<%BANNER%>

Willingness-to-Pay for Red Tide Mitigation, Control and Prevention Strategies

Permanent Link: http://ufdc.ufl.edu/UFE0042280/00001

Material Information

Title: Willingness-to-Pay for Red Tide Mitigation, Control and Prevention Strategies A Case Study of Florida Costal Residents
Physical Description: 1 online resource (126 p.)
Language: english
Creator: Lucas, Kriste
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: contingent, florida, logit, red, willingness
Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Harmful algal blooms (HABS) are natural events with ecological and economic consequences, and are referred to locally as ?red tides?. In Florida, Karenia brevis is the algae species that has accounted for nearly all HABs. This species produces potent neurotoxins that can kill marine life and affect the respiratory system of humans. The fact HABs can affect humans is potentially disastrous to Florida, which is heavily dependent on coastal tourism. Several types of strategies for addressing HABS are being researched and implemented, however some strategies are likely to face severe opposition. To determine the potential acceptance of alternative red tide strategies in Florida, mail and online surveys were sent to households in coastal counties. Results can be used to help summarize public opinion, inform policy makers, and evaluate specific programs intended to address the potentially harmful effects of red tide events in Florida.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Kriste Lucas.
Thesis: Thesis (M.S.)--University of Florida, 2010.
Local: Adviser: Larkin, Sherry L.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042280:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042280/00001

Material Information

Title: Willingness-to-Pay for Red Tide Mitigation, Control and Prevention Strategies A Case Study of Florida Costal Residents
Physical Description: 1 online resource (126 p.)
Language: english
Creator: Lucas, Kriste
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: contingent, florida, logit, red, willingness
Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Harmful algal blooms (HABS) are natural events with ecological and economic consequences, and are referred to locally as ?red tides?. In Florida, Karenia brevis is the algae species that has accounted for nearly all HABs. This species produces potent neurotoxins that can kill marine life and affect the respiratory system of humans. The fact HABs can affect humans is potentially disastrous to Florida, which is heavily dependent on coastal tourism. Several types of strategies for addressing HABS are being researched and implemented, however some strategies are likely to face severe opposition. To determine the potential acceptance of alternative red tide strategies in Florida, mail and online surveys were sent to households in coastal counties. Results can be used to help summarize public opinion, inform policy makers, and evaluate specific programs intended to address the potentially harmful effects of red tide events in Florida.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Kriste Lucas.
Thesis: Thesis (M.S.)--University of Florida, 2010.
Local: Adviser: Larkin, Sherry L.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042280:00001


This item has the following downloads:


Full Text

PAGE 1

1 WILLINGNESS TO PAY FOR RED TIDE MIT IGATION, CONTROL AND PREVENTION STRATEGIE S: A CASE STUDY OF F LORIDA COASTAL RESIDENTS By KRISTEN MARIE LUCAS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Kristen Marie Lucas

PAGE 3

3 ACKNOWLEDGMENTS I thank my family and friends for their constant love and encouragement, especially during the last part of my masters program. In particular, I want to thank my parents Michael and Karen, my sister Amber, and Matthew Rembold. Without all their love, help and support, completing this journey would not have been possible. I want to recognize my committee chair, Dr. Sherry Larkin, for her guidance and advice during the progress of this thesis, and for her support understanding during stressful times. I am very thankful to Dr. Charles Ada ms, as we ll, for his valuable input and assistance I want to thank the Florida Wildlife Commission for funding this study. I give special recognition to Dr. Michael Scicchitano, director of the Florida Survey Research Center and his staff, namely Dr. Tra cy Johns and Janet Heffner, for their tireless work on the mail survey. Finally, I extend my gratitude to the Food and Resource Economics Department for the financial support that made my graduate education possible. A special recognition goes to Dr. Jeffr ey Burkhardt, Graduate Coordinator, and Jess Herman for al l their help, advice and consideration.

PAGE 4

4 TABLE OF CONTENTS P age ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 3 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 9 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................... 1 Harmful Algal Blooms (HABs) ................................ ................................ ................................ 1 Problem Statement and Study Objective ................................ ................................ .................. 4 Stated Preference Methodology ................................ ................................ ................................ 6 The Contingent Valuation Method (CVM) ................................ ................................ ....... 6 Probability Model and Evaluation ................................ ................................ ..................... 8 Application to Red Tide Strategies ................................ ................................ .................. 12 Data Collection ................................ ................................ ................................ ....................... 13 Organization of Thesis ................................ ................................ ................................ ............ 18 2 SUMMARY OF SURVEY RESPONSES ................................ ................................ ............. 20 Awareness, Experience, Knowledge and Concern ................................ ................................ 20 Question 1: Are y ou aware of the coastal condition known as red tide? ........................ 20 Question 2: What has been your experience with red tide events in Florida? ................ 21 Question 3: Do you believe each statement is true or false with respect to Florida red tides? ................................ ................................ ................................ ...................... 21 Question 4: How concerned are you, if at all, about Florida red tide events? ................. 21 Information ................................ ................................ ................................ ............................. 22 Question 5: Do you agree or disagree with each statement concerning scientific research on red tides in Florida? ................................ ................................ .................. 22 Question 6: How frequently do you seek information about Flori da red tides? ............. 23 Question 7: How familiar are you with red tide information available from each agency? ................................ ................................ ................................ ........................ 24 Question 8: How frequently do you get Florida red tide information from each source? ................................ ................................ ................................ ......................... 24 Red Tide Strategies ................................ ................................ ................................ ................. 25 Question 10: Do you use plant fertilizers in Florida? ................................ ...................... 25 Question 10A: Would you vote for an X% tax on all fertilizer sales? ............................ 26 Question 10B: If so, how sure are you of this decision? ................................ ................. 26 Question 10C: If not, is there any% that you would vote for? ................................ ........ 27 Question 11: Are you aware of the Beach Conditions Reporting System for the Gulf Coast of Florida TM ? ................................ ................................ ............................. 27 Question 11A: Would you pay a one time donation of $X into this trust fund? ............. 27 Question 11B: If so, how sure are you of this decision? ................................ ................. 28

PAGE 5

5 Question 11C: If not, is there any amount that you would pay? ................................ ..... 28 Question 12: Approximately how many days do you spend at the beach each year? ..... 28 Question 13: Did you pay property taxes in Florida in 2009? ................................ ........ 28 Question 13A: Would you vote for a 3 year county property tax to fund control programs? ................................ ................................ ................................ ..................... 28 Question 13B: If so, how sure are you of this decision? ................................ ................. 29 Question 13C: If not, is there any amount that you would pay? ................................ ..... 29 Question 14: In general, which type of control most appeals to you? ............................ 29 Question 15: Which program would you prefer if the State of Florida only had funds for one? ................................ ................................ ................................ .............. 29 Demographics ................................ ................................ ................................ ......................... 30 Question 16: How long have you resided in Florida? ................................ ..................... 30 Question 17: How many months out of the year do you reside in Florida? .................... 30 Question 18: What is the ZIP code of your residence in Florida? ................................ ... 30 Question 19: How many miles by car do you live from the coast? ................................ 31 Question 20: In what year were you born? ................................ ................................ ...... 31 Question 21: What is the highest level of education that you have completed? ............. 31 Questi on 22: Which of the following describe your race or ethnicity? ........................... 31 income before taxes? ................................ ................................ ................................ ............................ 32 Comparison of Mail and Internet Responses ................................ ................................ .......... 32 3 RESULTS OF EMPIRICAL MODELS ................................ ................................ ................. 54 Overview ................................ ................................ ................................ ................................ 54 Comparison of Binary Models ................................ ................................ ................................ 55 Binary Logit Models for the Prevention Strategy ................................ ........................... 55 Binary Logit Models for the Control Strategy ................................ ................................ 59 Binary Logit Models for the Mitigation Strategy ................................ ............................ 61 Summary of Binary Logit Models ................................ ................................ ................... 63 Ordered Models ................................ ................................ ................................ ...................... 64 Ordered Logit Mode l for Prevention Strategy ................................ ................................ 65 Ordered Logit Model for Control Strategy ................................ ................................ ...... 67 Ordered Logit Model for Mitigation Strategy ................................ ................................ 70 Summary of the Ordered Logit Models ................................ ................................ ........... 72 Non Parametric Estimates of Willingness to Pay ................................ ................................ ... 73 Summary of Empirical Results ................................ ................................ ............................... 73 4 CONCLUSION AND DISCUSSION ................................ ................................ .................... 91 APPENDIX A Mail survey cover letter and questionnaire ................................ ................................ ............. 96 B Revised stated preference questions for online questionnaire ................................ .............. 105 C Open ended comments from online respondents (N = 47) ................................ ................... 109

PAGE 6

6 LIST OF REFERENCES ................................ ................................ ................................ ............. 113 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 115

PAGE 7

7 LIST OF TABLES Table page 2 1 ................................ .................. 37 2 2 True or false statements about red tide events in Florida ................................ .................. 37 2 3 Primary reason for whether respondent was concerned or not abo ut red tides ................. 38 2 4 Level of agreement about scientific research on red tides in Florida ................................ 38 2 5 Familiarity of respondents with various agencies that supply red tide information in Florida ................................ ................................ ................................ ................................ 39 2 6 Frequency of obtaining information by alternative sources ................................ .............. 39 2 7 Description of willingness to pay scenarios ................................ ................................ ...... 40 2 8 Percentage of willingness to pay responses, by price level ................................ ............... 40 2 9 Responses by region ................................ ................................ ................................ .......... 41 2 1 0 Percentage of respondents that have e xperienced effects of a red tide event by survey ... 41 2 11 Comparison of responses regarding red tide knowledge ................................ ................... 42 2 12 Primary reason for whether respondent was concerned or not about red tides by survey ................................ ................................ ................................ ................................ 43 2 13 red tides in Florida between mail and internet respondents ................................ ............... 44 2 14 Comparison of the familiarity of respondents with various agencies that supply red tide information in Florida by survey ................................ ................................ ................ 44 2 15 Comparing the percentage of respondents that respondents obtain information by alternative source and survey ................................ ................................ ............................. 45 2 16 Comparison of willingness to pay scenarios by survey ................................ .................... 45 2 17 Comparison of WTP responses by bid level and survey ................................ ................... 46 2 18 Comparison of responses related to WTP questions by survey ................................ ......... 46 2 19 Comparison of respondent geographic characteristics and tenure in Florida by survey ... 47 2 20 Comparison of age, education, income and race by survey ................................ ............... 48

PAGE 8

8 3 1 Description of dependent variables for the WTP models ................................ .................. 76 3 2 Description of independent variables for the WTP models ................................ ............... 77 3 3 Estimated binary logit models for the proposed prevention strategy ................................ 78 3 4 Estimated binary logit models for the proposed control strategy ................................ ...... 79 3 5 Estimated binary logit models for the proposed mitigation strategy ................................ 80 3 6 Estimated marginal values for the binary logit models of the prevention strategy ........... 81 3 7 Estimated marginal values for the binary logit models of the control s trategy ................. 82 3 8 Estimated marginal values for the binary logit models of the control strategy ................. 83 3 9 Estimated ordered logit models for the prevent, control, and mitigation strategies .......... 84 3 10 Estimated marginal values for the ordered logit models of the prevention strategy ......... 86 3 11 Estimated marginal values for the binary logit models of the control strategy ................. 87 3 12 Estimated marginal values for the binary logit models of the mitigation strategy ............ 88 3 13 Summary of estimated marginal values for the extreme levels from the ordered logit models for each strategy ................................ ................................ ................................ .... 89 3 1 4 Turnbull estimates using Model 1 data ................................ ................................ .............. 90

PAGE 9

9 LIST OF FIGURES Figure page 2 1 Concern over red tide events in Florida (N = 1,302) ................................ ......................... 49 2 2 How frequently respondents seek information about red tides (N = 1,330) ...................... 49 2 3 Dependence on coastal water quality (N = 1,338) ................................ ............................. 49 2 4 Share of respondents willing to pay a fertilizer tax (N = 1,326) and how sure they are of that decision (N = 789). ................................ ................................ ................................ 50 2 5 Share of respondents willing to donate to a trust fund (N = 1,320) and how sure they are of that decision (N = 474). ................................ ................................ ........................... 50 2 6 Share of respo ndents willing to vote for a property tax (N = 1,343) and how sure they are of that decision (N = 653). ................................ ................................ ........................... 50 2 7 Respondents preferred strategy (N = 1,311) ................................ ................................ ...... 51 2 8 Frequency of Months Residing in Florida (N=1,439) ................................ ....................... 51 2 9 Distribution of highest level of formal education among respondents (N = 1,431) .......... 52 2 10 Distribution of household income of respondents (N = 1,322) ................................ ......... 52 2 11 Comparison of frequency that respondents seek information about red tides by survey format ................................ ................................ ................................ ..................... 53 2 12 Comparison of dependence on coastal water quality by survey ................................ ........ 53

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science WILLINGNESS TO PAY FOR R ED TIDE MITIGATION, CONTROL AND PREVENTION STRATEGIE S: A CASE STUDY OF F LORIDA COASTAL RESIDENTS By Kristen Marie Lucas December 2010 Chair: Sherry Larkin Major: Food and Resource Economics Harmful algal blooms (HABS) are natural events with ecological and economic consequences worldwide. In Florida, Karenia brevis is the algae species that has accounted for nearly all HABs. Karenia brevis produces potent neurotoxins that can kill fish and ma rine mammals and become airborne and affect the respiratory system of humans. The fact that such like Florida that is heavi ly dependent on coastal tourism. A variety of strategies for addressing HABS have be en implemented around the world, and these strategies can be broken down into three main categories: prevention, control and mitigation. S ome strategies are likely to face severe opposition so t o determine the potential acceptance of alternative red tide strategies in Florida, mail and online surveys were sent to households in coastal counties. The questionnaire uses a polytomous choice contingent valuation framework to estimate the willingness to pay for t hree hypothetical strategies : a fertilizer tax to improve general water quality (prevention strategy that is uncertain for red tides), a trust fund donation for a beach conditions reporting service (mitigation strategy designed to change behavior), and a p roperty tax to fund pilot control programs (biological or chemical). For each strategy both binary and ordered logit models were

PAGE 11

11 estimated Results can be used to help summarize public opinion, inform policy makers, and evaluate specific programs intended to address the potentially harmful effects of red tide events in Florida.

PAGE 12

1 CHAPTER 1 INTRODUCTION Harmful Algal Blooms (HABs) In most marine and fresh water environments, microscopic, plant like organisms occur naturally in the surface layer of the water. These organisms, which are referred to as phytoplankton or microalgae, form the base of the food chain upon which nearly all o ther marine organisms depend. An algal bloom occurs when there is an increase in concentration of phytoplankton to the extent that it dominates the local planktonic community. This can occur for several reasons. Most often, an increase in the nutrients the algae feed on, or some environmental condition like a change in water temperature or patterns in water circulation, are the cause of the population explosion. These are naturally occurring events, however many scientists agree that human activity can exacerbate the severity of bloom events. Many algal blooms are relatively benign in their effects; however, depending on the species of algae involved, some blooms can have considerable negative impacts on the affected area. The extent of the impacts ca n vary depending on a number of factors, including the length and size of the bloom. Larger blooms have been known to last for more than a year and stretch along several miles of coastline. Harmful algal blooms (HABs) occur when algal blooms produce toxi c or otherwise harmful effects on humans, fish and marine mammals, and the surrounding ecosystem. Some species of algae that are responsible for HABs release powerful toxins in to the water and air. These toxins can paralyze fish and marine mammals, caus ing them to drown. They can also cause toxicity poisoning in humans that eat shellfish caught in affected waters. One common neurotoxin that can be found in contaminated shellfish is domoic acid, which can cause amnesic shellfish poisoning in humans. Ex treme cases, though rare, can lead to coma or death. Some HABs release toxins not only into the water, but into the air as well.

PAGE 13

2 Airborne toxins are responsible for causing or exacerbating respiratory ailments in humans, depending on the severity of the b loom. They can also cause coughing scratchy throats, burning eyes, and skin irritation. Even blooms that do not release toxins have the potent ial to cause massive fish kills by depleting the dissolved oxygen in the water The negative impacts of HABs not only affect the environment and human health but also affect local economies (Jin et al. 2008) One case study estimated the negative economic impacts of HABs in the United States at $82 million annually (Hoagland and Scatasta 2006) This study found that the commercial fishing industry has been especially hard hit, with annual losses estimated at $38 million. The tourism and recreation industry losses were estimated at $4 million per year, mainly due to beach closures during HAB events. These resea rchers also estimated the annual health related costs of HABs to be $37 million per year. In addition, it is very costly to manage and control bloom populations. Approximately $3 million per year has been spent on coastal monitoring and management Nearly all HABs in Florida are caused by the species Karenia brevis Generally, this algae Karenia brevis releases a potent neurotoxin into the water that kills fish and, in severe cases, dolphins and manatees as well. It also releases airborne toxins that make it difficult to breathe which, depending on the severity of the bloom and other environmental conditions (e.g., wind direction and speed), makes it virtually impos sible for residents and tourists to participate in marine based activities (e.g., beach going, diving, and fishing). Florida red tide also can also result in the closure of shellfish harvesting areas due to the release of toxins that make shellfish danger ous to consume (Fleming et al. 2009). Florida red tide occurs nearly every summer along

PAGE 14

3 the Gulf Coast and causes millions of dollars in damage and lost revenue (Morgan et al. 2010 ; Morgan et al. 2009; Larkin and Adams 2007 ; Backer 2009 ). In the summer of 1971 the Tampa Bay area experienced a particularly bad red tide event, which prompted the first study of the wide coastal communities. It was estimated that the 1971 red tide caused $20 million in lost revenue in both the commercial/recreational fishing and tourism industries in seven coastal counties alone. In addition to this lost revenue, these counties incurred thousands of dollars in clean up costs from removing dead fish and debris from the coastl ine. The study forecasted that a red tide event of similar magnitude in the future could cause up to 40% more in economic damage (Habas and Gilbert 1974). In a 2007 study, it was found that beach attendance during a red tide event decreased by 13.5%, or 50,000 visitors, in the two counties under study (Larkin and Adams 2007). In addition, a study in 2006 found that hospital admissions of patients with respiratory illnesses increased significantly during a red tide, adding to the burden of local health ca re facilities (Kirkpatrick et al. 2006). Clearly, HABs are costly in environmental, health and economic terms. Fortunately, scientists have developed, and continue to research, a variety of methods for managing bloom events. HAB management practices can be divided into three general categories: mitigation, control and prevention strategies. It is important to understand the difference in these strategies. Mitigation strategies focus on minimizing the effects of a bloom on humans, the environment and the economy after the bloom has already occurred. This can include, but is not limited to, monitoring coastal conditions and disseminating that information to the public. Mitigation strategies are proven to be effective, but they do not take any direct acti on against a bloom population.

