Appendix E Verbatim Comments to Open-Ended Questions E-1 1 Section I. Executive Summary Restoration of impaired waters is gaining increasing attention from the State of Minnesota. Improving the quality of impaired waters will yield environmental benefits that will also translate into economic and social benefits. The estimation of the economic value of these environmental benefits by assessing the total willingness-to- pay (WTP) of property owners for restoring water quality in impaired lakes within two watersheds in the Upper Mississippi River Basin of Minnesota is the primary objective of this study. The watersheds are the Sauk River (also known as the Horseshoe) Chain of Lakes (Sauk COL) and the Lake Margaret-Gull Lake Chain. Estimating the economic value of improvements in public goods, such as environmental goods and services, requires a method that utilizes non-price (non-market) data. A stated- preference estimation technique known as the contingent valuation method (CVM) is utilized to estimate the WTP of property owners for water quality improvements resulting from reduced nutrient loads, particularly phosphorus. Contingent valuation employs a survey that describes the prospective policy and its effects. The percentages of respondents favoring the proposal at different household costs provide information on how much households value the changes, if at all. Logistical regression is utilized to relate the percentages voting YES with the household costs and other variables. Economic theory suggests the level of support should vary inversely with the costs. The results are consistent with these economic principles. Censored logistic regression allows conversion of these relationships into a valuation function that estimates mean WTP of respondents. 2 The causes of the impairments differ between the two watersheds, so different management options may generate different levels of net benefits. The analysis demonstrates that the watersheds are also different in terms of how property owners in the watershed relate to the impaired lakes. Many property owners are not residents of the watersheds (67% have ZIP codes outside the watershed for Margaret-Gull) and are wealthier and older than the average residents of the area. The pattern is less severe in the Sauk COL Watershed, as about 11% of the property owners have mailing addresses outside of the watershed and Stearns County. The Margaret-Gull Chain has a high degree of surface water as percentage of watershed acreage compared to Sauk COL, and consequently a high proportion (64 percent versus 16 percent of respondents per watershed respectively) of lakeshore owners relative to the overall population of property owners in the watershed. The Margaret-Gull Chain also has many highly-valued lake properties owned by people with high incomes and a large amount of recreational use by lake owners and visitors. A mail survey was sent to a randomly selected sample of 1,500 property owners in the Sauk COL Watershed and to the entire population of 1,044 property owners in the Margaret-Gull Watershed. The total Sauk COL response of 571 over the pilot and main mailings as a percentage of 1,380 potential respondents is 41.4%. The overall response including the pilot was 510 in Margaret-Gull which is 49.9% of 1,022 potential respondents. The simplest specification of the logistic regression explained the percentage voting YES on the proposal with COST and WATERSHED as the explanatory variables. Support for the proposal is significantly higher in the Margaret-Gull Watershed. When additional explanatory variables are included, the watershed is no longer significant because the characteristics of the respondents within the watersheds dominate. Robust results are yielded in that Lakeshore 3 Ownership, Frequency of Lake Use, Perception of Effectiveness of the Policy, and Income were consistently found to be significant at the 1% level under alternative assumptions. The alternative models also generate a range of WTP estimates, with the means for Margaret-Gull substantially higher in all models. Margaret estimates were clustered in the $200- $300 range, while the estimates for Sauk COL ranged in the double digits. The preferred model is converted using censored logistic regression to estimate mean WTP. The mean for respondents in the Margaret sample is $267 and for Sauk it is $17. While the estimates from the censored logistic regression are the most theoretically appealing, the higher mean WTP for Sauk yielded from alternative approaches implies that the $17 value, should be taken as a lower bound. The stark differences between Margaret-Gull and Sauk COL fulfill the methodological goal of studying watershed property owners that are at opposite ends of the spectrum. While the estimated equations for the two watersheds have slightly different coefficients, the extreme differences in WTP result from huge differences in the mean values for the variables between the watersheds. Following the economic theories that are the foundation of the contingent valuation method a mean of $145 between the two watersheds was calculated. However, this mean value is misleading if applied inappropriately to other watersheds or the State of Minnesota in general. Rather, the range of values could be transferred to approximate mean WTP in other watersheds where enough is known about the underlying characteristics of the watershed and the population. For example in other systems with impaired lakes, the closer they are on the Margaret end of the spectrum (high proportion of lakeshore ownership, frequent lake use, high confidence in policy effectiveness - especially protection against future impairment downstream, rather than reversing it - and high income) the closer the mean value will be to the Margaret value. If the 4 characteristics more closely mimic the Sauk, the mean WTP will be much lower. The wide variation of WTP estimates also implies that some people stand to gain a great deal from lake restoration while others will feel the costs are not worth it to them. Respondents provided a great deal of feedback on which land-use management options they would or would not support. While respondents indicated a willingness to participate in best-management practices, there is also a strong tendency to want others (those who benefit the most or contribute the most pollution) to make changes or pay for activities that improve water quality. Management of aquatic plants, particularly curly leaf pondweed, is closely related to the management of nutrients as part of a total lake management program. Responses to a series of questions about aquatic plants reveal which approaches are preferred by different groups. The respondents are concerned about invasive aquatic plants but are not sure about the importance of native plants in resisting the spread of invasive species and in-cycling nutrients within the system.