PAGE 15

4 Control strategies focus on managing a bloom after the bloom has occurred by reducing biological or chemical controls. Biological co ntrols mainly involve using a predator species that will feed on the algae, or one that will complete for the same nutrients the algae is feeding on. A chemical control involves applying a natural material that attaches to the algae and removes it from su rface waters. Control practices have been effective in small scale lab testing, but have not been tried in a large scale application in the United States. Finally, prevention strategies differ from the previous two strategies in that they attempt to add ress the problem of algal blooms before they occur. They aim to reduce the frequency and the severity of future bloom events and, thus, are long term HAB management strategies. Most focus on reducing human activity that increases the amount of nutrients in coastal waters. Prevention strategies are still largely untested for HABs; and, since blooms are natural events whose causes are still relatively unknown, it is unclear whether or not these strategies would be effective. Problem Statement and Study Obj ective Though HABs occur worldwide, this study focuses on HABs occurring along the Florida coast. In Florida, the most common HAB is known as a red tide, and is caused by the species Karenia brevis. In a state like Florida, where the economy is heavily dependent on the commercial fishing and tourism industries, a red tide can be a potentially catastrophic environmental and economic event (Larkin and Adams 2007). A variety of HAB management practices have been implemented around the world and in the stat e and much important research is still being done into these, and new, strategies. However, it would be folly for researchers and administrators to assume that all strategies will be equally acceptable to the public. Since it is Florida residents who wil l ultimately be paying for any management practice implemented in the

PAGE 16

5 state, it is important to understand what type of strategy is most appealing to them (Morgan et al. 2008). This type of information will allow researchers to focus their time and money on a strategy that is likely to be accepted by the public. In addition, if the public is particularly adverse to a method that scientists believe to be the most effective means of managing bloom populations, administrators may be able to change their opin ion through education and dissemination of information. The goal of this study is to determine public preferences for three alternative red tide mitigation, control and prevention strategies. This accomplished by administering a survey to residents in co astal counties where red tides are a common occurrence. familiarity with and use of red tide information supplied by alternative a gencies, organizations and media outlets. To evaluate these responses, residents were also asked about their fertilizer use, beach use and dependence on coastal water quality, in addition to socio demographic characteristics. The survey also uses a state d preference methodology to evaluate the three types of strategies. In particular, three willingness to pay (WTP) scenarios were presented in random order for evaluation: a fertilizer tax to improve general water quality (prevention strategy that is uncer tain for red tides), a trust fund donation for a beach conditions reporting service (mitigation strategy designed to change behavior), and a property tax to fund pilot control programs (biological or chemical). Respondents were first asked whether or not they would be willing to pay a specific amount and then asked how sure they were of the response and allowed to provide an alternate amount they would be willing to pay. This sequential and multiple response format allowed for the use of either dichotomou s choice or ordered preference models, which are

PAGE 17

6 standard for analysis of this type of stated preference data. In addition to providing WTP estimates, t he results of this study will be used to help summarize public opinion, inform policy makers, and evalu ate specific programs intended to address the potentially harmful effects of red tide events in Florida. Stated Preference Methodology The Contingent Valuation Method (CVM) Most environmental goods and services, water quality or environmental preservation are not traded i n markets. Therefore their economic value cannot be revealed through market prices. The only option for assigning monetary values to them is to rely on non market valuation methods. The c ontingent valuation method (CVM) is a type of non m arket valuation technique market good or service It is a valuable tool for performing cost benefit analysis of environmental projects, such as ecosystem restoration projects The CVM is referred to as a stated preference technique, because it asks people to directly state their values or preferences, rather than inferr ing values from actual choices as revealed preference (Carson et al. 1997). These stated pre ferences are discovered through the implementation of carefully designed and administered sample surveys. In these surveys a description of the good or service in question is provided, after which some hypothetical scenario in which the institutional mech anism that will provide or finance the good is introduced. The willingness to pay (WTP) is then elicited through a question taking on one of two basic formats: an open ended question asking the respondent what is the maximum amount they would be willing t o pay for the good/service in question or a dichotomous or discrete choice format in which the respondent is presented with a price (or cost) for the good/service in question

PAGE 18

7 and is then asked whether they would be willing to pay that amount (i.e., a yes, response) In 1993 in wake of the Exxon Valdez oil spill, a panel of economists convened with the support of the National Oceanic Atmospheric Administration ( NOAA ) to ev aluate the validity of using the CVM method to estimate non market values for environmental goods or services. This panel created a list of guidelines that should be followed to ensure that WTP estimates from a CVM study are accurate (i.e., unbiased) The NOAA panel suggests t hat, for example, a closed ended, discrete c hoice format should be used for several reasons: it is more realistic in that individuals typically make decisions faced with fixed prices; it provides less of an incentive to engage in strategic behavior; and it provides a more clear cut decision rule as it is analogous to a referendum ( i.e., has the support of the population ). the institutional mechanism that will provide and finance the non market good. The payment vehicle should be familiar, credible and feasible. With many CVM studies the payment vehicle of choice is some form of taxation. The respondent is generally asked if they would be willing to vot e for a referendum (i.e., a closed ended response format) that would implement the tax in order to generate funds to provide or conserve the environmental good or service in question. However, the payment vehicle can also take the form of an increase in t he price (or cost) of a related market good, or the respondent can be asked if they would be willing to donate a specific amount to a trust fund or non governmental group (NGO). Several approaches have recently been developed that incorporate uncertainty i nto CVM studies. This is important because people are being asked to evaluate something for the first

PAGE 19

8 time; there is likely to be some uncertainty, since they may be unfamiliar about how to evaluate this new good. For example, an approach was employed b y Li and Mattsson (1996) that used a two step method. They first used a tradi tional dichotomous choice WTP question but then included a follow up question that required the respondent perform a "post decisional" rating of the certainty of the ir response t o the WTP q uestion. This certainty rating wa s incorporated into the empirical model that was used to estimate the WTP for the sample respondents. The result reduced both the mean WTP and th e variance of the estimated WTP; as such this method provides a mor e conservative estimate of willingness to pay. Conservative estimates of WTP are preferred since they are needed to ensure that a proposed program will produce benefits that exceed costs. Probability Model and Evaluation Using a dichotomous choice model e valuation format, as recommended by the NOAA panel (i.e., asking for a yes or no response to a proposed program, at a given price or cost to the respondent, which would help to provide a non market good), allows for modeling the choice as follows: (1) where Pr is a probability, Y = 1 corresponds to respondents that were WTP the price specified for that particular stra tegy ( Y = 0 if they were not) and is the logistic cumulative density function that captures all the factors believe d to be correlated with the probability of being WTP. In particular, is a vector of the independent variable coefficient estimates and X is a vector of the independent variables believed to explain the probability of being WTP. In order to account for the uncertainty in their responses and obtain a more conservative estimate of WTP different models can be estimated if respondents are asked about how sure they were about their decision,

PAGE 20

9 response by asking them if they were unsure, somewhat sure or very sure, the following three models can be specified to capture a range of WTP estimates : Model 1: (2) Model 2: (3) Model 3: (4) The models vary the level of commitment to the alternative red tide strategies as measured by the mos leas t inclusiv e. The use of three binary models allows for the calculation of thre e mean WTP estimates (e.g., Welsh and Poe 1998 ). The benefit of this approach is the ability to calculate WTP estimates for different measures of commitment and to see if the variables e xplaining WTP change as the definition of WTP changes. The WTP was first (5) where is a row vector for the sample means of the independent variables (1 is used for the constant term), variables, and 0 is the coefficient estimate for the price variable. The advantage of this approach is that it can be used to evaluate WTP estimates for different subgroups of respondents. The disadvantage is that this approach does not produce a reliable estimate of statistical significance ( Loomis and Gonzales Caban 1998)

PAGE 21

10 T he second WTP was then estimated using the Delta method. With this method, a simple binary logit model is estimated with the constant as the only explanatory factor in the model. A Wald test is then conducted on the WTP estimate. The Delta method has as its advantage the ability to produce a WTP estimate that can be statistically test ed for difference with zero and that it provides an estimate of the variance that can be used to generate a 95% confidence interval around the result (Dumas et al. 2007) Fin ally, the third WTP was estimated using the Turnbull Lower Bound method, which is a non parametric approach. To calculate the WTP using the Turnbull method, you only need used in the to verify that Fj < Fj+1. If Fj is not less than Fj +1 then the Nj and Nj+1 need to be pooled together, as well as Tj and Tj+1, to ensure that there is a monotonically increasing CDF. Once increasing Fjs are calculated, Fk+1 is set equal to one and F0 is set equal to zero. Then fj+1 = Fj+1 FJ is calculate d for each price level. These numbers represent the probability that WTP falls between price j and price j+1. These probabilities are multiplied by the lower price level (j) which provides a WTP at each price level and, when summed, provides the expected value (or lower bound) of WTP, ELB (WTP). The corresponding variance of the lower bound is calculated as: (6) While alternative WTP estimates can be obtained from redefining the definition of a and

PAGE 22

11 valuation studies is that it was adopted in response to concerns related to the use of the dichotomous choice framework for non market valuation, which was outlined in the Report of the NOAA Panel on Contingent Valuation. The multiple response question format, and in particular one that included an option for not supporting the proposal in general, can be useful for cases where there are a sufficient num ber of responses in each category (enough for the model to estimate distinct effects of each explanatory variable for each price level). For an ordered response model, assume that: (7) where X is a vector of explanatory variables, is a vector of associated parameters a nd u is a logistic random variable with mean zero and unknown scale parameter. Also assume that: (8) where the J are thresho ld parameters differentiating each response level and 0 is normalized to zero so that J 2 threshold parameters are estimated. Using this arrangement, the probability of observing the different ordered outcomes for each i are then calculated using a logit model: (9) As the sign and magnitude of the estimated coefficients for an ordered response model are not directly interpretable, the marginal effects are generally used to discuss the impacts of the explana tory variables on the probability of observing the different outcomes. For a discrete explanatory variable x i the marginal effects are estimated by taking the difference between

PAGE 23

12 estimated probabilities when the discrete e xplanatory variable x i takes the value of 0 and 1 holding all other variables constant, usually at their means and are calculated using the equation: (10) Application to Red Tide Strategies In this study a dicho tomous choice (DC) model was first used to estimate the initial scenario (Model 1). A separate model was run for each type of management strategy. For the mail survey, the response data for the dependent variable was obtained dire ctly and the price they were asked to consider was included as an independent variable in the model. For the internet survey, a comparable model could use the zero values as ables include d traditional demographics ( e. g., age, education, ethnicity, income, length of Florida residency, and how many months the respondent resides in Florida per year), location (i.e., region of residence, number of miles from the coastline that the residence is located), level of concern with red tide, level of dependence on local water quality, and the order in which the particular scenario in question appears in the survey. In addition, each model included one or more distinct variable s to captur e strategic bias that is based on whether the respondent will be affected by the proposed strategy (e.g., if they maintain a lawn, are familiar with the beach reporting system in the Southwest region, or if they paid property taxes in Florida last year). In the mail survey, respondents were asked about their level of uncertainty if they used to recode the answers to the WTP questions in order to determine a more accurate (i.e., conservative) estima te. Then, a model was followed by a certainty assessment of their response were coded as (Model 3) The second model

PAGE 24

13 This protocol was similar to a study by Vossler et al. in which the respondents from a survey asking if they would vote for a referendum that would implement a tax on hom e values were able to answer yes no or undecided and the (Vossler et al 2003) The models were estimated using LIMDEP 8.0. The models were compared to one another (within each strategy type ) using various statistics on model fit (i.e., chi squared statistics, pseudo R2 statistics, and percent correctly predicted), as well the number of statistically significant variables and the stability of the signs on those variables. In both the mail and the internet surveys the responses to the WTP scenarios can be modeled using an ordered response framework. In the mail survey this was possible by using the level of certainty responses to order the yes responses ( yes ). Similarly, higher values, or the value could be estimated using a gen eralized least squares methodology The ordered models can also be estimated using LIMDEP. In addition, WTP estimates can be defined and compared to the estimates from the discrete models. Lastly, these models generate marginal values for each ordered le vel being modeled. This information can be used to determine how each explanatory variable affects the probability of a coastal resident being not WTP or very sure of their WTP. Data Collection to pay for diffe rent red tide management practices, 14,400 mail surveys were sent to residents in three different coastal regions in Florida where red tide is a common occurrence. The number of surveys sent to each region was

PAGE 25

14 determined based upon the population from the most recent census. A total of 1,674 surveys were sent to the northwest region, which included the following four counties: Gulf, Franklin, Bay and Okaloosa. Along the Gulf of Mexico in the southwest region, 6,624 surveys were sent to residents in the f ollowing four counties: Manatee, Sarasota, Charlotte and Lee. Finally, in the northeast region, 6,102 surveys were sent to residents in St. Johns, Flagler, Volusia and Brevard counties. To address order bias, 18 versions of the questionnaire were develop ed (which is discussed further below) such that 800 questionnaires of each version were sent. The questionnaires were color coded based on region because the pretest of 100 revealed that the majority failed to provide their zip code. Since regional diffe rences are hypothesized, the color coding was added to the survey implementation protocol. The internet survey was administered through Expedite Media Group (EMG), which had a total of 692,431 email addresses of residents located throughout the 12 county study region. EMG maintains email addresses for marketing purposes but organizations also use the agency to send newsletters and press releases in addition to solicitations or advertisements. EMG designed the invitation and sent a total of three messages The questionnaire was organized into four sections. The first section of the survey (titled s with a series of questions designed to experience with it. Respondents were also asked a series of true or false questions regarding the causes and effects of red tide to determine their level of knowledge of the phenomenon. They were then asked about their level of concern for red tide events and their main reason why in a follow up question.

PAGE 26

15 pinion on nine issues related to scientific research on red tide. Respondents were then asked about how frequently they seek out information on red tide. These questions were designed to evaluate the awareness of and frequency of use of existing red tide information sources for the community in order to determine if any improvements could be made in the dissemination of information. Lastly, this section asked about the level of dependency on coast water quality. The third section included the evaluation mitigation, control and prevention. In the mail survey, each scenario (i.e.,, red tide strategy) has three different versions based on price: one with a low price level, one with a medium price level, and one with a high price. In addition, the scenarios had to be randomized with respect to order resulting in eighteen versions of the questionnaire. If a respondent received a survey with high price levels this meant that for every scenario the price level presente d was high. For example, one version of the survey presented the prevention strategy first, second and third, all with low price levels. In this way order bias controlled for. Respondents were also instructed to respond to each scenario independently of the others, that is, they were asked to evaluate each as if they were the only option available. For each scenario, the respondents were presented with background information describing the type of management strategy, its risks and its benefits. In add ition, a behavioral or experience question specific to each scenario was asked before each willingness to pay question was introduced to better assess strategic bias. After the willingness to pay section, they were asked which of the three scenarios they preferred most, if any. Each of these scenarios are discussed in turn. wide retail tax on all fertilizer sales was proposed. It was explained that the tax on fertilizer was chosen in order to discourage its u se in

PAGE 27

16 coastal areas where runoff into coastal waters could provide the increased nutrients needed for an algal bloom to occur, spread or intensify. It was also explained that regardless of its impact on future red tide events, it is believed that this sce nario would ultimately help to improve the overall quality of coastal waters. The respondents were presented with the following willingness to pay scenario: Potential Prevention Strategy: Establish a state wide retail tax on fertilizer that would encourage a reduction in fertilizer use and raise funds to pay for continual monitoring of coastal water quality, and research to determine water quality improvements. If no measurable improvements were found within three years, the law would be automatica lly repealed. Depending on which price level the respondent received, they were asked if they would be willing to vote for a 5%, 10% or 15% tax on fertilizer sales. Additionally, the respondent was asked if such a fertilizer tax would affect them more due to personal fertilizer use, with the expectation that this information would be a determining variable in the willingness to pay for this strategy. A real strategy since it invol ves monitoring coastal conditions and broadcasting this information to the public, and such a system already exists in some coastal areas. It was stressed to respondents that this type of strategy would accrue benefits regardless of red tide conditions in the area because the reporting system also monitors tidal conditions, weather conditions and a whole suite of additional coastal information. The following scenario was presented: Potential Mitigation Strategy: Establish a Beach Conditions Reporting Ser vice Trust Fund to support the training of observers, initial equipment expenditures and maintenance of an electronic reporting system. It is anticipated that one time donations to this fund would be sufficient to establish and support this program over th e next three years. Only people who donate would able to access the system. Again, depending on the randomized price level, the respondent was asked if they would be willing to pay a one time donation of $5, $15 or $25 for access to the reporting service f or three

PAGE 28

17 years. Since this system has been launched at some beaches in the Southeast region, respondents were asked if they were familiar with the existing system. Some pre test respondents from the Southeast region were familiar and, therefore, not will ing to pay since it is available for free. This is important information for our study since funding for the current system is not guaranteed but there is a potential for it to be expanded state wide. year prop erty tax to fund pilot red tide control methods was proposed. Respondents were told that control programs have been widely successful on a small scale level and have been used in different countries, however research on a larger scale in the U.S. is still needed. Part of the funds raised would go towards pilot testing for ecological impacts from large scale applications. The scenario was worded as follows: Potential Control Strategy: Establish a 3 year tax on the assessed value of all taxable property t o fund red tide control programs, including pilot testing. If no measurable improvements were found within three years, the law would be automatically repealed. Depending on price level, it was asked if respondents would be willing to vote for a three year tax of $5, $10 or $15 per $100,000 of assessed value of all taxable property at the county level, for the funding of a local red tide control program. It was also asked if the respondent paid property taxes in Florida last year, as this could affect thei r response. In addition, each WTP scenario had follow up questions regarding the certainty of the respondents answer. Respondents that answer ed s any amount t hey would be willing to support using in an open ended format. Respon dents that answer ed yes were asked to indicate whether they were very sure, somewhat sure or unsure of their response. In addition, t ended response) the maximum amount they would be willing to pay would be.

PAGE 29

18 The online questionnaire differed in the treatment of the WTP question. The internet survey asked for their maximum willingness to pay for each scenario and they were provided with five response choices (i.e.,, a closed ended format). Th ree of the choices were the three levels used in the mail survey, which had been based on the incidence of pre test responses to the level of the three values presented. The final section of the questionnaire asked a series of questions to allow the models to control for various socio demographic characteristics. Th in the survey included the length of residency in Florida, location in the state (including distance to the nearest coast), and the age, education, ethnicity and household income of the respondent. Internet respondents were also provided a space to enter feedback regarding the content of the survey. Organization of Thesis In the following chapters the responses from the surveys, as well as the empirical results and WTP estimates will be discussed in depth. In Chapter 2 the responses from the mail surveys will be summarized individually, in order of their appearance in the survey. At the end of the chapter are several tables and figures that summarize these results in a clear and concise fashion. Also, the cover letter and questionnaire, as they appeared to the respondents, are shown in Appendix A. The final section of Chapter 2 will compare the responses from the mail survey to those from the internet survey. Chapter 3 is an in depth discussion of the empirical resul ts from the estimation of the models detailed above. This chapter includes tables that define and describe the independent and dependent variables, as well as a series of tables that summarize the estimated coefficients and marginal effects for each type of model. The nine binary models are

PAGE 30

19 discussed first, followed by a discussion of the three ordered models. The estimated WTP values are also addressed in this section, as well as the non parametric calculations of WTP (i.e., the Turnbull WTP). Finall y, Chapter 4 is a conclusion and discussion of results, implications, limitations and future steps of the study.

PAGE 31

20 CHAPTER 2 SUMMARY OF SURVEY RESPONSES public preferences for different red tide management practices, 14,40 0 mail surveys were sent to residents in three different coastal regions in Florida where red tide is a common occurrence. Out of those, 1,454 were returned, giving a response rate of 10.1%. This response rate is considered to be conservative since the s urveys were sent by bulk mail to reduce costs, which mean that undeliverable questionnaires were not returned. The results from the mail survey are summarized in detail in this section. Each question is addressed individually and in order of appearance i n the survey. The verbatim question, as it appeared in the survey, is also presented. The cover letter and questionnaire are shown in Appendix A. The final section of the chapter compares the responses to the mail and internet versions of the questionna ire. Since only 115 completed surveys were obtained from the internet version, responses are not discussed in detail here. Awareness, Experience, Knowledge and Concern knowle dge of red tide, their level of concern with the issue, and their experience with it. Question 1: Are You A ware of t he Coastal Condition Known a s Red Tide? A total of 93.5 % of respondents answered yes to this question, indicating that they were aware of red tide, while only 6.5 % responded no (N = 1,431). Respondents who answered no were asked to skip directly to the demographic section of the survey, as the survey was intended to target those reside nts who had at the very least a passing knowledge of red tide. All subsequent questions, summarized below, were directed to those respondents who answered yes

PAGE 32

21 Question 2: What Has Been Your Experience with Red Tide Events i n Florida ? For this questio statements regarding their personal experience with red tide (Table 2 1) A significant portion of respondents (82 .0 %) reported having experienced the smell of dead fish at the beach a t some point in the past, as well as having seen dead animals on the shore (73.9%). In contrast, very few respondents indicated that they had changed a hotel or restaurant reservation (6.2% and 19.2%, respectively). Question 3: Do You Believe Each Stateme nt Is True or False with Respect t o Florida Red Tides ? In question three respondents were asked a series of true or false questions regarding the causes and effects of red tide in order to determine the ir level of knowledge of red tide. third statements regarding the safety of seafood consumption during a red tide event. The majority of respondents either answere d the statement incorrectly or indicated they did not know the correct answer (Table 2 name for red tide (77.1%) and if red tides are the same worldwide (65.1%). However, most responden ts answered correctly when asked if red tide conditions vary greatly within a small area (92%) and if people with asthma are more likely to experience health effects from red tide (78%). Many of these statements are directly comparable to those asked to 1 ,000 residents of Manatee and Sarasota counties in 2000 (Larkin and Adams 2007). Question 4 : How Concerned Are You, If at All, a bout Florida Red Tide Events? were ask ed (Figure 2 1) Respondents were then asked for the reasoning for their answer in a

PAGE 33

22 follow up question. Out of the 1,302 respondents who answered this question 23.6 % in dicated that they were generally unconcerned about red tides in Florida, while the remaining 76.4 % indicated that they were at least somewhat, if not very, concerned about the issue. Those that responded no to the initial question were asked the follow up question about their primary reason (Table 2 3) Of the 307 respondents who indicated they were not concerned in the initial question, 287 went on to answer the follow up question. The response rates for each answe r are listed in the table below, but the majority of respondents indicated that they were unconcerned because red tides are a natural occurrence (44.6%). The respondents that answered yes to the initial question were also asked a similar questi on regarding the reasoning behind their response. Of the 995 respondents who answered yes to the initial question, 858 went on to complete the follow up question. These respondents man health (31.9%). Information Next, a series of questions regarding information sources for red tide were presented. These questions were designed to determine the information seeking behaviors of respondents as well as to evaluate the quality of exist ing red tide information sources for the community in order to determine if any improvements could be made in the dissemination of information. Question 5 : Do You Agree or Disagree w ith Each Statement Concerning Scienti fic Research on Red Tides i n Florida ? This question presented respondents with a series of statements that were designed to determine their feelings towards scientific research on red tide. Respondents were asked to

PAGE 34

23 2 4. The first four statements asked respon dents to comment on the quality of research on red tide and the quality of monitoring and prediction systems. The average ratings for these for questions ranged from 3.00 to 3.86, indicating that respondents were largely unaware or unconcerned about the i ssue. The next three questions asked respondents to comment on how important they felt it was that research be done on the different effects that red tide has on human health, on people in general, and on marine animals. The statement regarding human hea lth received the highest average Likert rating (4.54), indicating that people strongly agreed that learning about how red tide affects human health is important. The average ratings for the statements on red tide effects on people and marine animals were 4.46 and 4.47, respectively. In the final statement respondents were asked whether they thought that determining the costs and benefits of different red tide strategies is important. The average Likert rating for this statement was 4.29 indicating that r espondents agreed that this issue was important. Question 6: H ow Frequently Do You Seek Information a bout Florida Red Tides? Question number six aimed to determine the frequency with which respondents search for information about red tide in general (Figu re 2 2) Overall, nearly three quarters of respondents actively seek out information. In terms of frequency, the majority of respondents ( 43 .1%) responded that they only look for information when a red tide affects near shore waters. Another 2.1% indica ted that they only look for information when something new is reported about Florida red tides and 7.6% indicated that they look for information about Florida red tides on a regular basis to see what is new Finally, only 23.1% responded that they never l ook for information about red tide events (N=1,330).

PAGE 35

24 Question 7: How Familiar Are You w ith Red Tide Information Available f rom Each Agency ? Respondents were asked about their familiarity with eight different sources of information on red tides in Florida They were able to choose whether they were or familiar (Table 2 5). The agencies in question were the Florida Fish & Wildlife Institute (FWRI), major universities in the state of Florida, Mote Marine Lab, the Sierra Club, Solutions to Avoiding Red Tide (START), Florida Red Tide Coalition (FRTC), Florida Red Tide Alliance FRTA) and the Beach Conditions Reporting System ( BCRS) TM The majority of respondents indicated that they were not at all familiar with any of the organizations listed. The source with the highest percentage of respondents who were very familiar or somewhat familiar with it was the BCRS (15.2% and 39.7% respectively). Mote Marine Lab closely followed the BCRS with 13.5%. The organization that respondents were the least familiar with was START (91.1% indicated that they were not at all familiar). Following closely behind START were FRTA with 90.8%, th e Sierra Club with 89.6% and FRTC with 88.9%. These results, however, are likely regional in nature since not all organizations operate equally in every area of the state. Question 8: How Frequently Do You Get Florida Red Tide Information f rom Each Sourc e? In this question respondents we re asked to indicate how often they search for information (Table 2 6) The information source most commonly used by e was public forums, with

PAGE 36

25 Question 9: How Dependent Are You on Coastal Water Quality a nd Quantity? This question ndence on coastal water quality. Of the 1,338 respondents who answered the question, 6.9 % were not at all dependent (Figure 2 3) 52.6 % of respondents were somewhat dependent, and the remaining 40.5 % were very dependent. Red Tide Strategies Following t he information section was a series of three contingent valuation (i.e., WTP) questions for the purpose of evaluating public preference for the different types of management practices. There was a scenario for each type: mitigation, control and prevention For each scenario, the respondents were presented with background information describing the type of management strategy, its risks and its benefits. In addition, a behavioral or experience question specific to each scenario was asked before each willi ngness t o pay question was introduced. Finally, after the willingness to pay section, they were asked which of the three scenarios (if any) they approved of the most. Since there were three different price levels for each scenario, the price levels are d in the descriptions that follow In addition, the order in which the scenarios appear within this summary is not indicative as to the order in which they appeared within the actual surveys (in reality, the order of appearance was randomized among surveys). The three scenarios are summarized in Table 2 7 but the questionnaire contains the exact questions and the background information provided on each (Appendix A). Question 10: Do You Use Plant Fertilizers i n Florida ? This program relies on funds generated from those who buy fertilizer so those individuals might be less likely to support the program since it will cost them more Conversely, their response since they would not have to pay. In stated preference questionnaires it is critical

PAGE 37

26 to remind the respondent of how the proposed program might affect them financially, so this question is needed to account for their use of fertilizers. A fter pr efacing the question with this information, 54.9 % indicated that the used plant fertilizers (N = 1,321 ) Question 10A : Would You Vote for a n X % Tax o n All Fertilizer Sales? Sixty percent of respondents who answered the question indicated that they would be willing to vote for the scenario described above, while 40 % indicated that they would not be willing resp onses broken down by price level are provided in Table 2 8. Question 10B : If So, How Sure Are You o f This Decision? Those that answered yes were then asked how sure they were of that decision using a closed form, three level format for the response Of t he 795 respondents who answered yes, 789 went on to answer this question. Of those, 8.1 % were unsure, 45.6 % were somewhat sure and 46.3 % were very sure (Figure 2 4) the opportunity to provide a maximum percentage that they would be willing to pay. However, a very large number of respondents were confused by this question, thinking that they were being asked to provide their level of certainty in percentage form. Due to this misinterpretation by at least some respondents there were a high number of 100% values, which skewed the responses too much to be useful to this study. Therefore the average maximum WTP amount is not discussed or summarized. The same is true for the corresponding questions for each type of strategy.

PAGE 38

27 Question 10C : I f Not, I s There Any % That You Would Vote f or? Those respondents who answered no to the initial questions about their willingness to pay were asked if there was any amount at all that they might be willing to pay and if so how much. This question was needed in case a respondents was asked to respond to an amount that they could not afford so amounts lower than the maximum (i.e., 10%) were expected. A total of 515 of the 531 respondents who answered no to the or iginal question went on to answer this question. Of those, 83.3 % responded that there was no amount th at they would be willing to pay but 16.7 % responded that there was, indeed, an amount they would be willing to pay, and the average of those amounts was 3.89%. Question 11: Are You Aware of t he Beach Conditions Reporting System f or t he Gulf Coast o f Florida tm ? Of the 1,334 respondents who chose to answer the question, 74.1 % were un familiar with the system, while 25.9 % had at least heard of it. This is another question where the responses are most likely regional in nature, since this system only operates in southwest Florida. Question 11A : Would You Pay a One Time Donation Of $ X i nto This Trust Fund? For this scenario, respondents would be do nating to the trust fund for access to the proposed system for three years. The majority of respondents (64%) who answered the question indicated that they would not be willing to make a one time donation to a beach conditions reporting service trust fund as described above. Thirty six percent indicated that they would be willing to donate (N=1,320). level in Table 2 8.

PAGE 39

28 Question 11B : If So, How Sure Are You o f T his Decision? As with the p revious two scenarios t hose that answered yes were asked how sure they were of that decision. Of the 475 respondents who answered yes, 474 went on to answer this. Of those, 11.6 % were unsure, 49.6 % were somewhat sure and 38.8 % were very sure of their st ated willingness to pay (Figure 2 5) Question 11C : I f Not, Is There Any Amount T hat You Would Pay? Those respondents who answ ered no to the initial question were asked if there was any amount at all that they might be willing to pay, and if so how much A total 806 of the 875 respondents who answered no to the original question went on to answer this question. Of those, 88.8 % responded that there was no amount th at they would be willing to pay but 11.2 % responded that there was an amount the y would be willing to pay. T he average of those open ended responses from those respondents who were very sure of their willingness to pay was $6.91, which was less than the maximum proposed donation of $25. Question 12: Approximately How Many Days Do You Spend at t he Beach Each Year? The average days spent at the beach per year among respondents was 48.9 days per year with a standard deviation of 87.4 days. The minimum number of days spent at the beach was 0 days and the maximum number of days was 365 days. Question 13: Did You Pay Pr operty Taxes i n Florida i n 2009? Out of the 1,344 respondents who answered the question, 87.4 % indicated that they paid property taxes in Florida in 2009. Question 13A : Would You Vote For a 3 Year County Property Tax t o Fund Con trol Programs ? Approximately 49 % of respondents who answered the question indicated that they would be willing to vote for a 3 yeat property tax fo fund local control programs while 50.7 % indicated

PAGE 40

29 that they would not be willing (N=1,343). The percentag e of responses by price level are shown in Table 2 8. Question 13B : If So, How Sure Are You o f This Decision? As with the previous scenario s those that answered yes were asked how sure they were of that decision. Of the 662 respondents who answered yes, 652 went on to answer this follow up Of those, 11 .0% were unsure, 54.6 % were somewhat sure and 34.4 % were very sure of their stated WTP for a control program (Figure 2 6) Question 13C : I f Not, I s There Any Amount That You Would Pay? Those responden ts who answered no to the initial questions were asked if there was any amount at all that they might be willing to pay, and if so how much. A total of 653 of the 681 respondents who answered no to the original question went on to answer this question. O f those, 85.2 % responded that there was no amount that they would be willing to pay but 14.9 % responded that there was, indeed, an amount they would be willing to pay. T he ave rage of those amounts was $4.69 per $100,000 of assessed taxable property value, which was less than the maximum of 415 proposed in the question. Question 14: In General, Which Type of Control Most Appeals t o You? O ver half (56.2 % ) of respondents indicated that they would prefer biological controls (e.g., introducing a predator speci es or a species that completes for food with the algae) while 20.7 % preferred chemical controls (e.g., introducing a substance that would alter the composition of the algae or absorb the algae) and the remaining 23.1 % preferred neither. Question 15: W hich Program Would You Prefer If The State o f Florida Only Had Funds f or One? The final question of the survey asked respondents which of the three scenarios, if any, they would prefer if the state of Florida only had funds to implement one. Of the three proposed

PAGE 41

30 programs the prevention program had the most support with 43.1 % of respondents indicating that they favored that strategy (Figure 2 7) Twenty percent favored the control strategy and 16.6 % favored the mitigation strategy. Finally, 20.4 % indica ted that they would prefer that no strategy be implemented and funds used for some other purpose (N=1,311). Demographics Question 16: How Long Have You Resided i n Florida ? The average length of Florida residency among respondents was 24.2 years (N=1,445) but responses ranged from 0 to 82 years. The standard deviation was 16.87 years. Question 17: How Many Months o ut of t h e Year Do You Reside in Florida ? The majority of respondents reside in Florida for the entire year (93.4%). The average of the month s of residency in Flor ida each year was 11.7 months, with a standard deviation of 1.2 months (N=1,439). Figure 2 8 displays a frequency distribution for months residing in Florida. Question 18: What Is t he ZIP Code of Your Residence i n Florida ? The ZIP code information was first used to determine the distribution of responses in the study area. The overall response rate for each region is provided in the Table 2 8, as well as the proportion of the total responses represented by the region. To r eview, the northeast region was comprised of St. Johns, Flagler, Volusia and Brevard counties. The southwest region included Manatee, Sarasota, Charlotte and Lee counties. Finally, the northwest region encompassed Gulf, Franklin, Bay and Okaloosa countie s. At 10.7%, the response rate for the southwest region was slightly better than the 10.4% overall response rate, however, the response rates for the northeast and northwest regions (9.9% and 8.4%, respectively) were slightly under the overall response ra te. Of the total 1,454 responses, the southwest region made up the highest percentage of responses at 48.9%. The northeast region made up 41.7%, and the northwest region brought up

PAGE 42

31 the rear at only 9.6%. While these distributions may seem disproportiona te, they are roughly equivalent to the proportion of surveys mailed out to each area (which was based on population). Question 19: How Many Miles by Car Do You Live from t he Coast? nearest coastline was 8.7 miles with a standard deviation of 10.9 miles. The reported minimum distance of residency to the coastline was 0 miles and the maximum distance was 200 miles. Finally, 43.2% of respondents lived within five miles of the coast (N=1,428). Question 20: In What Year Were You Born? The information gathered in this question was converted from year born into the age of the respondent. T he average age was 59.9 years with a standard deviation of 14.4 years. The minimum age was 18 years and the maximum was 96 years. Finally, 50.2% of respondents were over the age of sixty (N = 1,426). Question 21: What Is the Highest Level o f Education That You Have Completed? The majority of respondents had a college degree or higher ( Figure 2 9 ). Thirty two percent degree. Only 11% held a high school degree or less. Question 22: Which of t h e Following Describe Your Race o r Ethnicity? To describe their ethnicity respondents were allow ed to choose from more than one category. The majority of respondents (92.4 % ) considered themselves to be Caucasian. Nearly two percent (1.9%) considered themselves to be African American and another 3.0% indicated that they were of Hispanic descent. Fi nally, 2.7% indicated that they were of some other race or ethnicity (N = 1,433).

PAGE 43

32 Question 23: Which Category Includes efore Taxes? T 56.3%) reported an annual income level fell between $30,000 a nd $90,000 (Figure 2 10) Approximately 17% made less than $30,000 a year, and around 7% earned over $150,000 per year. Comparison of Mail and Internet Responses To supplement the mail surveys, an additional 692,431 email invitations for an internet vers ion of the survey were sent out to residents in the same 12 counties. The two surveys were very similar, only treating the format of the WTP questions in a slightly different manner (due to complications with the online format). Due to this, the WTP from the mail and internet surveys must be estimated separately in slightly different ways. The rest of the surveys can be compared directly. Since there were only 115 responses from the internet survey were returned, they are not summarized fully on their o wn. They are, however, compared below to the responses from the mail survey, for questions of particular interest. Appendix B shows difference in WTP questions in online survey, while Appendix C provides verbatim responses to open ended final question as king for comments about survey. The open ended responses are not included due to the tremendous number of comments; however they are available from the author upon request. Table 2 10 compares the percentage of respondents who have experienced the effect s of red tide. The percentage of yes responses is provided. It appears that the respondents from the internet survey have slightly more personal experience with red tide. Of those respondents, 86% indicated they had noticed red tide conditions in the wa ter, as compared to 74% from the mail survey. Similarly, 93% indicated that they had experienced the odor of decaying fish as compared to 82% from the mail survey. In fact, there was a higher percentage of internet ndicator of level of experience.

PAGE 44

33 The responses from the true/false questions are compared in Table 2 11. Again, the respondents from the internet survey seemed to have more knowledge about the causes and effects of red tide. For each question a higher p ercentage of those respondents answered correctly. For example, 54% of internet respondents were aware that seafood bought in stores or restaurants is safe to consume, as compared to only 44% of mail respondents. Of note, too, is that 50% of internet res pondents were aware that reddish brown water does not always indicate that humans will experience respiratory problems, while only 32% of mail survey respondents were aware of this. In Table 2 12 the primary reasons why a respondent was concerned or uncon cerned about red tide is compared between the two surveys. First, of note is that 81% of internet respondents reported being concerned about red tide, compared to 76% of mail respondents. While, the reasons for being concerned about red tide do not diffe r greatly, the reasons for being not concerned were slightly dissimilar. The In the internet survey 73% of respondents indicated that they were unconcerned about red tide because it is a natural occurrence, while only 44% of mail respondents felt that way In addition, only 4% of internet respondents were unconcerned because red tide had not affected them, as compared to 29% for the mail respondents. This is consistent with the responses from question two, which indicated that internet respondents had mo re personal experience with red tide than mail respondents. Table 2 13 compares the percentages that agree with statements about scientific research on red tides in Florida. These responses were remarkably similar across both surveys. Table 2 14 is a co mparison of the familiarity of respondents with various agencies that supply red tide information. These, too, are very similar, though respondents from the internet survey were slightly more familiar with each of the agencies. Of note is Mote Marine Lab with whom 47%

PAGE 45

34 of internet respondents were at least somewhat familiar with as compared to only 37% of mail respondents. In addition, 21% of internet respondents were familiar with the Sierra Club, while only 10% of mail respondents were. Finally, 42% o f internet respondents were familiar with red tide research from the universities in Florida while only 30% of mail respondents reported a familiarity. Table 2 15 compares the percentage of respondents that obtain information about red tide from several d ifferent informational sources. Like the last two questions, the responses from the internet and mail surveys were for the most part similar to one another. A notable difference was the response rate for internet sites, where 58% percent of internet resp ondents used this informational source sometimes or often, while only 34% of mail respondents did the same. The percentage of respondents who were willing to pay for each strategy is broken down by price level in Table 2 16. For each strategy type ther e were more respondents from the mail survey who were not willing to pay than from the internet survey (40% versus 36% for prevention, 64% versus 46% for mitigation, and 51% versus 44% for control). The responses for the rest of the price levels did not d iffer a great deal. For the prevention strategy, 33% of internet respondents were willing to pay the lowest price level (1%), as compared with 22% of mail respondents. Another notable difference was that 19% of internet respondents were willing to pay th e highest price level for the mitigation strategy ($25), while only 9% of mail respondents were willing to do the same. A comparison of the responses from the strategy specific questions can be found in Table 2 17. The percentage of respondents who use f ertilizer, the number of days spent at the beach and the percentage of respondents who paid property taxes were not notably different. In addition, respondents from both surveys preferred biological controls to chemical controls or no

PAGE 46

35 controls at all. Ov erall preference for the strategies was similar: respondents from both surveys preferred prevention strategies the most, followed by control, no strategy at all and finally, mitigation. Finally, more respondents from the internet survey were aware of the BCRS (31%) than respondents from the mail survey (26%). Geographic characteristics and tenure in Florida are summarized in Table 2 18. There were more responses from the northeast region in the mail survey (41%, versus 25% from the internet), while there were more responses from the southwest region in the internet survey (68%, versus 49% from the mail survey). The responses from the northwest region were similar for both. Length of residency in Florida seemed to be distributed similarly in both surveys though the internet survey had more respondents who had resided for over 20 years than the mail survey (65% versus 50%). The majority of respondents for both the internet and mail surveys resided in Florida for more than 9 months out of the year (96% an d 95%, respectively). Finally, for both surveys the majority of respondents lived within 20 miles of the coast (96% for the internet survey and 95% for the mail survey). Table 2 19 summarizes the remaining demographic characteristics. The internet respo ndents did not differ in age (an average of 60 years for both). They also did not differ greatly in ethnicity (95% Caucasian for the internet survey versus 92% for the mail survey). Respondents from the internet survey were very slightly more educated, w ith 33% having a graduate or professional degree as compared with 25% of mail survey respondents. Finally, internet respondents were somewhat wealthier than mail respondents. Only 6% of internet respondents were in the lowest income bracket as compared t o 17% of mail respondents, and 16% were in the highest income bracket as compared to 7% of mail respondents.

PAGE 47

36 Finally, Figures 2 11 and 2 12 compare frequency at which respondents search for information and the dependence on coastal water quality, respectively. The information seeking behaviors reported by respondents from the internet survey and respondents from the mail survey are almost identical. However, a larger percentage of internet respondents reported being highly dependent on coastal wa ter quality than did the mail respondents (57% versus 40%).

PAGE 48

37 Table 2 Alternative possible effects of a red tide on humans Yes No N I have noticed red tide conditions in the water 74.3% 25.7% 1,325 I have seen dead animals on shore during a red tide 73.9% 26.1% 1,326 I have experienced the odor of decaying fish at the beach 82.0% 18.0% 1,327 I have (or a member of my family has) experienced burning eyes, scratchy throat, or coughing that could have been from a red tide 69.1% 30.9% 1,336 I have changed plans to visit a beach because of a red tide 63.7% 36.3% 1,327 I have changed hotel reservations because of a red tide event 6.2% 93.8% 1,302 I have changed a restaurant reservation because of a red tide event 19.2% 80.8% 1,317 Table 2 2. True or false statements about red tide events in Florida Statements about red tide True False Know Correct Answer N Red tide conditions can vary greatly from one area to another (within a few miles) due to winds and currents 92.0% 1.0% 7.0% True 1,347 Seafood bought in stores or restaurants is safe to eat during red tides 44.3% 16.4% 39.3% True 1,342 Recreationally caught shrimp and crab are safe to eat during a red tide 12.8% 37.1% 50.1% True 1,337 Recreationally caught finfish are unsafe to eat during a red tide 32.3% 14.8% 52.9% False 1,342 Recreationally caught oysters are unsafe to eat during a red tide 43.1% 9.1% 47.8% True 1,342 People with asthma are more likely to notice the effects of red tide 78.0% 1.3% 20.6% True 1,343 16.8% 6.1% 77.1% False 1,334 Reddish brown water indicates that humans will experience respiratory problems 17.7% 32.4% 49.9% False 1,344 The algae that causes red tides is always present in the Gulf of Mexico 44.9% 7.8% 47.3% True 1,345 Red tides are the same all over the world 9.1% 25.8% 65.1% False 1,344

PAGE 49

38 Table 2 3. Primary reason for whether respondent was concerned or not about red tides Primary reason Distribution of responses For generally being not concerned: Red tides have not affected me 28.9% Red tides are unpredictable so being concerned serves no purpose 9.8% Scientists are working on the issue 5.9% Red tides are a natural occurrence 44.6% Other 10.8% Total (N = 287) 100.0% For being somewhat or very concerned: Red tides cause economic losses 9.8% Red tides prevent fishing, beach going, and other marine activities. 24.1% Red tides affect human health 31.9% Red tides indicate poor water quality 11.5% Other 22.6% Total (N = 858) 100.0% Table 2 4. Level of agreement about scientific research on red tides in Florida Statements about scientific research N Average Rating % Rated Scientific research on red tides in Florida has generated a lot of knowledge 1,346 3.36 9.1% Scientific research on red tides in Florida has generated practical applications 1,343 3.00 2.3% Results from the scientific research on Florida red tides is confusing 1,344 3.20 6.2% Better monitoring and prediction systems are needed for red tides in Florida 1,345 3.86 26.7% Learning about how Florida red tides affect marine animals is important 1,346 4.46 56.2% Learning about how Florida red tides affect human health is important 1,346 4.54 64.6% Learning about how people are affected by Florida red tides is important 1,345 4.47 58.9% Learning about how we can control or prevent Florida red tides is important 1,344 4.41 59.6% Determining the costs and benefits of different red tide strategies is important 1,345 4.29 48.6% Notes: The rating scale is a five point Likert

PAGE 50

39 Table 2 5. Familiarity of respondents with various agencies that supply red tide information in Florida Agency or Organization N Not at all Somewhat Very FWRI (Fish & Wildlife Research Institute) 1,344 65.4% 29.2% 5.4% Universities 1,336 70.5% 25.8% 3.6% Mote Marine Lab 1,342 63.4% 23.0% 13.5% Sierra Club 1,337 89.6% 9.2% 1.2% START (Solutions to Avoid Red Tide) 1,337 91.1% 7.0% 1.9% Florida Red Tide Coalition 1,339 88.9% 9.9% 1.2% Florida Red Tide Alliance 1,338 90.8% 8.2% 1.0% Beach Conditions Reporting Systems 1,340 45.1% 39.7% 15.2% Table 2 6. Frequency of obtaining information by alternative sources Information delivery mechanism N Never Sometimes Often Television 1,358 11.3% 55.7% 33.0% Radio 1,347 43.7% 46.1% 10.2% Local newspapers 1,353 18.0% 53.3% 28.7% Internet websites 1,341 65.7% 27.4% 6.9% Public forums, meetings or workshops 1,344 90.2% 9.0% 0.8% Printed brochures or pamphlets 1,347 78.7% 20.3% 1.0% Friends or family 1,347 33.1% 52.1% 14.8%

PAGE 51

40 Table 2 7. Description of willingness to pay scenarios Description of strategy Payment vehicle Bid levels (X) Prevention: Establish a state wide retail tax on fertilizer that would encourage a reduction in fertilizer use and raise funds to pay for continual monitoring of coastal water quality, and research to determine water quality improvements. If no measurable improvements were found within three years, the l aw would be automatically repealed. Fertilizer tax 1% 5%, 10% Mitigation: Establish a Beach Conditions Reporting Service Trust Fund to support the training of observers, initial equipment expenditures and maintenance of an electronic reporting system. It is anticipated that one time donations to this fund would be sufficient to establish and support this program over the next three years. Only people who donate would able to access the system. One time trust fund donation $5, $15, $25 Control: Estab lish a 3 year tax on the assessed value of all private property to fund red tide control programs, including pilot testing. If no measurable improvements were found within three years, the law would be automatically repealed. Property tax $5, $10, $15 Table 2 8. Percentage of willingness to pay responses, by price level Price level N # Yes %Yes # No % No Prevention: High (10%) 457 251 54.9% 206 45.1% Med (5%) 415 246 59.3% 169 40.7% Low (1%) 454 298 65.6% 156 34.4% Total 1,326 795 531 Mitigation: High ($25) 452 115 25.4% 337 74.6% Med ($15) 413 141 31.4% 272 65.9% Low ($5) 455 219 48.1% 236 51.9% Total 1,320 475 845 Control: High ($15) 464 206 44.4% 258 55.6% Med ($10) 418 181 43.3% 237 56.7% Low ($5) 461 275 59.7% 186 40.3% Total 1,343 662 681

PAGE 52

41 Table 2 9. Responses by region Region N Distribution of responses Response rate Northeast region 607 41.7% 9.9% Southwest region 707 48.9% 10.7% Northwest region 140 9.6% 8.4% Total 1,454 100.0% Table 2 10. Percentage of respondents that have experienced effects of a red tide event by survey Potential experience with red tides Internet ( N=111 115 ) Mail (N=1,302 1,336) I have noticed red tide conditions in the water 86% 74% I have seen dead animals on the shore during a red tide 88% 74% I have experienced the odor of decaying fish on the beach 93% 82% I have (or a member of my family has) experienced burning eyes, scratchy throat, or coughing that could have been from a red tide 85% 69% I have changed plans to visit a beach because of a red tide event 74% 64% I have changed a hotel reservation because of a red tide event 7% 6% I have changed a restaurant reservation because of a red tide event 37% 19%

PAGE 53

42 Table 2 11. Comparison of responses regarding red tide knowledge Internet (N = 112 115) Mail (N = 1,337 1,347) Correct Answer Correct Know Correct Know Red tide conditions can vary greatly from one area to another (within a few miles) due to winds and currents True 97% 3% 92% 7% Seafood bought in stores or restaurants is safe to eat during red tides True 54% 30% 44% 39% Recreationally caught shrimp and crab are safe to eat during a red tide True 18% 46% 13% 50% Recreationally caught finfish are unsafe to eat during a red tide False 26% 41% 15% 53% Recreationally caught oysters are unsafe to eat during a red tide True 45% 25% 43% 48% People with asthma are more likely to notice the effects of red tide True 86% 14% 78% 21% tide False 7% 77% 6% 77% Reddish brown water indicates that humans will experience respiratory problems False 50% 38% 32% 50% The algae that causes red tides is always present in the Gulf of Mexico True 61% 36% 45% 46% Red tides are the same all over the world False 33% 55% 26% 65%

PAGE 54

43 Table 2 12. Primary reason for whether respondent was concerned or not about red tides by survey Primary reason Internet (N=114) Mail (N=1,145) For generally being not concerned: Red tides have not affected me 4% 29% Red tides are unpredictable so being concerned serves no purpose 9% 10% Scientists are working on the issue 9% 6% Red tides are a natural occurrence 73% 44% Other 5% 11% Total 100% 100% For being somewhat or very concerned: Red tides cause economic losses 9% 10% Red tides prevent fishing, beach going, and other marine activities. 25% 24% Red tides affect human health 40% 32% Red tides indicate poor water quality 18% 11% Other 8% 23% Total 100% 100% events in Florida compared with 76% from the mail survey.

PAGE 55

44 Table 2 red tides in Florida between mail and internet respondents Statements about scientific research Internet (N=109 to 110) Mail (N=1,343 to 1,436) Scientific research on red tides in Florida has generated a lot of knowledge 47% 46% Scientific research on red tides in Florida has generated practical applications 21% 20% Results from the scientific research on Florida red tides is confusing 46% 33% Better monitoring and prediction systems are needed for red tides in Florida 75% 69% Learning about how Florida red tides affect marine animals is important 94% 94% Learning about how Florida red tides affect human health is important 98% 94% Learning about how people are affected by Florida red tides is important 96% 93% Learning about how we can control or prevent Florida red tides is important 91% 88% Determining the costs and benefits of different red tide strategies is important 89% 88% Notes: The mail survey responses were obtained on a 5 point Likert Table 2 14. Comparison of the familiarity of respondents with various agencies that supply red tide information in Florida by s urvey Agency or Organization Internet (N = 107 110) Mail (N = 1,342 1,344) FWRI (Fish & Wildlife Research Institute) 37% 35% Universities 42% 30% Mote Marine Lab 47% 37% Sierra Club 21% 10% START (Solutions to Avoid Red Tide) 15% 9% Florida Red Tide Coalition 16% 11% Florida Red Tide Alliance 12% 9% Beach Conditions Reporting Systems 56% 55%

PAGE 56

45 Table 2 15. Comparing the percentage of respondents that respondents obtain information by alternative source and survey Information delivery mechanism Internet (N = 106 109) Mail (N = 1,341 1,358) Television 92% 89% Radio 64% 56% Local newspapers 81% 82% Internet websites 58% 34% Public forums, meetings or workshops 18% 10% Printed brochures or pamphlets 26% 21% Friends or family 75% 67% Note: Respondents were considered to obtain information if they sought it out either Table 2 16. Comparison of willingness to pay scenarios by survey Strategy Payment vehicle Internet (maximum WTP?) Mail (WTP X? yes or no) Prevention Fertilizer tax 0%, 1%, 5%, 10%,or more than 10% 1%, 5%, or 10% Mitigation One time trust fund donation (3 years access) $0, $5, $15, $25, or more than $25 $5, $15, or $25 Control 3 year property tax $0, $5, $10, $15, or more than $15 $5, $10, or $15 Notes: Zero values for the internet survey are equivalent to not being WTP in the first stage of questioning in the mail survey. The three levels identified are equivalent between surveys. For comparison the highest two categories in the internet survey can be consolidated.

PAGE 57

46 Table 2 17. Comparison of WTP responses by bid level and survey Strategy and bid level Internet Mail Prevention: (N=109) (N=1,326) Not willing to pay (WTP = 0%) 36% 40% Willing to pay level 1 (WTP = 1%) 33% 22% Willing to pay level 2 (WTP = 5%) 18% 19% Willing to pay at least level 3 (WTP 10%) 13% 19% Mitigation: (N=103) (N=1,320) Not willing to pay (WTP = $0) 46% 64% Willing to pay level 1 (WTP = $5) 18% 16% Willing to pay level 2 (WTP = $15) 17% 11% Willing to pay at least level 3 (WTP $25) 19% 9% Control: (N=108) (N=1,343) Not willing to pay (WTP = $0) 44% 51% Willing to pay level 1 (WTP = $5) 19% 20% Willing to pay level 2 (WTP = $10) 19% 13% Willing to pay at least level 3 (WTP $15) 16% 16% Table 2 18. Comparison of responses related to WTP questions by survey Strategy and related question Internet Mail Prevention: (N=110) (N=1,321) Respondent uses fertilizers (%) 47% 55% Mitigation: (N=108) (N=1,334) Respondent is aware of the BCRS (%) 31% 26% (N=109) (N=1,087) No beach use (0 days) 4% 11% 1 7 days per year 29% 24% 8 14 days per year 22% 13% 15 21 days per year 10% 13% 22 days per year 35% 39% Control: (N=110) (N=1,334) Respondent paid property taxes in Florida last year (%) 92% 87% Respondents preference for type of control: (N=106) (N=1,297) Biological 60% 56% Chemical 18% 21% Neither 22% 23% Overall preference for red tide strategy: (N=109) (N=1,311) Prevention 42% 43% Mitigation 17% 17% Control 19% 20% None of them 21% 20%

PAGE 58

47 Table 2 19. Comparison of respondent geographic characteristics and tenure in Florida by survey Characteristic Internet Mail Region: (N=110) (N=1,454) East/North central region 25% 41% Southwest region 68% 49% Northwest region 6% 10% Length of residency in Florida (years): (N=109) (N=1,445) Less than one year 0% 1% 1 5 years 2% 10% 6 10 years 10% 17% 11 20 years 23% 22% 21 30 years 30% 19% More than 30 years 35% 31% Length of residency in Florida (months/year): (N=109) (N=1,439) Less than one month 0% <1% 1 3 months 1% <1% 4 6 months 1% 2% 7 9 months 2% 2% More than 9 months 96% 95% Distance to coast (miles by car): 0 miles (live on the water) 0% <1% 1 5 miles 1% <1% 6 10 miles 1% 2% 11 20 miles 2% 2% More than 20 miles 96% 95%

PAGE 59

48 Table 2 20. Comparison of age, education, income and race by survey Characteristic Internet Mail Average age of respondent (years; N=96, 1,426) 60 (10 s.d.) 60 (14 s.d.) Highest level of education completed: (N=110) (N=1,431) Some elementary or high school 1% 2% High school graduate or GED 4% 9% Technical/vocational 2% 6% Some college 29% 26% College graduate 32% 32% Graduate/professional degree 33% 25% Race or ethnicity (check all that apply): (N=110) (N=1,433) White/Caucasian (% yes) 95% 92% African American/Black (% yes) 0% 2% Asian (% yes) 0% 0% Native Hawaiian/Pacific Islander (% yes) 0% 0% American Indian/Alaskan Native (% yes) < 1% 2% Hispanic/Latino (% yes) 2% 3% Other (% yes) 2% 3% Annual household income before taxes ($/year): (N=98) (N=1,322) Less than $30,000 6% 17% $30,000 $60,000 23% 32% $60,001 $90,000 26% 25% $90,001 $120,000 23% 14% $120,001 $150,000 5% 6% More than $150,000 16% 7%

PAGE 60

49 Figure 2 1. Concern over red tide events in Florida (N = 1,302) Figure 2 2. How frequently respondents seek information about red tides (N = 1,330) Figure 2 3. Dependence on coastal water quality (N = 1,338)

PAGE 61

50 Figure 2 4. Share of respondents willing to pay a fertilizer tax (N = 1,326) and how sure they are of that decision (N = 789). Figure 2 5. Share of respondents willing to donate to a trust fund (N = 1,32 0 ) and how sure they are of that decision (N = 474). Figure 2 6. Share of respondents willing to vote for a property tax (N = 1,3 43 ) and how sure they are of that decision (N = 653). Yes 60% No 40% Not sure 8% Some what sure 46% Very sure 46% Yes 36% No 64% Not sure 12% Some what sure 46% Very sure 44% Yes 49% No 51% Not sure 11% Some what sure 46% Very sure 34%

PAGE 62

51 Figure 2 7 Respondents preferred strategy (N = 1,311) Figure 2 8. Frequency of Months Residing in Florida (N=1,439) 43% 17% 20% 20% Prevention: 3 year fertilizer tax for wataer quality monitoring and enforcement Mitigation: Donation to fund beach conditions reporting system for 3 years Control: 3 year tax on assessed home values to fund red tide control program None of them 5 7 44 39 1344 0 200 400 600 800 1000 1200 1400 1600 0 to 2 3 to 5 6 to 8 9 to 11 12

PAGE 63

52 Figure 2 9. Distribution of highest level of formal education among respondents (N = 1,431) Figure 2 10. Distribution of household income of respondents (N = 1,322) 2% 9% 6% 26% 32% 25% Some elementary or high school High school or GED Technical/vocational Some college College degree Graduate/professional degree 17.2% 31.6% 24.7% 13.6% 5.9% 7.0% Less than $30,000 $30,001 to $60,000 $60,001 to $90,000 $90,001 to $120,000 $120,001 to $150,000 Over $150,000

PAGE 64

53 Figure 2 11. Comparison of frequency that respondents seek information about red tides by survey format Figure 2 12. Comparison of dependence on coastal water quality by survey 18. 0% 44. 0% 25. 0% 13. 0% 23.1% 43.1% 26.2% 7.6% Never When red tide is in the area When something new is reported On a regular basis 6% 37% 57% 7% 53% 40% "Not at all dependent since I don't water any plants, fish, or go to the beach" "Somewhat dependent since I have a lawn and occasionally visit the beach or fish" "Very dependent since water is important to my recreation or livelihood" Internet survey (N = 11 1 ) Mail survey (N = 1 ,338 ) Internet survey (N = 110) Mail survey (N = 1 ,33 0)

PAGE 65

54 CHAPTER 3 RESULTS OF EMPIRICAL MODELS Overview All models were estimated in LIMDEP 8.0 using maximum likelihood estimation (MLE). As discussed in Chapter 1, models were run for each management strategy (i.e. prevention, control and mit igation) separately. There were a total of four models run for each strategy. ordered model that captured four levels: zero for those not willing to pay the price they were asked to evaluate This research analysis plan led to twelve models being estimated in total. A description of the dependent and independent variables used in these models (including definitions and statistics) can be found in Tables 3 1 and 3 2, respectively. The next section will focus primarily on a comparison of the three binary models for each management strategy based on model statistics and the signs and statistical significance of the explanatory variables. The alt ernative estimates of willingness to pay (WTP) will also be discussed for each model. The third section will cover the results of the ordered models, including the marginal values and comparable WTP estimates. The fourth section will include a brief disc ussion and summary of the non parametric estimates of WTP (i.e. the Turnbull approach). The final section of chapter 3 includes a summary and discussion of the empirical results.

PAGE 66

55 Comparison of Binary Models For each management strategy, three binary logit models were run in which the definition response. All three estimated models are shown in a single table for each strategy (Tables 3 3, 3 4 and 3 5) and the corresponding estimated marginal values are similarly shown in a tables for each strategy (Tables 3 6, 3 7, and 3 8). To review, the first model used the data in its raw espondent to them. Throughout this discussion this model will be referred to as Model 1. In the second models for each management strategy are discussed separately below. Binary Logit Models for th e Prevention Strategy The empirical results for the prevention binary logit models are shown in Table 3 3, including the coefficients and their corresponding p values, as well as model performance results and WTP estimates. In addition, the marginal effec ts are summarized in Table 3 6, though they are not discussed here. There were 1,016 useable observations for estimating the prevention models once all missing observations were thrown out. There are several different indicators of whether the estimated models are good in the sense that they adequately explain the responses; even though the goal of this study was not to create a model for prediction (i.e., maximize the explanatory power of each model), it is

PAGE 67

56 important to have a set of explanatory variable s that are justified by economic theory and that do have explanatory power. First, a likelihood ratio t est is used to test the significance of each estimated model. This test compares the log s. The restricted model assumes all model coefficients, except the intercept, equal zero. The unrestricted model is the model as specified with the explanatory variables that are hypothesized to have a correlation with the dependent variable (i.e., wheth er respondents indicated they were willing to pay to support the proposed program). LIMDEP automatically computes both log likelihood values and the resulting likelihood ratio statistic, which is the absolute value of two times the difference in the log l ikelihood values. If this statistic is greater than the 5% critical value of the Chi square distribution with 18 degrees of freedom (i.e., the number of omitted variables in the constrained model ), which is 28.87, then the null hypothesis of the restricted model is rejected in favor of the unrestricted model. As a result, we use the unrestricted model results reported in Table 3 3 for each model. A second method of examining model fit in general is the proportion of the variance in the dependent v ariable that is explained by the variance in the independent variables using the R 2 statistic, but is no equivalent measure in logistic regressio n. However, there are several "p seudo" R 2 statistics that can be used. LIMDEP computes the McFadden R 2 statist ic, which is calculated using the restricted and unrestricted log likelihoods of the data. Because pseudo R 2 statistics are generally much lower than the traditional R 2 the McFadden R 2 for models 1, 2 and 3 (0.217, 0.196, and 0.142, respectively) are exp ected and are considered to explain an adequate amount of the variation in the dependent variables.

PAGE 68

57 A third method for discussing model fit is the percent of responses that were correctly predicted by the estimated models. The percents correctly predict ed among the three models varied only slightly from model to model (73.9% for Model 1, 72.2% for Model 2 and 72.5% for wer decreased slightly as the models became less inclusive (i.e. from Model 1 to Model 3), all three models were of adequate fit and superior to constant only specifications. In addition, the signs of the coefficients of the significant variables in all t hree models were consistent across all three models indicating a robust specification. Seven of the total eighteen explanatory variables were significant determinants of WTP for Model 1 at the 10% level or lower (i.e., p < 0.10). Since the magnitude of th e coefficients is not directly interpretable in logistic regression, the sign of the coefficients will be discussed. As one would expect, the variable for price level (P_Price) was significant and negative, indicating that an increase in price would decre ase the probability of a WTP greater than zero. The only other significant variable that was negative was length of Florida residency (Q16). Level of concern for red tide (Q4), information seeking behavior (Q6A) and preference for the prevention strategy (Pref_P) all increased the probability of a WTP greater than zero. The dummy variables for region (DV_EST and DV_SW) were also significant and positive, indicating that being in either the southwest coast region or the northeast coast region increased th e probability of WTP other than zero more so than being in the base region (the northwest region). For Model 2 seven out of the eighteen variables were statistically significant at the 10% level or lower. As with Model 1, the variable for price level was a significant determinant of WTP. Price level was negative for Model 2, also, indicating that an increase in this variable would decrease the probability of a WTP greater than zero. However, length of Florida

PAGE 69

58 residency was not a significant determinant o f WTP for Model 2. Again, level of concern for red tide, information seeking behavior, preference for the prevention strategy and the dummy variables for region were all statistically significant and positive. In addition, dependence on coastal water qua lity and quantity was significant and positive for Model 2, indicating that an increase in this variable will increase the probability of a WTP greater than zero. Model 3 had the most significant variables of the three models, nine in total. The significa nt variables also differed for this model as compared with the first two. As with Model 1 and Model 2, price was significant and negative, as well as the length of Florida residency. The level of concern for red tide, information seeking behavior and pre ference for the prevention were, again, all significant and positive; however, the regional dummy variables were not significant for Model 3, which means that region is not a factor in explaining those that were very sure of their willingness to pay for th e proposed prevention strategy. Unlike the previous two models, age was a significant determinant for Model 3, having a negative impact on WTP. Older respondents were less likely to be very sure of their willingness to pay for the proposed prevention str ategy. Also unlike the previous models, education level (Q21A) and ethnicity (Q22) were significant for Model 3. These variables were positive, indicating that a higher education level would increase the probability of being very sure of their WTP a 1% t o 10% fertilizer tax, as would being of Caucasian ethnicity. The average willingness to pay associated with each model was estimated using two The grand constant was calculated using all es timate parameters, regardless of their statistical significance (Dumas et al. 2007). WTP estimates for the prevention models are included in Table 3 4. For Model 1, the estimated WTP (i.e., average fertilizer tax that would be supported by residents of c oastal

PAGE 70

59 counties) calculated using the GC method was 10.3% while the WTP calculated using the Delta method was higher, at 16.3%. For Model 2 the estimated WTPs were 5.2% and 11.2% for the GC method and Delta method, respectively. Finally, for Model 3 the GC method estimated WTP was 14.0% while the Delta method estimated WTP was 19.4%. Of the three WTPs calculated using the Delta method, however, none were statistically significant. Note that negative WTP estimates are not uncommon when logit models are estimated using linear variables so redefining the variables and using a probit distribution would likely solve this problem; however, given that we have several models and several methods of calculating WTP, and the negative estimates are only associated with the model attempting to explain those that were very sure of their response, these approaches are not investigated here. Binary Logit Models for the Control Strategy The empirical results for the binary logit models of the WTP for the proposed contro l strategy are shown in Table 3 4, and the corresponding marginal effects are summarized in Table 3 7. There were 975 observations used in the estimation of the control models. As with the prevention models, the likelihood ratio test for all three contro l models indicate that the unrestricted models (with all the explanatory variables) perform better than the restricted models. The McFadden R 2 for control models 1, 2 and 3 were 0.21, 0.22, and 0.14, respectively, indicating that the model explaining only the very sure responses does not explain as much of the variation in the dependent variables as the first two models. There was slightly more variation in the percents correctly predicted among the three models than for the prevention models (i.e., 70.6% for Model 1, 71.3% for Model 2 and 80.5% for Model 3), with the very sure model being able to best predict the responses. Taking into consideration all of the different measures for r to constant only specifications.

PAGE 71

60 The control models had significantly more statistically significant variables than the prevention models. The signs of the coefficients of the all the significant variables in all three models were consistent. The expla natory variables that were significant determinants of WTP across all three models were the price level (C_Price), the strategy specific dummy variable for northe ast region (DV_EST), whether or not control was the preferred strategy (Pref_C), the length of Florida residency (Q16), the level of concern for red tide issues (Q4), information seeking behavior (Q6A) and dependence on coastal water quality and quantity ( Q9A). The variables that increased the probability of a WTP greater than zero across all models were the dummy variable for the northeast region, whether or not control was the preferred strategy, the level of concern for red tide issues, the frequency at which respondents look for red tide information and dependence on coastal water quality and quantity. Those that decreased the probability of a WTP greater than zero across all models were the price level, the dummy variable for whether or not the prefer residency. These variables were statistically significant determinants of WTP with the same signs no matter how inclusive the definition of a yes was. There were several more explanatory variabl es that were the significant for both Models 1 and 2, but not for Model 3. These included the variable indicating whether control came first in the order of appearance of the strategy types (CFST), which is an indication of order bias, the dummy variable for the southwest coast region (DV_SW), education level (Q21A), income level The variables that increased the probability of a WTP greater than zero were th e control first variable, the dummy variable for the southwest region and education level. Those that had the

PAGE 72

61 decreased the probability of a WTP a 3 year prop erty tax surcharge of $5 to $15 per $100,000 of assessed value. There was only one variable that was significant in Model 2 only. That variable was ethnicity (Q22), and it had the effect of increasing the probability of a WTP greater than zero. The only significant variable that was specific to Model 3 was the number of months a year residing in Florida (Q17). The estimated coefficient for this variable was negative, indicating that an increase in months decreased the probability of a WTP greater than ze ro. In total, there were fifteen out of twenty variables that were statistically significant in at least one of the models. WTP estimates for the control models are reported in Table 3 5. For Model 1, the estimated WTP calculated using the GC method was $0.70 while the WTP calculated using the Delta method was much higher, at $10.17. For Model 2 the estimated WTPs were $3.43 and $7.01 for the GC method and Delta method, respectively. Finally, for Model 3 the GC method estimated WTP was $28.79 while th e Delta method estimated WTP was $16.99. Again, none of the WTPs calculated using the Delta method were found to be statistically significant. Binary Logit Models for the Mitigation Strategy The empirical results for the binary logit models for the propo sed mitigation strategy are shown in Table 3 5, while the corresponding marginal effects are summarized in Table 3 8. The mitigation models were estimated using 836 observations. As with the prevention and control models, the likelihood ratio test statis tic (i.e., the chi squared value) for all three mitigation models indicate that they are preferable to the constant only (restricted) model. The McFadden R 2 for mitigation models 1, 2 and 3 were nearly identical at 0.12, 0.12, and 0.13, respectively. They too, are considered to explain an adequate amount of the variation in the dependent

PAGE 73

62 variables for these types of models. There was more variation in the percents correctly predicted among the three models than in the control models (69.9% for Model 1, 7 2.0% for Model 2 and 85.0% for Model 3), although the pattern was the same with Model 3 being a better predictor of models were of adequate fit and superior to constant only specifications, though slightly worse than the previous two models given the lower McFadden R 2 and higher variation within the percent correctly predicted. Like the prevention models, the mitigation models had, in total, far fewer statist ically significant variables than the control models. Again, the signs of the significant variables were consistent across models. The explanatory variables that were significant determinants of WTP across all three models were the price level (M_Price), the strategy specific variable for number of days spent at the beach (M5), whether or not mitigation was the preferred strategy (Pref_M) and the level of concern for red tide issues (Q4). The variables that increased the probability of a WTP greater than zero across all models were number of beach days, whether or not mitigation was the preferred strategy and the level of concern for red tide issues. The only significant variable that decreased the probability of a WTP greater than zero was the price lev el. The strategy specific variable indicating awareness of the Beach Conditions Reporting System for the Gulf Coast of Florida TM (BCRS) was a significant determinant of WTP for both Model 1 and Model 2. The estimated coefficient for this variable was po sitive for both models, indicating that awareness of the BCRS increased the probability of a WTP greater than zero. Also significant and positive for both Models 1 and 2 was the information seeking behavior variable (Q6A), indicating that as the frequency at which information on red tide is searched for increases, so does the probability of a WTP greater than zero. For Models 1 and 3, but not

PAGE 74

63 Model 2, dependence on coastal water quality and quantity (Q9A) was statistically significant. This variable was positive for both models indicating that an increase in dependence on coastal water quality increases the probability of a WTP greater than zero. While the dummy variable for the northeast region was significant across all models, the dummy variable for t he southwest region (DV_SW) was only significant for Models 2 and 3. Like the northeast variable, the estimated coefficients on this variable for both models were positive indicating that being in the southwest region increased the probability of a WTP gr eater than zero more so than being in the base region (northwest). Finally, education level was a significant determinant in Model 2 only and had a positive effect on WTP. In total, nine out of twenty explanatory variables were significant determinants f or at least one mitigation model. WTP estimates for the mitigation models are also reported in Table 3 5. For Model 1, the estimated WTP calculated using the GC method was $11.70 while the WTP calculated using the Delta method was higher and positive, at $2.98. For Model 2 the WTPs were $28.02 and $0.12 for the GC method and Delta method, respectively. Finally, for Model 3 the GC method estimated WTP was $32.09 while the Delta method estimated WTP was $16.37. For comparison, the values that respond ents were asked to evaluate ranged from $5 to $25. As with the prevention and control models, none of the Delta method WTPs were statistically significant. Summary of Binary Logit Models While all models were determined to be of adequate fit and predictiv e power, the models for prevention and control were slightly better than those for mitigation. For prevention models, the predictive power of the models decreased as the level of inclusivity decreased (i.e. as the e stringent). For the control models there is no clear relationship between inclusivity and predictive power, as the both the R 2 and the percent correctly predicted increased from Model 1 to Model 2 but then decreased from Model 2 to

PAGE 75

64 Model 3. The same ca n be said of the mitigation models, as the R 2 increased from Model 1 to Model 2 and decreased from Model 2 to Model 3 and the percent correctly predicted increased across models. For all models (across all management strategies) price level (P_Price, C_Pri ce and M_Price), strategy preference (Pref_P, Pref_C, and Pref_M) and level of concern regarding red tide issues (Q4) were significant and of the same sign (negative for price, positive for strategy preference and concern). These results are all intuitive In addition, the frequency at which information on red tide is searched for (Q6A) was significant and positive across all models with the exception of mitigation Model 3. The signs of all the models (within the management strategies) were consistent an d, though marginals were not discussed, the signs of the estimated coefficients and the marginals were consistent with each other as well. Though the WTPs for the prevention models decreased as the definition of a yes variable became stricter (which woul d be expected), they were not adequately similar between the different calculation methods (i.e., the grand constant method and the Delta method). In addition, the WTPs calculated using the Delta method were not statistically significant so there is littl e confidence that the WTP estimates are indeed accurate. The same issue arises with the control models and the mitigation models. All in all, it is determined that while the binary logit models are accurate enough in determining how the independent varia bles affect the WTP they do not produce adequate estimates of WTP. Ordered Models In addition to the three binary models run for each strategy type an ordered model was also run for each strategy. For these models the different ordered levels zero through three were

PAGE 76

65 In a s ense, the ordered models capture the certainty information that is implied by estimating the separate binary logit models. Ordered Logit Model for Prevention Strategy Parameter estimates for the ordered model of responses to the proposed prevention strategy (i.e., a fertilizer tax) are found in table 3 9. There were 1,004 observations used in the estimation of the ordered model. The likelihood ratio test as indicated by the chi squared statistic for the model indicated that the model, as specified, is superior to the restricted (constant only) model. In addition, the McFadden R 2 was 0.12, indicating that an adequate amount of the variation in the dependent variables can be explained by the independent variables in the model. All threshold paramet ers (1 and 2) were statistically significant indicating that the on ordered model, ten out of eighteen explanatory variables were significant determinants of WTP. Of the strategy specific variables, the price level (P_Price) and the strategy preference (Pref_P) were the only variables that had a significant effect on W TP. Other significant variables included the regional dummy variables (DV_EST and DV_SW), age, length of residency in Florida (Q16), income level (Q23A), level of concern (Q4), information seeking behavior (Q6A), and dependence on coastal water quality (Q 9A). Estimated WTP was calculated using the Delta method and is also summarized in Table 3 9. The estimated WTP for the prevention ordered model was 16.9% (per bag of fertilizer purchased) and was statistically significant at the 1% level. As the sign an d magnitude of the estimated coefficients for an ordered response model are not directly interpretable it is more productive to discuss the direction and extent of the impact of each variable using the marginal effects. The marginal effects from the preve ntion ordered

PAGE 77

66 model are summarized in Table 3 10. These effects were computed at the means of the covariates. Since the ordered response model best explains the impacts of the explanatory variables at the extremes of the ordered levels, only the marginal effects of y=0 and y=3 will be discussed here (Genio et al. 2007) For the prevention strategy, a higher price level reduced the probability of being very willing to pay by 1 percentage point. Conversely, a higher price level increased the probability of being not willing to pay at all also by 1 percentage point. Residing in the northeast region and residing in the southwest region both increased the probability of being very willing to pay by 14 and 11 percentage points, respectively. Those variables d ecreased the probability of not being willing to pay at all by 16 and 13 percentage points, respectively. The probability of being very willing to pay was increased by 32 percentage points when prevention was the preferred strategy, while the probability of not being willing to pay at all was decreased by 36 percentage points when the preferred strategy was prevention. Individuals who were concerned with red tide issues in Florida were about 11 percentage points more likely to be very willing to pay and 1 6 percentage points less likely to be completely unwilling to pay. As the frequency with which individuals seek out information on red tide increases one category, the probability of being very willing to pay increases by 7 percentage points. Conversely, as the frequency with which individuals seek out information on red tide increases one category the probability of being not dependence on water quality increased by one category, they were 7 percentage points more one category they were also 9 percentage points less likely to be completely unwilling to pay. As for demog raphic characteristics, as average age and the average length of residency in Florida

PAGE 78

67 increased each by one year the probability of being very willing to pay went down by 0.1 and 0.2 percentage points, respectively. As average age and the average length o f residency in Florida increased each by one year the probability of being not willing to pay at all went up by 0.2 percentage points for both. Individuals of Caucasian ethnicity increased their probability of being very willing to pay by 7 percentage poi nts. However, ethnicity did not significantly affect the probability of being not willing to pay at all. Finally, as income level increased by one category the probability of being very willing to pay decreased by 0.1 percentage points while the probabil ity of being not willing to pay at all increased by 0.01 percentage points. Ordered L ogit Model for Control Strategy Parameter estimates for the ordered model for the proposed control strategy (i.e., biological or chemical) are also found in Table 3 9. Th ere were 975 observations used in the estimation of the ordered model. The likelihood ratio test for the model indicated that it is better than the constant only model. In addition, the McFadden R 2 was 0.13, indicating that an adequate amount of the vari ation in the dependent variables can be explained by the independent variables in the model. All threshold parameters (1 and 2) were statistically significant indicating that the were significant for the control ordered model, fourteen out of twenty explanatory variables. Of the strategy specific variables, the pric e level (C_Price), the dummy variable for whether the of strategies (CFST) and the strategy preference (Pref_C) were the variables that had a significant effe ct on WTP. Other significant variables included the region dummy variables (DV_EST and DV_SW), length of residency in Florida (Q16), months of the year residing in

PAGE 79

68 Florida (Q17), ethnicity, income level (Q23A), number of true/false question answered corre ctly in the survey (Q3), level of concern (Q4), information seeking behavior (Q6A), and dependence on coastal water quality (Q9A). Estimated WTP was calculated using the Delta method and is also summarized in Table 3 9. The estimated WTP for the control ordered model was $10.11 and was statistically significant at the 1% level Again, signs and magnitudes of the impacts on WTP of the significant individual explanatory variables will be discussed using the marginal effects of those variables. The marginal effects from the ordered logit model for the control strategy are summarized in Table 3 11. As with prevention, only the marginal effects on being not willing to pay at all (y=0) and on being very willing to pay (y=3) will be discussed. For the ordered mo del for the control strategy, increasing price to the next highest level decreased the probability of being very willing to pay by 0.8 percentage points, while increasing the price to the next highest level increased the probability of being not willing to pay at all by 2 percentage points. control decreased their probability of being very willing to pay by 12 percentage points. methods of control strategies had the highest positive effect on being not willing to pay at all, increasing the probability by 33 percentage points, as would be expected. When control was the preferred strategy the probability of being very willing to p ay was increased by 19 percentage points. Conversely, when control was the preferred strategy, the probability of being not willing to pay at all decreased by 32 percentage points. The last strategy specific variable that had a significant effect was whe ther control came first in the order of presentation of scenarios. When control was presented as the first strategy for the respondent

PAGE 80

69 to evaluate, the probability of being very willing to pay increased by 3 percentage points, while the probability of bei ng not willing to pay at all increased by 8 percentage points (which indicates and controls for the degree of order bias). Residing in the northeast region and residing in the southwest region both increased the probability of being very willing to pay, by 8 and 5 percentage points, respectively. Residing in those areas decreased the probability of being not willing to pay at all by 18 and 12 percentage points, respectively. The number of true/false questions answered correctly in the survey also had a significant effect. Increasing the number of questions answered correctly by one question actually decreased the probability of being very willing to pay by 0.6 percentage points, while increasing the number of questions answered correctly by one question increased the probability of being not willing to pay at all by 2 percentage points. Individuals who indicated that they were concerned with issues about red tide were 7 percentage points more likely to be very willing to pay. Being concerned with red t ide issues also decreased the probability of being not willing to pay by 18 percentage points. As the frequency with which individuals seek out information on red tide increased by one category, the probability of being very willing to pay increased by 3 percentage points. Conversely, as the frequency with which individuals seek out information on red tide increases one category the probability of being not willing to pay at all decreases by 8 percentage points. ality increased by one category, they were 6 quality increased by one category they were also 13 percentage points less likely to be completely unwilling to pa y.

PAGE 81

7 0 Of the demographic characteristics, as the average length of residency in Florida increased by one year the probability of being very willing to pay decreased by 0.1 percentage points while the probability of being not willing to pay at all increased by 0.3 percentage points. As education level increased by one category, the probability of being very willing to pay increased by 0.7 percentage points and the probability of being completely unwilling to pay was decreased by 2 percentage points. Finally individuals of Caucasian ethnicity were 5 percentage points more likely to be very willing to pay. Those who were of Caucasian ethnicity were also 15 percentage points less likely to be not willing to pay at all. Ordered Logit Model for Mitigation Strat egy Parameter estimates for the ordered model for the mitigation strategy (i.e., Beach Conditions Reporting Service) are also found in Table 3 9. There were 836 observations used in the estimation of this ordered model. The likelihood ratio test for the model indicated that it is superior to the restricted model. In addition, the McFadden R 2 was 0.09, and though it is lower than would be hoped for, it still indicates that at least some of the variation in the dependent variables is explained by the indep endent variables in the model. All threshold parameters (1 and 2) were statistically significant indicating that the twenty explanatory variables were significant determinants of WTP for the mitigation strategy. Of the strategy specific variables, the price level (M_Price), awareness of the Beach Conditions Report ing System for the Gulf Coast of Florida TM (BCRS) (M1), number of days spent on the beach (M5) and the strategy preference (Pref_M) had a significant effect on WTP. Other significant variables included the southwest region dummy variable (DV_SW), educatio n level (Q21A), level of concern (Q4) and information seeking behavior (Q6A). Estimated WTP was

PAGE 82

71 calculated using the Delta method and is also summarized in Table 3 9. The estimated WTP for the mitigation ordered model was $3.16 and was statistically sign ificant at the 1% level. For comparison, respondents evaluated prices that ranged from $5 to $25 for a 3 year subscription to this proposed beach condition information system. Again, signs and magnitudes of the impacts on WTP of the significant individual explanatory variables will be discussed using the marginal effects of those variables. The marginal effects from the ordered logit model for the proposed mitigation strategy are summarized in Table 3 12. As with prevention and control models, only the m arginal effects on being not willing to pay at all (y=0) and on being very willing to pay (y=3) will be discussed. As price level increased to the next highest level the probability of being very willing to pay was raised by 0.6 percentage points. However as price level increased the probability of being not willing to pay at all was decreased by 1 percentage point. Individuals who were aware of the BCRS in the Gulf Coast area of Florida were 4 percentage points more likely to be very willing to pay. Th ose same individuals were also 8 percentage points less likely to be not willing to pay at all. As the average number of beach days among individuals increased by one day, the probability of being very willing to pay was increased by 0.1 percentage points As the average number of beach days increased by one, the probability of being not willing to pay at all decreased by 0.1 percentage points. The last strategy specific variable that was significant was the strategy preference. When mitigation was the preferred strategy the probability of being very willing to pay went up by 8 percentage points. Conversely, when mitigation was the preferred strategy the probability of being not willing to pay at all went down by 16 percentage points. Only one region du mmy variable had a significant effect, and that was the dummy variable for the southwest region. Residing the southwest region increased the probability of being very

PAGE 83

72 willing to pay by 5 percentage points, while it decreased the probability of not being w illing to pay at all by 11 percent; this larger impact is expected since this service is already available free to the public in this region. Individuals that were concerned with issues regarding red tide were 9 percentage points more likely to be very wi lling to pay and were 22 percentage points less likely to be not willing to pay at all. As the frequency with which individuals seek out information on red tide increases one category, the probability of being very willing to pay increases by 4 percentage points. Conversely, as the frequency with which individuals seek out information on red tide increases one category the probability of being not willing to pay at all decreases by 8 percentage points. Finally, the only demographic characteristic that w as significant was education level. As education level increased by one category the probability of being very willing to pay increased by 8 percentage points and the probability of being not willing to pay at all decreased by 2 percentage points. Summary of the Ordered Logit Models All three ordered models were of significant fit and predictive power. The likelihood ratio test results (i.e., chi squared values) indicate that the models perform better than constant only specifications. In addition, the p arameter estimates for each model ( 1 and 2) were statistically significant. This result indicates that there is indeed an ordered effect among the response categories zero through three. For each model the signs and statistically significant variables were consistent between the parameter estimates and the marginal effects. The control model had the most explanatory variables that were significant determinants of WTP. The willingness to pay for each model calculated using the Delta was statistically s ignificant for all three models, meaning one can be confident that they are an accurate estimate of WTP. Due to the better

PAGE 84

73 model fit, the significant threshold parameters, the higher number of significant variables, and the significant estimated willingne ss the ordered models for prevention, control and mitigation are considered to be superior to the corresponding binary logit models, and is why the impacts of the significant variables are discussed in more detail for those models. Non Parametric Estimates of Willingness to Pay non parametric approach known as the Turnbull lower bound mean approach was also used. The Turnbull method constructs an interval estimate for th e willingness to pay based on the fraction levels for each management strategy. The Turnbull WTP is calculated by multiplying the lower bound of each price in terval by the fraction of the sample estimated to be in that interval and summing the results. This process results in a conservative estimate of WTP. The necessary components for the process, as well as the WTP estimates, are summarized in Table 3 14. The lower bound WTP for the prevention strategy was calculated to be 5.77%. The WTP calculated for the control strategy was $8.96, and the WTP for the mitigation strategy was calculated to be $8.36. Summary of Empirical Results The natural question resu lting from the information above is which model, out of the four for each strategy, is superior. There are a number of determinants that can be used to judge the issue. One criterion that could be used is the set of statistics on model fit. For the bina ry models the model fit statistics gradually got worse as the models became less inclusive (i.e., from Model 1 to Model 3), with the exception of the models for the proposed control strategy. In addition, the model fit statistics were worse for the ordere d models for each strategy, though only very slightly. However, despite this, the statistics indicate that each model is still better than a

PAGE 85

74 constant only specification of the models. Therefore this criterion alone is not enough to determine which model is best. Another criterion that could be applied to determine which model is superior is the number of significant explanatory variables that the model produces. For each strategy, the ordered models had the highest amount of statistically significant de terminants. Finally, the estimated WTPs can be compared and used to determine the best model. The WTP estimates for the binary models receive a mixed review. The discrepancies between the WTPs using the grand constant method and those using the Delta me thod are too large for one to be entirely comfortable with. In addition, the WTPs for prevention Model 3, control Model 3 and mitigation models 2 and 3 are all negative. F inally none of the WTPs using the Delta method from any of the models are statisti cally significant, meaning that very little confidence can be placed in those values. On the other hand, for all three ordered models, the estimated WTP using the Delta method were all positive and significant at the 1% level. Therefore, though the three binary models and the ordered models may be comparable in their ability to determine which characteristics of an individual are significant determinants of willingness to pay, the ultimate goal of this study is to produce an accurate and reasonable estima te of willingness to pay. Thus, with this objective in mind, it can be said that the ordered logit models for the proposed prevention, control and mitigation strategies are superior to their respective binary logit models. Using the corresponding margina l values from the ordered models, it can be determined which explanatory variables were significant determinants of WTP. Though the models for each strategy type had a set of strategy specific explanatory variables, the rest of the independent variables w ere the same. For each model, price was significant and negative. Also significant for each model, but increasing the probability of support for the programs, were the level of concern about red tides, information seeking behavior about red tides, depend ence on water

PAGE 86

75 quality and strategy preference variables. All of these results are intuitive. Obviously, as prices increase there will be less individuals who are willing to pay for a given management strategy. It can also be surmised that individuals wh o are concerned with red tide and those whose livelihoods depend on water quality would be more likely to be willing to pay than those who are not. In the same way, if an individual searches for red tide information more frequently it could be an indicati on that they are more concerned about the issues, making them more likely to be three, it makes intuitive sense that they would be more likely to be willi ng to pay for that strategy. Although they are intuitive, the models are important in that they provide statistical evidence of support and (more importantly) they control for these known factors in the WTP estimates. Surprisingly, the strategy specific variables that indicated whether a respondent would be more economically impacted by the proposed strategy, because they buy fertilizer (P1) or because they own taxable property in Florida (C1) were not statistically significant. It would make sense that using fertilizer or owning property would reduce the probability of being willing to pay, however, this apparently is not the case.

PAGE 87

76 Table 3 1. Description of dependent variables for the WTP models Variable N Mean Std Dev Min Max Prevention: 1,326 0.60 0.49 0.00 1.00 1,326 0.00 1.00 1,326 0.00 1.00 1,326 0.00 3.00 Control: 1,343 0.49 0.50 0.00 1.00 1,343 0.00 1.00 1,343 0.00 1.00 1,343 0.00 3.00 Mitigation: 1,320 0.36 0.48 0.00 1.00 1,320 0.00 1.00 1,320 0.00 1.00 1,320 0.00 3.00

PAGE 88

77 Table 3 2 Description of independent variables for the WTP models Variable N Mean Std Dev Min Max P_Price (1%, 5%, or 10% tax on all fertilizer sales) 1,454 5.33 3.73 1.00 10.00 C_Price ($5, $10, or $15 ad valorem fee on all taxable property) 1,454 9.99 4.14 5.00 15.00 M_Price ($5, $15 or $25 donation to trust fund) 1,454 14.9 8.28 5.00 25.00 P1 (1 if uses fertilizer, 0 else) 1,321 0.55 0.49 0.00 1.00 M1 (1 if heard of BCRS, 0 else) 1,334 0.25 0.44 0.00 1.00 M5 (number of days spent at beach) 1,088 48.9 87.4 0.00 365.00 C1 (1 if pays property tax in FL, 0 else) 1,344 0.87 0.33 0.00 1.00 1,301 0.23 0.42 0.00 1.00 Pref_P (1 if prevention was preferred strategy, 0 else) 1,311 0.43 0.50 0.00 1.00 Pref_C (1 if control was preferred strategy, 0 else) 1,311 0.20 0.40 0.00 1.00 Pref_M (1 if mitigation was preferred strategy, 0 else) 1,311 0.17 0.37 0.00 1.00 PFST (1 if prevention was ordered first, 0 else) 1,454 0.35 0.48 0.00 1.00 CFST (1 if control was ordered first, 0 else) 1,454 0.33 0.47 0.00 1.00 MFST (1 if mitigation was ordered first, 0 else) 1,454 0.32 0.47 0.00 1.00 Q2 (level of experience with red tide, 1 through 7) 1,452 3.55 2.07 0.00 7.00 Q3 ( number of true/false questions answered correctly, 1 through 10) 1,454 3.64 2.11 0.00 10.00 Q4 (1 if concerned with red tide issues, 0 else) 1,302 0.76 0.42 0.00 1.00 Q6A (1 if never, 2 if sometimes, 3 if frequently) 1,330 1.85 0.53 1.00 3.00 Q9A (1 if very dependent on coastal water quality, 0 else) 1,338 0.41 0.49 0.00 1.00 Q16 (number of years residing in FL) 1,445 24.15 16.87 0.00 82.00 Q17 (number of months per year residing in FL) 1,439 11.73 1.21 0.00 12.00 Q19 (miles of residence to coast) 1,428 8.67 10.91 0.00 200.00 AGE (age of respondent in years) 1,426 59.92 14.38 18.00 96.00 Q21 A (education level ) 1,430 4.52 1.28 1.00 6.00 Q22 (1 if Caucasian, 0 else) 1,433 0.94 0.24 0.00 1.00 Q23 A (income level) 1,322 2.80 1.40 1.00 6.00 DV_EST (1 if from east coast region, 0 else) 1,454 0.42 0.49 0.00 1.00 DV_SW (1 if from southwest coast region, 0 else) 1,454 0.48 0.50 0.00 1.00

PAGE 89

78 Table 3 3. Estimated binary logit models for the proposed prevention strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 1.00 2.143 0.072 3.297 0.005 4.059 0.001 P_Price 5.36 0.070 0.001 0.074 0.000 0.057 0.006 P1 0.56 0.052 0.747 0.108 0.483 0.181 0.255 PFST 0.36 0.163 0.310 0.192 0.213 0.128 0.418 DV_EST 0.39 1.078 0.000 1.024 0.000 0.215 0.444 DV_SW 0.51 0.844 0.001 0.720 0.005 0.314 0.257 AGE 59.92 0.005 0.398 0.008 0.162 0.013 0.026 Pref_P 0.44 2.162 0.000 1.816 0.000 1.329 0.000 Q16 24.07 0.010 0.042 0.005 0.261 0.009 0.079 Q17 11.77 0.001 0.989 0.055 0.403 0.004 0.951 Q19 8.54 0.004 0.557 0.008 0.249 0.002 0.798 Q2 3.81 0.010 0.828 0.003 0.948 0.017 0.719 Q21A 15.4 0.031 0.453 0.038 0.332 0.068 0.098 Q22 0.95 0.377 0.294 0.369 0.292 0.719 0.083 Q23A 73.23 0.002 0.214 0.003 0.140 0.003 0.163 Q3 4.01 0.012 0.771 0.016 0.693 0.009 0.834 Q4 0.76 0.739 0.000 0.906 0.000 0.665 0.003 Q6A 1.87 0.354 0.032 0.505 0.002 0.489 0.005 Q9A 0.41 0.222 0.179 0.273 0.084 0.568 0.000 Model Statistics: N 1,016 1,016 1,016 Log Likelihood 525.97 557.706 525.007 McFadden R 2 0.217 0.196 0.142 Chi Squared 291.33 0.000 272.010 0.000 174.201 0.000 Correct Pred. (%) 73.92 72.146 72.539 WTP (Delta) 16.35% 0.992 11.15% 0.919 14.02% 0.919 WTP (Grand Constant) 10.28% 5.17% 19.38%

PAGE 90

79 Table 3 4. Estimated binary logit models for the proposed control strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 2.607 0.032 2.343 0.055 2.697 0.055 C_Price 10.000 0.074 0.000 0.082 0.000 0.066 0.003 C1 0.876 0.120 0.621 0.163 0.506 0.003 0.993 C5N 0.221 1.405 0.000 1.478 0.000 1.320 0.000 CFST 0.308 0.418 0.010 0.412 0.011 0.120 0.535 DV_EST 0.381 0.729 0.006 0.711 0.009 0.578 0.100 DV_SW 0.509 0.565 0.030 0.490 0.066 0.317 0.360 AGE 58.136 0.008 0.185 0.003 0.633 0.006 0.404 Pref_C 0.206 1.856 0.000 1.704 0.000 0.982 0.000 Q16 23.811 0.013 0.008 0.010 0.036 0.009 0.129 Q17 11.760 0.017 0.791 0.032 0.640 0.108 0.129 Q19 8.684 0.003 0.602 0.005 0.457 0.004 0.599 Q2 3.818 0.035 0.439 0.025 0.594 0.035 0.544 Q21A 15.393 0.068 0.089 0.064 0.119 0.060 0.219 Q22 0.952 0.528 0.155 0.816 0.040 0.563 0.280 Q23A 73.185 0.004 0.056 0.004 0.041 0.002 0.306 Q3 4.028 0.086 0.034 0.060 0.149 0.003 0.949 Q4 0.762 0.605 0.002 0.895 0.000 0.907 0.004 Q6A 1.863 0.199 0.036 0.297 0.075 0.298 0.146 Q9A 0.417 0.330 0.018 0.527 0.001 0.684 0.000 Model Statistics: N 975 975 975 Log Likelihood 534.064 525.190 391.625 McFadden R 2 0.210 0.218 0.138 Chi Squared 283.483 0.000 292.374 0.000 125.267 0.000 Correct Pred. (%) 70.560 71.282 83.487 WTP (Delta) $10.17 0.901 $7.01 0.897 $16.99 0.879 WTP (Grand Constant) $0.70 $3.43 $28.79

PAGE 91

80 Table 3 5. Estimated binary logit models for the proposed mitigation strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 2.257 0.077 1.748 0.176 2.051 0.190 M_Price 0.912 0.062 0.000 0.062 0.000 0.065 0.000 M1 0.077 0.309 0.092 0.433 0.020 0.368 0.122 M5 0.135 0.003 0.005 0.002 0.029 0.003 0.015 MFST 0.017 0.051 0.760 0.157 0.362 0.102 0.653 DV_EST 0.005 0.014 0.960 0.421 0.160 0.466 0.256 DV_SW 0.130 0.260 0.350 0.573 0.054 0.651 0.111 AGE 0.227 0.004 0.523 0.001 0.934 0.002 0.835 Pref_M 0.125 0.737 0.000 0.582 0.006 0.740 0.004 Q16 0.072 0.003 0.551 0.001 0.785 0.008 0.212 Q17 0.641 0.054 0.445 0.113 0.130 0.125 0.132 Q19 0.091 0.010 0.136 0.000 0.991 0.018 0.149 Q2 0.204 0.054 0.276 0.065 0.214 0.011 0.872 Q21A 0.907 0.059 0.156 0.084 0.052 0.051 0.351 Q22 0.316 0.330 0.401 0.573 0.149 0.629 0.216 Q23A 0.054 0.001 0.716 0.001 0.714 0.001 0.590 Q3 0.067 0.017 0.697 0.038 0.401 0.028 0.630 Q4 0.795 1.058 0.000 1.133 0.000 1.378 0.000 Q6A 0.402 0.185 0.073 0.307 0.104 0.150 0.532 Q9A 0.373 0.159 0.299 0.097 0.593 0.434 0.064 Model Statistics: N 836 836 836 Log Likelihood 476.351 457.211 303.529 McFadden R 2 0.122 0.120 0.125 Chi Squared 132.770 0.000 125.213 0.000 86.859 0.000 Correct Pred. (%) 69.856 72.010 84.994 WTP (Delta) $2.98 0.915 $0.12 0.989 $16.37 0.863 WTP (Grand Constant) $11.70 $28.20 $32.09

PAGE 92

81 Table 3 6. Estimated marginal values for the binary logit models of the prevention strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 1.00 0.471 0.073 0.795 0.005 0.762 0.001 P_Price 5.36 0.015 0.001 0.018 0.000 0.011 0.006 P1 0.56 0.011 0.747 0.026 0.483 0.034 0.252 PFST 0.36 0.036 0.313 0.047 0.214 0.024 0.422 DV_EST 0.39 0.224 0.000 0.236 0.000 0.041 0.449 DV_SW 0.51 0.184 0.009 0.172 0.004 0.059 0.256 AGE 59.92 0.001 0.398 0.002 0.162 0.002 0.025 Pref_P 0.44 0.431 0.000 0.405 0.000 0.256 0.000 Q16 24.07 0.002 0.042 0.001 0.261 0.002 0.078 Q17 11.77 0.000 0.989 0.013 0.403 0.001 0.951 Q19 8.54 0.001 0.557 0.002 0.249 0.000 0.798 Q2 3.81 0.002 0.828 0.001 0.948 0.003 0.719 Q21A 15.4 0.007 0.453 0.009 0.332 0.013 0.098 Q22 0.95 0.087 0.313 0.091 0.298 0.113 0.032 Q23A 73.23 0.001 0.214 0.001 0.140 0.001 0.163 Q3 4.01 0.003 0.771 0.004 0.693 0.002 0.834 Q4 0.76 0.171 0.000 0.221 0.000 0.114 0.001 Q6A 1.87 0.078 0.032 0.122 0.002 0.092 0.004 Q9A 0.41 0.048 0.175 0.065 0.081 0.109 0.001

PAGE 93

82 Table 3 7. Estimated marginal values for the binary logit models of the control strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 0.439 0.141 0.574 0.054 0.304 0.054 C_Price 10.000 0.019 0.000 0.020 0.000 0.007 0.003 C1 0.876 0.038 0.530 0.040 0.509 0.000 0.993 C5N 0.221 0.324 0.000 0.319 0.000 0.115 0.000 CFST 0.308 0.107 0.007 0.102 0.011 0.014 0.541 DV_EST 0.381 0.184 0.004 0.174 0.008 0.069 0.119 DV_SW 0.509 0.142 0.026 0.120 0.063 0.036 0.360 AGE 58.136 0.002 0.212 0.001 0.633 0.001 0.402 Pref_C 0.206 0.405 0.000 0.398 0.000 0.136 0.000 Q16 23.811 0.003 0.013 0.003 0.036 0.001 0.128 Q17 11.760 0.006 0.722 0.008 0.640 0.012 0.128 Q19 8.684 0.001 0.637 0.001 0.457 0.000 0.599 Q2 3.818 0.010 0.393 0.006 0.594 0.004 0.544 Q21A 15.393 0.014 0.166 0.016 0.119 0.007 0.219 Q22 0.952 0.123 0.163 0.182 0.018 0.052 0.182 Q23A 73.185 0.001 0.072 0.001 0.041 0.000 0.305 Q3 4.028 0.022 0.028 0.015 0.149 0.000 0.949 Q4 0.762 0.160 0.001 0.207 0.000 0.086 0.000 Q6A 1.863 0.059 0.149 0.073 0.075 0.034 0.145 Q9A 0.417 0.109 0.006 0.129 0.001 0.081 0.001

PAGE 94

83 Table 3 8. Estimated marginal values for the binary logit models of the control strategy Model 1 Model 2 Model 3 Variable Mean Pr > z Pr > z Pr > z Constant 0.481 0.085 0.353 0.260 0.199 0.187 M_Price 0.912 0.013 0.000 0.013 0.000 0.006 0.000 M1 0.077 0.071 0.093 0.091 0.025 0.038 0.149 M5 0.135 0.001 0.002 0.000 0.029 0.000 0.016 MFST 0.017 0.008 0.821 0.032 0.368 0.010 0.648 DV_EST 0.005 0.005 0.940 0.087 0.166 0.047 0.278 DV_SW 0.130 0.064 0.296 0.115 0.052 0.064 0.115 AGE 0.227 0.001 0.684 0.000 0.934 0.000 0.835 Pref_M 0.125 0.170 0.001 0.126 0.009 0.087 0.015 Q16 0.072 0.001 0.582 0.000 0.785 0.001 0.211 Q17 0.641 0.015 0.362 0.023 0.130 0.012 0.133 Q19 0.091 0.002 0.135 0.000 0.991 0.002 0.145 Q2 0.204 0.012 0.292 0.013 0.213 0.001 0.872 Q21A 0.907 0.015 0.102 0.017 0.052 0.005 0.351 Q22 0.316 0.073 0.438 0.127 0.179 0.076 0.306 Q23A 0.054 0.000 0.778 0.000 0.714 0.000 0.590 Q3 0.067 0.004 0.643 0.008 0.401 0.003 0.630 Q4 0.795 0.210 0.000 0.198 0.000 0.105 0.000 Q6A 0.402 0.083 0.039 0.062 0.103 0.015 0.532 Q9A 0.373 0.014 0.711 0.020 0.594 0.044 0.073

PAGE 95

84 Table 3 9. Estimated ordered logit models for the prevent, control, and mitigation strategies Prevention Control Mitigation Variable Pr > z Pr > z Pr > z Constant 2.092 0.031 1.192 0.252 1.995 0.095 P_Price 0.062 0.000 --------P1 0.065 0.613 --------PFST 0.065 0.617 --------Pref_P 1.688 0.000 --------C_Price ----0.072 0.000 ----C1 ----0.126 0.548 ----C5N ----1.454 0.000 ----CFST ----0.306 0.028 ----Pref_C ----1.356 0.000 ----M_Price --------0.063 0.000 M1 --------0.334 0.052 M5 --------0.003 0.002 MFST --------0.046 0.775 Pref_M --------0.686 0.000 DV_EST 0.745 0.001 0.727 0.003 0.280 0.301 DV_SW 0.599 0.007 0.500 0.036 0.484 0.071 AGE 0.008 0.109 0.000 0.968 0.002 0.738 Q16 0.009 0.019 1.356 0.000 0.000 0.990 Q17 0.003 0.951 0.011 0.015 0.088 0.201 Q19 0.003 0.556 0.074 0.204 0.003 0.523 Q2 0.015 0.695 0.005 0.444 0.053 0.268 Q21A 0.042 0.196 0.024 0.560 0.075 0.060 Q22 0.454 0.127 0.064 0.071 0.433 0.238 Q23A 0.003 0.074 0.620 0.067 0.000 0.943 Q3 0.012 0.728 0.004 0.035 0.031 0.460 Q4 0.692 0.000 0.058 0.103 1.117 0.000 Q6A 0.388 0.005 0.721 0.000 0.363 0.039 Q9A 0.385 0.004 0.312 0.030 0.153 0.362 Mu(1) 0.316 0.000 0.522 0.000 0.202 0.000

PAGE 96

85 Mu(2) 1.812 0.000 0.264 0.000 1.321 0.000 Model Statistics: N 1,004 975 836 Log Likelihood 1,100.791 1,127.715 752.155 McFadden R 2 0.123 0.134 0.087 Chi Squared 305.341 0.000 302.864 0.000 144.240 0.000 WTP (Delta) 19.40% 0.000 $10.11 0.000 $3.16 0.140 Std. Err. (Delta) 3.73% $1.05 $2.14

PAGE 97

86 Table 3 10. Estimated marginal values for the ordered logit models of the prevention strategy Y=0 Y=1 Y=2 Y=3 Variable Pr > z Pr > z Pr > z Pr > z P_Price 0.014 0.000 0.001 0.001 0.004 0.001 0.011 0.000 P1 0.015 0.613 0.001 0.611 0.004 0.616 0.012 0.612 PFST 0.015 0.618 0.001 0.611 0.004 0.625 0.012 0.615 DV_EST 0.162 0.001 0.015 0.002 0.036 0.001 0.141 0.001 DV_SW 0.134 0.007 0.010 0.007 0.036 0.011 0.108 0.007 AGE 0.002 0.109 0.000 0.117 0.000 0.119 0.001 0.111 Pref_P 0.355 0.000 0.028 0.000 0.067 0.000 0.316 0.000 Q16 0.002 0.019 0.000 0.024 0.001 0.028 0.002 0.019 Q17 0.001 0.951 0.000 0.951 0.000 0.951 0.001 0.951 Q19 0.001 0.556 0.000 0.557 0.000 0.558 0.001 0.556 Q2 0.003 0.695 0.000 0.695 0.001 0.695 0.003 0.695 Q21A 0.010 0.196 0.001 0.202 0.003 0.204 0.008 0.197 Q22 0.108 0.141 0.005 0.001 0.039 0.221 0.073 0.084 Q23A 0.001 0.074 0.000 0.082 0.000 0.085 0.001 0.075 Q3 0.003 0.728 0.000 0.728 0.001 0.728 0.002 0.728 Q4 0.162 0.000 0.008 0.000 0.056 0.001 0.114 0.000 Q6A 0.087 0.005 0.007 0.008 0.024 0.009 0.071 0.005 Q9A 0.086 0.003 0.007 0.008 0.022 0.004 0.071 0.005

PAGE 98

87 Table 3 11. Estimated marginal values for the binary logit models of the control strategy Y=0 Y=1 Y=2 Y=3 Variable Pr > z Pr > z Pr > z Pr > z C_Price 0.018 0.000 0.000 0.043 0.010 0.000 0.008 0.000 C1 0.032 0.547 0.000 0.389 0.017 0.540 0.014 0.562 C5N 0.334 0.000 0.021 0.000 0.196 0.000 0.117 0.000 CFST 0.076 0.027 0.001 0.166 0.041 0.024 0.034 0.036 DV_EST 0.180 0.002 0.002 0.289 0.095 0.001 0.083 0.005 DV_SW 0.124 0.034 0.003 0.121 0.068 0.033 0.053 0.037 AGE 0.000 0.968 0.000 0.968 0.000 0.968 0.000 0.968 Pref_C 0.317 0.000 0.009 0.013 0.132 0.000 0.193 0.000 Q16 0.003 0.015 0.000 0.099 0.001 0.015 0.001 0.015 Q17 0.018 0.204 0.000 0.269 0.010 0.205 0.008 0.205 Q19 0.001 0.444 0.000 0.469 0.001 0.444 0.000 0.444 Q2 0.006 0.560 0.000 0.572 0.003 0.560 0.003 0.560 Q21A 0.016 0.071 0.000 0.159 0.009 0.072 0.007 0.072 Q22 0.150 0.052 0.008 0.229 0.089 0.063 0.053 0.021 Q23A 0.001 0.035 0.000 0.127 0.001 0.036 0.000 0.036 Q3 0.015 0.103 0.000 0.190 0.008 0.104 0.006 0.104 Q4 0.176 0.000 0.008 0.016 0.102 0.000 0.067 0.000 Q6A 0.078 0.030 0.002 0.119 0.043 0.030 0.033 0.031 Q9A 0.130 0.000 0.002 0.105 0.070 0.000 0.058 0.000

PAGE 99

88 Table 3 12. Estimated marginal values for the binary logit models of the mitigation strategy Y=0 Y=1 Y=2 Y=3 Variable Pr > z Pr > z Pr > z Pr > z M_Price 0.014 0.000 0.001 0.000 0.006 0.000 0.006 0.000 M1 0.075 0.058 0.005 0.036 0.034 0.051 0.036 0.070 M5 0.001 0.002 0.000 0.003 0.000 0.002 0.000 0.002 MFST 0.010 0.775 0.001 0.774 0.005 0.775 0.005 0.776 DV_EST 0.062 0.305 0.005 0.285 0.028 0.299 0.029 0.315 DV_SW 0.106 0.069 0.008 0.065 0.049 0.067 0.049 0.074 AGE 0.000 0.738 0.000 0.738 0.000 0.738 0.000 0.738 Pref_M 0.159 0.001 0.009 0.000 0.068 0.000 0.082 0.002 Q16 0.000 0.990 0.000 0.990 0.000 0.990 0.000 0.990 Q17 0.019 0.201 0.001 0.202 0.009 0.201 0.009 0.203 Q19 0.001 0.522 0.000 0.524 0.000 0.523 0.000 0.522 Q2 0.012 0.268 0.001 0.270 0.005 0.268 0.005 0.268 Q21A 0.016 0.060 0.001 0.063 0.008 0.060 0.008 0.061 Q22 0.100 0.259 0.006 0.118 0.044 0.223 0.051 0.305 Q23A 0.000 0.943 0.000 0.943 0.000 0.943 0.000 0.943 Q3 0.007 0.460 0.001 0.461 0.003 0.460 0.003 0.461 Q4 0.216 0.000 0.020 0.000 0.104 0.000 0.092 0.000 Q6A 0.079 0.039 0.006 0.042 0.037 0.040 0.036 0.040 Q9A 0.034 0.364 0.003 0.356 0.016 0.362 0.016 0.368

PAGE 100

89 Table 3 13. Summary of estimated marginal values for the extreme levels from the ordered logit models for each strategy Prevention Control Mitigation Variable y=0 y=3 y=0 y=3 y=0 y=3 P_Price 0.014* 0.012* --------P1 0.015 0.012 --------PFST 0.019 0.015 --------Pref_P 0.352* 0.314* --------C_Price ----0.018* 0.008* ----C1 ----0.032 0.014 ----C5N ----0.334* 0.117* ----CFST ----0.076* 0.034* ----Pref_C ----0.317* 0.193* ----M_Price --------0.014* 0.006* M1 --------0.075* 0.036* M5 --------0.001* 0.000* MFST --------0.010 0.005 Pref_M --------0.159* 0.082* DV_EST 0.163* 0.142* 0.180* 0.083* 0.062 0.029 DV_SW 0.138* 0.111* 0.124* 0.053* 0.106* 0.049* AGE 0.002* 0.001* 0.000 0.000 0.000 0.000 Q16 0.002* 0.002* 0.003* 0.001* 0.000 0.000 Q17 0.002 0.002 0.018 0.008 0.019 0.009 Q19 0.001 0.001 0.001 0.000 0.001 0.000 Q2 0.003 0.003 0.006 0.003 0.012 0.005 Q21A 0.010 0.008 0.016* 0.007* 0.016* 0.008* Q22 0.123* 0.082* 0.150* 0.053* 0.100 0.051 Q23A 0.001* 0.001* 0.001* 0.000* 0.000 0.000 Q3 0.002 0.002 0.015 0.006 0.007 0.003 Q4 0.162* 0.114* 0.176* 0.067* 0.216* 0.092* Q6A 0.095* 0.077* 0.078* 0.033* 0.079* 0.036* Q9A 0.084* 0.070* 0.130* 0.058* 0.034 0.016 *S ignificant at the 10% level

PAGE 101

90 Table 3 14. Turnbull estimates using Model 1 data Group j Fee Fee Range (N j ) Total obs. (T j ) CDF=F j (N j /T j ) PDF=P j (F j F j 1 ) E LB (WTP) Prevention: 0 1% 0 1% 156 454 0.344 0.344 0.00% 1 5% 1 5% 169 415 0.407 0.063 0.06% 2 10% 5 10% 206 457 0.451 0.044 0.22% 3 >10% 1.000 0.549 5.49% Total 531 1,326 1.000 5.77% Std. Err. 0.15% Control: 0 $5.00 $0 5 186 461 0.403 0.403 $0.00 1 $10.00 $5 10 237 418 N/A N/A N/A 2 $15.00 $10 15 258 464 N/A N/A N/A 3 >$15.00 1.000 0.597 $8.96 Total 681 1,343 1.000 $8.96 $0.11 Mitigation: 0 $5.00 $0 5 236 455 0.519 0.519 $0.00 1 $15.00 $5 15 272 413 0.659 0.140 $0.70 2 $25.00 $15 25 337 452 0.746 0.087 $1.31 3 >$25.00 1.000 0.254 $6.35 Total 845 1,320 1.000 $8.36 $0.33 *Note: The standard error is the square root of the variance

PAGE 102

91 CHAPTER 4 CONCLUSION AND DISCU SSION H armful algal blooms (HABS) are natural events with ecological and economic consequences worldwide. In Florida, Karenia brevis is the algae species that has accounted for nearly all of blooms. This algae species is unique in that the toxins produced during the bloom are a neurotoxin that can kill fish and marine mammals as well as become airborne and affect the respiratory system of humans. Red tides are potentially disastrous to a state like Florida that is heavily dependent on coastal tourism and commercial fishing. The three most common types of strategies for dealing with red tide are control, mitigation and prevention strategies. Research is continually being done on all three types of management practice s ; however it is possible that some str ategies may face opposition from the public. This study attempts to determine public preference for the three different types of management practices by conducting a mail survey of Florida residents in coastal areas where red tide is a frequent nuisance. The main findings from the study are summarized below: Most respondents were at least somewhat aware of the issue of red tide in Florida. Many indicated they had at least experienced red tide conditions in the water at some point in their lives. A consi derable number of respondents revealed they were largely unaware if fish, shellfish and oysters were safe to consume during a red tide event. A significant percentage of those indicated they were, indeed, concerned about the issue. The majority of those who were concerned indicated they were concerned mostly because red tides negatively affect human health and also prevent beach and water activities. Despite their concern, most respondents were unfamiliar with the existing sources of red tide information such as START, FWRI, FRTA and FRTC. They were also fairly unfamiliar with the stat us of research on red tide (i.e., effectiveness and results produced to date ). For the prevention strategy, 60 percent of respondents indicated they were willing to pay t he proposed tax. Of those, 46 percent were very sure of their response. Forty nine percent of respondents indicated they would be willing to pay for the proposed control strategy, and of those, 34 percent indicated they were very sure of their decision. In addition, when asked what their preferred method of control was, the majority indicated they preferred biological controls to chemical controls (or neither). For the mitigation strategy 36 percent indicated they would be willing to pay the proposed do nation. Of those respondents, 44 percent were very sure of their answer.

PAGE 103

92 Three binary models were run for each type of strategy (prevention, control and mitigation). All models were found to be of adequate fit and predictive power. For the prevention strategies, the model fit and predictive power decreased as the models became less inclusive of a definition of a success (i.e., a yes answer). The same can b e said for the mitigation models. For the control strategy, the models actually improved from Mo del 1 to Model 2 and then declined from Model 2 to Model 3. The statistically significant variables, and their signs, were consistent between models. The variables that were significant across strategy types included price, strategy preference, level of concern and information seeking behaviors. In addition to the binary logit models, one ordered logit model was run for each strategy type. The ordered models were found to be superior to the binary models These, too, were all found to be of adequate fit and predictive power. For the prevention, control and mitigation models the threshold parameters were all statistically significant, indicating the models were indeed ordered. The price level, age length of Florida residency and income level were found to have a negative impact on WTP for the prevention strategy, while being in the southwest or northeast regions, being concerned about red tide, seeking information on red tide frequently and being dependent on local water quality all increased the probab ility of being WTP. For the control strategy a high price level, a longer length of residency, and a higher level of knowledge about red tide all decreased the likelihood of being WTP, while residing in the northeast or southwest region, being of Caucasia n descent, being concerned about red tide, seeking information on red tide frequently and being dependent on local water quality all increased the probability of being WTP. Finall y, for the mitigation strategy only higher price level reduced the probabili ty of being WTP at a significant level. However, residing in the southwest region, spending more days at the beach, being familiar with the BCRS, being concerned about red tide and seeking information on red tide frequently all increased the likelihood of being WTP. Using the Delta method from the ordered logit model, the prevention strategy was found to have an estimated WTP of 19.4 percent per sale of fertilizer. The Turnbull lower bound calculation yielded a lower bound WTP of 5.77 percent. For contro l, the estimated WTP using the Delta method from the ordered model was $10.11, while the Turnbull lower bound WTP was $8.96. Finally, the Delta method estimated WTP for the mitigation strategy was $3.16, while the Turnbull lower bound WTP was $8.36. All of the estimated WTP using the Delta method were statistically significant at the 1% level, indicating that the estimates are accurate. A total of 115 complete responses were received from 692,432 residents in the 12 county study r egion that were invited t o complete the survey online by email from Expedite Media Group. The low response rate (0.02%) is not a surprise since Expedite is a marketing s ervice; recipients receive multiple solicitations each day. For this campaign, 10.2% v iewed the message and 1.4% of those began the survey. This response is comparable to o ther surveys conducted through Expedite. The mail survey and internet respondents were similar with respect to many factors, for example the average age was identical (60 years), distance from their residence to the coast

PAGE 104

93 (95% are more than 20 miles), race (92% 95% Caucasian), and residency during the year (95% spend more than 9 months per year in Florida); however, there were notable differences. The internet respondents had generally lived in Florida longer, are more highly educated and have higher incomes. The internet sample reported more experience with all seven of the potential negative effects of red tide (by one to 19 percentage points). The internet sample had a higher percentage of c orrect answers to all 10 questions regarding red tide knowledge (by one to 18 percentage points). The internet respondents were more familiar with the red tide information provided by Universities, Mote Marine Lab, and Sierra Club (by at least 10 percenta ge points). The internet respondents are more likely to obtain red tide information by internet (58% to 34% for mail survey respondents). The internet respondents reported being more dependent on coastal water quality or quantity than mail survey respond ents (57% versus 40%, respectively, reported being The internet respondents were less willing to support any of the three red tide strategies, from four to 18 percentage points less for the prevention and mitigation strategies, respectively. These findings have several potentially significant implications. First, from the answers to question three in the survey it is apparent that there is a lack of knowledge among the public regar ding the causes and effects of red tide, especially with regard to the safety of seafood consumption. Although this lack of knowledge did not significantly impact willingness to pay, it is still a cause for concern and additional education and outreach ef forts should made throughout the state. The WTP findings indicate that the public is most willing to pay for prevention programs, followed by control programs, with mitigation programs being the management strategy that the y are least likely to pay for T he strategy that had the most support overall was prevention. This is at first slightly surprising, since it was emphasized in the survey that this strategy carried the most amount of uncertainty with regards to its effectiveness for dealing with red tide However, other studies have shown that people are more likely to be willing to pay when humans are part of the cause of the environmental problem in question (Bulte et al. 2005) Since it was indicated that the prevention strategy was more directly related to human causes than the control or mitigation strategies the high level of support for the prevention strategy could be expected. In addition, it was indicated to the respondents that the prevention strategy

PAGE 105

94 would improve overall water quality. quality is driving the high level of support, which suggests that red tide issues be included in general water quality programs. These findings should be taken into account as new programs are devised and introduced to the public. Though a significant amount of confidence can be placed in the results found by this study, a few caveats should be mentioned. First, the WTP estimates for the mitigation strategy may be (Polome et al. 2006) This occurs when the respondent does not understand the provided description of the public good they are being asked to value. For the mitigation strategy many respondents indicated their confusion over whether the p roposed program would be offered in their region. responses than would have been given had the respondents understood the description of the strategy, resulting in WTP estimates that are too low. In addition, the W TP r esults may be low efficiently (Polome et al. 2006) Many respondents, for all three strategy types, indicated doubt that the hypothetical programs could or would be implemented in the manner indicated in the description there w as complete confidence that the strategy in question could be implemented easi ly and efficiently management practice, and would lower the estimated WTP for each strategy. There are several steps that should be addressed in the future. First, it is apparent from the survey respon ses and the estimated coefficients from the binary and ordered models that new models should be estimated based on the three regions included in the study. Addit ionally,

PAGE 106

95 amount that they would be willing to pay at. In essence, the respondents who indicated that they would be willing to pay at a different price level could be c new price level. In addition, the study sample population should be compared to the total population of Florida using current census data, to ensure that the sample population is representative of the total population. Finally, the estimated WTPs from this study should be extrapolated to the relevant population to calculate the total WTP that can be compared to the total costs of the proposed programs. Total WTP can be extrapolated using the formula: where EV j is the economic value of prop osed strategy j across the i study regions (NW=northwest, NE=northeast, SW=southwest), which is calculated by multiplying the estimated household level willingness to pay estimate for each region, number of households in each region ( H ), and the total cost per household ( V ). Note that V will be zero for the mitigation strategy, it will equal the average fertilizer expenditures for the prevention strategy (e.g., average pounds us ed times retail cost per pound) and the median assessed value of private reside nces divided by $100,000 for the control strategy.

PAGE 107

96 APPENDIX A MAIL SURVEY COVER LE TTER AND QUESTIONNAI RE

PAGE 108

97 Food and Resource Economics Department PO Box 110240 Gainesville, FL 32611 0240 U.S.A Telephone: (352) 392 1845 January 20 10 Dear Florida Resident: We need your help in assessing awareness of Florida red tide events and issues related to this your level of concern for knowledge about and experience with red tide ev ents in Florida. We also ask about where you get information on red tides and your preferences for strategies to address future red tides This information will help education and outreach efforts and provide valuable information to the state as it decides how best to spend its limited resources. Your household was randomly selected to participate in this survey by the Florida Survey Research Center at the University of Florida. We do not have the mailing list and are not linking addresses with responses s ince w e are only interested in the response of the sample as a whole Thus, by responding to this survey you cannot not be solicited by any of the agencies or groups mentioned in the survey. The questionnaire should take about 10 minutes to complete. Since this survey is being conducted through the University of Florida, it has been approved by the Institutional Review Board (IRB). According to this approval we need to tell you that there are no direct benefits or risks to you for answering the questio ns. Participation is voluntary and you will not be compensated. Your identity will remain anonymous. You can not be penalized for choosing not to answer certain questions. If you have any questions, please contact me directly at 1 352 846 2874 For questio ns about your rights as a research participant, please contact the IRB at 1 352 392 0433 (protocol# 2009 U 1087 ). Thank you in advance for your time in helping us to better understand how Florida residents feel about red tide events in Florida Best regards, Michael J. Scicchitano, Ph.D. Director, FSRC mscicc@ufl.edu Pleas e answer today and mail using the self addressed, postage paid envelope, even if you are

PAGE 109

98 1. Are you No (last page). 2. What has been your experience w ith red tide events in Florida? Circle one for each. I have noticed red tide conditions in the water Yes No I have seen dead animals on the shore during a red tide Yes No I have experienced the odor of decaying fish on the beach Yes No I have (or a member of my family has) experienced burning eyes, scratchy throat, or coughing that could have been from a red tide Yes No I have changed plans to visit a beach because of a red tide event Yes No I have changed hotel reservations because of a red tide event Yes No I have changed a restaurant reservation because of a red tide event Yes No 3. Do you believe each statement is True (T) or False (F) with respect to red tides in Florida? Red tide conditions can vary greatly from one area to another (within a few miles) due to wind s and currents T F DK Seafood bought in stores or restaurants is safe to eat during red tides T F DK Recreationally caught shrimp and crab are safe to eat during a red tide T F DK Recreationally caught finfish are unsafe to eat during a red tide T F DK Recreationally caught oysters are unsafe to eat during a red tide T F DK People with asthma are more likely to notice the effects of red tide T F DK T F DK Reddish brown water indicates that humans will experience respiratory problems T F DK The algae that causes red tides is always present in the Gulf of Mexico T F DK Red tides are the same all over the world T F DK Awareness Experience, Knowledge, and Concern

PAGE 110

99 4. How concerned are you, if at all, about Florida red tide events? OR 4A. What is the one reason you are generally not concerned about red tide events in Florida ? 4B. What is the one reason you are at least somewhat concerned about red tide events in Florida ? Red tides have not affected me Red tides cause economic losses Red tides are unpredictable so being concerned serves no purpose Red tides prevent fishing, beach going, and other marine activities Scientists are working on the issue Red tides affect human health Red tides are a natural occurrence Red tides indicate poor water quality Other: Other: 5. Do you agree or disagree with each of the following statements concerning scientific research on red tides in Florida? Please use the following 5 point rating scale: disagree know agree Rating S cientific research on red tides in Florida has generated a lot of knowledge S cientific research on red tides in Florida has generated practical applications Results from the scientific research on Florida red tides is confusing Better monitoring and prediction systems are needed for red tides in Florida Learning about how Florida red tides affect marine animals is important Learning about how Florida red tides affect hu man health is important Learning about how people are affected by Florida red tides is important Learning about how we can control or prevent Florida red tides is important Determining the costs and benefits of different red tide strategies is important Information

PAGE 111

100 6. Which of the following statements best describes how frequently you seek information about Florida red tides? Please check one. I never look for information about red tide events in Florida I only look for information when a red tide event affects near shore waters I only look for information when something new is reported about Florida red tides I look for information about Florida red tides on a regular basis to see what is new 7. How familiar are you with red tide information available from each of the following agencies and organizations in Florida? Please circle one number for each. Not at all Somewhat Very FWRI (Fish & Wildlife Research Institute) 1 2 3 Universities 1 2 3 Mote Marine Lab 1 2 3 Sierra Club 1 2 3 START (Solutions to Avoid Red Tide) 1 2 3 Florida Red Tide Coalition 1 2 3 Florida Red Tide Alliance 1 2 3 Beach Conditions Reporting System 1 2 3 8. How frequently do you get Florida red tide information from each source? Please circle one number for each. Never Sometimes Often Television 1 2 3 Radio 1 2 3 Local newspapers 1 2 3 Internet websites 1 2 3 Public forums, meetings, or workshops 1 2 3 Printed brochures or pamphlets 1 2 3 Friends or family 1 2 3 9. How dependent are you on coastal water quality and quantity ? Check one. Not at all dependent since I Somewhat dependent since I have a lawn and occasionally visit the beach or fish Very dependent since water is important to my recreation or livelihood

PAGE 112

101 We will describe three different types of red tide programs: prevention mitigation and control ast program is described, we will ask you which you would prefer the most (if any). PREVENTION: Background: We know that plants including algae need nutrients to grow and that when excess nutrients drain from the land to the sea, these waters can from land can be naturally occurring or can come from the human use of fertilizers for lawns and crops. We know that sometimes this enrichment can provide fuel for naturally occurring algae and lead to a bloom. Some blooms can be harmless and provide food for tiny animals. or even kill them. Some harmful algal blooms, including Florida red tide, can be harmful to people. Reducing the amou nt of nutrients washing into the sea from land whether Florida red tides are involved or not would help improve some aspects of water quality. Potential Prevention Strategy : Establish a state wide retail tax on fertilizer that would encourage a reduc tion in fertilizer use and raise funds to pay for continual monitoring of coastal water quality, and research to determine water quality improvements If no measurable improvements were found within three years, the law would be automatically repealed. 10. Th is program relies on funds raised from those that buy fertilizer. Would this program cost you more because you use plant fertilizers in Florida? Yes, I use fertilizers No 10.A. Would you vote for a 10 % ta x on all fertilizer sales to support this type of long run prevention program? Yes, I would vote for it No I would not vote for it OR 10.B. How sure are you of this decision? 10. C Is there any % you would vote for ? Not sure Yes, I would vote for _______% Somewhat sure No Very sure that you would vote for? _____% Red Tide Strategies

PAGE 113

102 MITIGATION: Background : Many factors determine when and where someone might visit a particular beach or coastal area in Florida, including unpredictable weather and environmental conditions. In the case of red tides, these conditions can vary within the hour and between nearby beaches. Providing accurate and timely information on a variety of coastal conditions would help the public decide when and where to visit. This service would maximize the enjoyment of the trip to the public and help stabilize tourism expenditures across t he state. I t would also mitigate, or Since this system includes a variety of beach related information the benefits of this short run mitigation strategy will accr ue regardless of red tide conditions. Potential Mitigation Strategy : Establish a Beach Conditions Reporting Service Trust Fund to support the training of observers, initial equipment expenditures and maintenance of an electronic reporting system. It is a nticipated that one time donations to this fund would be sufficient to establish and support this program over the next three years. Only people who donate would able to access the system. 11. Are you aware of the Beach Conditions Reporting System for the Gulf Coast of Florida TM ? Yes, I know of this system No 1 1 A. Would you pay a one time donation of $ 2 5 into this trust fund for access to information provided by this service for the next three years? Yes, I would donate No I would not donate OR 1 1 B. How sure are you of this decision? 11C. Is there any amount you would pay? Not sure Yes, I would pay $ _______ Somewhat sure No Very sure that you would pay ? $ _____ 12. Approximately how many days do you spend at the beach each year? _____ days

PAGE 114

103 CONTROL: Background : We know that red tides are caused by rapid growth in one species of algae. This growth can be stopped using biological or chemical controls. A biological control would involve releasing a predator species. A chemical control would involve spreading a nat ural material on the bloom to inactivate the toxin or remove it from surface waters. Both types of strategies have been used to successfully treat blooms throughout the world, but not in conditions identical to Florida; thus, pilot testing would be require d to determine any ecological impacts. and is similar in nature to mosquito control efforts. Benefits include protecting beachside and marine recreation and coast al tourism. Potential Control Strategy : Establish a 3 year tax on the assessed value of all private property to fund red tide control programs, including pilot testing. If no measurable improvements were found within three years, the law would be automati cally repealed. 13. Did you pay property taxes in Florida last year? Yes, I paid this tax No 13A. Would you vote for a 3 year ad valorem fee o f $ 10 per $100,000 of the assessed value of all taxable property in your county to fund a local Red Tide Control program? Yes, I would vote for it No I would not vote for it OR 1 3 B. How sure are you of this decision? 1 3C Is there any % you would vote for ? Not sure Yes, I would vote for _______% Somewhat sure No Very sure that you would vote for ? _____% 14. In general, which type of control most appeals to you? Please select one: Biological Chemical Neither

PAGE 115

104 15. Of the three types of red tide programs you just evaluated, which would you prefer if the S tate of Florida only had funds for one? Please check one. Prevention: 3 year fertilizer tax for water quality monitoring and enforcement Control: 3 year tax on assessed home values to fund red tide control programs Mitigation: D onation to fund beach conditions reporting program for 3 years None of them 16. How long have you resided in Florida? ____ years 17. How many months of the year do you reside in Florida? ____ months 18. What is the ZIP code of your residence in Florida? _ 19. How many miles by car do you live from the coast? ___ miles 20. In what year were you born? 1 9 21. What is the highest level of education tha t you have completed? Some elementary or high school Some college High school graduate/GED College graduate Technical/Vocational Graduate/Professional degree 22. Which of the following describe your race or ethnicity? Please mark all that apply. White/Caucasian Native Hawaiian/Pacific Islander African American/Black American Indian/Alaskan Native Asian Hispanic/Latino Other 23. Which category includes your household's annual income before taxes? Less than $30,000 $90,001 to $120,000 $30,000 to $60,000 $120,001 to $150,000 $60,001 to $90,000 More than $150,000 Thank you. Please mail in the postage paid envelope today Demographics

PAGE 116

105 APPENDIX B REVISED STATED PREFE RENCE QUESTIONS FOR ONLINE QUESTIONNAIRE

PAGE 117

106 We will describe three different types of red tide programs: prevention mitigation and control ask you which you would prefer the most (if any). Please sel ect one of the three terms that includes the month you were born. You will be randomly choosing the order you will evaluate these three programs, which could affect your evaluation. January April May August September December PREVENTION: Background: We know that plants including algae need nutrients to grow and that when excess nutrients drain from the land to the sea, these waters can from land can be naturally occurring or can come from the human use of f ertilizers for lawns and crops. We know that sometimes this enrichment can provide fuel for naturally occurring algae and lead to a bloom. Some blooms can be harmless and provide food for tiny animals. have toxins and can make animals sick or even kill them. Some harmful algal blooms, including Florida red tide, can be harmful to people. Reducing the amount of nutrients washing into the sea from land whether Florida red tides are involved or not woul d help improve some aspects of water quality. Potential Prevention Strategy : Establish a state wide retail tax on fertilizer that would encourage a reduc tion in fertilizer use and raise funds to pay for continual monitoring of coastal water quality, and research to determine water quality improvements If no measurable improvements were found within three years, the law would be automatically repealed. 12. This program relies on funds raised from those that buy fertilizer. Would this program cost you more because you use plant fertilizers in Florida? Yes, I use fertilizers No Red Tide Strategies

PAGE 118

107 10.A. What is the highest fertilizer sales ta x that you would vote for to support this type of long run prevention program? 0% I do not support this 1% 5% 10% More than 10% Please select the value that best represents the most you would be willing to vote for even. MITIGATION: Background : Many factors determine when and where someone might visit a particular beach or coastal area in Florida, including unpredictable weather and environmental conditions. In the case of red tides, these conditions can vary within the hour and between nearby beaches. Providing accurate and timely information on a variety of coastal conditions would help the public decide when and where to visit. This service would maximize the enjoyment of the trip to the public and help stabilize tourism expenditures across t he state. I t would also mitigate, or Since this system includes a variety of beach related information the benefits of this short run mitigation strategy will accr ue regardless of red tide conditions. Potential Mitigation Strategy : Establish a Beach Conditions Reporting Service Trust Fund to support the training of observers, initial equipment expenditures and maintenance of an electronic reporting system. It is a nticipated that one time donations to this fund would be sufficient to establish and support this program over the next three years. Only people who donate would able to access the system. 24. Are you aware of the Beach Conditions Reporting System for the Gulf Coast of Florida TM ? Yes, I know of this system No 1 1 A. What is the highest one time donation that you would be willing to make to this trust fund for access to information provided by this service for the next three years? $0 I do not support this $5 $15 $25 More than $25 11.B. Approximately how many days do you spend at the beach each year? Please count partial days as full days. Pull down [none, 1 7, 8 14, 15 21, more than 21].

PAGE 119

108 CONTROL: Background : We know that red tides are caused by rapid growth in one species of algae. This growth can be stopped using biological or chemical controls. A biological control would involve releasing a predator species. A chemical control would involve spreading a natural material on the bloom to inactivate the toxin or remove it from surface waters. Both types of stra tegies have been used to successfully treat blooms throughout the world, but not in conditions identical to Florida; thus, pilot testing would be required to determine any ecological impacts. rrence within a specific area and is similar in nature to mosquito control efforts. Benefits include protecting beachside and marine recreation and coastal tourism. Potential Control Strategy : Establish a 3 year tax on the assessed value of all private p roperty to fund red tide control programs, including pilot testing. If no measurable improvements were found within three years, the law would be automatically repealed. 25. Did you pay property taxes in Florida last year? Yes, I paid this tax No I 12.A. If a 3 year ad valorem fee were on the next ballot, what is the highest fee that you vote for? The fee would apply to each $100,000 of the assessed value of all taxable property in your county and it would be used to fund a local Red Tide Control program in your region only. $0 I do not support this $5 $10 $15 More than $15 12.B. In general, which type of control most appeals to you? Biological Chemical Neither 26. Of the three types of red tide programs you just evaluated, which would you prefer if the S tate of Florida only had funds for one? Prevention: 3 year fertilizer tax for water quality monitoring and enforcement Control: 3 year tax on assessed home values to fund red tide control programs Mitigation: D onation to fund beach conditions reporting program for 3 years None of them

PAGE 120

109 APPENDIX C OPEN ENDED COMMENTS FROM ONLINE RESPONDENTS ( N = 47) Our s tate which is already too dependent on t ourism cannot afford any ecological disasters be it red tide or oil drilling off shore I am a recreational fisherman and sea turtle patroller and very concerned about the effects of red tide on my community I am concerned that few people know much about red ti de, even in our area where it severely impacts our economy and quality of life. L ike the building code, establish types of ferti li zer that can be sold in counties on the water. NO NEW TAX. Go back to areas of critical concern and let everyone keep their storm water runoff on their own property. None of the funding approaches that are mentioned are levied upon the tourists who frequently visit our area and occupy the resorts located on the beach. It is this group that tends to take advantage of our beaches Add a 1 to 2% visitors or sales tax since they and the hotels rely very heavily on the water and beaches health and welfare. I don't know if fertilizer control is in the category of chemical or biological. I answered that I would prefer biological, if there should be some organism that prevents the algae from making the toxin. However, I also support fertilizer control. I am not so sure about the clay proposals, seems as if it might be a worse problem than the red tide. Regarding the proposed "pay to g et warnings" system, for it to help me it would need to be very current. I go to the beach in the morning, before 10 am in warm weather. And conditions can change from day to day. I have looked up some info on the internet in the past and found that it app lied to different beaches than I go to or was out dated. Keep up the good job. Too much emphasis on taxes and donations (which would come mostly from people living near the water). Our universities should be funded (grants sponsored by commercial and volun teer sources) to study and perform some of the suggested approaches named above. I don't know much about what causes red tide, but I do know that when Lake Ochocobee (not spelled right) is not opened and let to drain into the river Ft Myers beach and all the waters going to Ft Myers beach are much cleaner.

PAGE 121

110 I DON'T BELIEVE THAT RED TIDES WILL BE REDUCED UNTIL THE FERTILIZER COMING D OWN THE MISSISSIPPI IS REDUCED. THIS I BELIEVE IS THE PRIME SOURCE FOR RED TIDE IN THE GULF. R ed tide is nothing new to F lorid a waters, just better reported. it has be e n reported in the old logs of sailing ships I would consider contributing time. I think prevention is key and that people need to be educated about what causes this problem and that g o lf courses add to the proble m with the amount of water they consume and the fertilizer used. However, I also feel that as a home owner on the river I need to also curb my water and fertilizer use just as much. No new taxes! Forget controlling or eliminating red tide. It is natural, a nd there will be unintended consequences. Red tides have been around l ong before man or fertilizers. When I was a small kid growing up on the island we had awful red tides. There was almost no population then. The fish always come back. The red tide always subsides. The tourists always come back. Gover nment needs to stay out of it. We don't need more taxes or meddling wi th nature! I learned that you want my money to do your study of red tides and you should get your money from a private grant. I don't want to p ay any more taxes for anything. Human health consequences should be investigated thoroughly. I believe that immune compromised people are at extremely high risk for complications and even death during red tide episodes. I know of 2 cases of death in th e 2005 06 red tide episode that were immune compromised indiv iduals that lived on the coast. Let heaven and nature sing Stop letting companies dump their waste chemicals into Tampa Bay. I feel we fertilize too much, but towns and states are responsible. T hey fertilize freely with no thought of the harm they are doing. First we need to develop an earth friendly fertilizer. Second, a place in Ocala has developed a fire ant sterilizer that does not disturb the eco system. I believe that if we eradicated the use of over the counter fire ant control, this would be of benefit in helping with r ed tide. Stop opening the locks at Lake Okechobee.

PAGE 122

111 Red tide is a matter that needs to be paid close attention. In general I am confident additional alternate sol utions will be found. Meanwhile, people n eed to be made aware of its importance and impact. I appreciate your efforts to develop knowledge. Most people are probably most concerned about the things which most affect them. For me red tide is an afterthought. Life is so hectic red tide issues aren't much more than a blip on my radar of life. Not to say the research is unimportant because there is a lot of important information being researched every day that I am unaware of which greatly affects my life. I have wondered ab out the effect of mining debris e.g. francolite, on the severity of red tide. I wrote a letter to the Herald Tribune that they never published. Does phosphate detritus play a role in red ti d e? One unbiased organization should consolidate the research, i nformation and issues for public support communications. The most extensive and easily accessible literature appears to be coming from non objective purveyors. God created red tide as a part of the system, it is a natural way of controlling other things an d fish. As far as I know restricting fert ilizer in the summer is a huge mistake and prevents nothing other than unhealthy plants and lawns, so leave it alone! Don't mess with it ; it will only cause other problems that may be a greater burden than a little inconve nience like going to the beach. B e careful thinking taxing will do the job... those of us who still work for a living can't continue to pay taxes for the rest who aren't While it seems most probable that fertilizer runoff is a major contributor the practices employed by the Army Corps and SFWMD may have greater impact on our coastal waters. I am of the opinion that mismanagement of C 43 has done tremendous damage. J ust a thank you to whomever is working diligently to control the problems associated with red tide, including the offering of the survey for completion/education. ; if this a political tactic, we'll fall right into the other worthless surveys made the last twelve months. A HEALTH AND ECOOMIC PROBLEM. Because of my asthma I cannot leave my house or open the windows during a red tide or red drift algae bloom. It is very important to our family to have a solution to this problem. It is definitely a drawback to the tourists to have to walk around dead and rotting fish on the beach.

PAGE 123

112 I think you missed the most important cause of red tide on the west coast. Water releases out of Lake OKEECHOBEE When red tide is in the area we cannot go outside and enjoy our beach because of difficulty breathing. We have neighbors who have moved and tourists who have not returned because of the frequency of it in the past few years. The dead fish on the beach that are not cared for b y the county make the beach experience unacceptable for those who come to enjoy Florida as well as we who live here. Something has to be done to solve this problem. The million dollar condos that the developers have forced on us are going to remain empty i f a solution is not found shortly. D ue to the increase in red tide blooms since 2004 we need to take a more active role in control of the k. brevis algae. Through satellite monitoring I believe we can notify the public of potential outbreaks in red tide u sing the Sea WiFS sensor aboard the orbitbiew 2 satellite. M y suggestion is to initiate another red tide conference with the University of Florida, Rosenstiel School at the University of Miami and Mote Marine Laboratory to be located in the Lee County Area shortly and to initiate potential solutions to the Red Tide problem. Chuck Weisinger : weisguy@weisbuy.com J ust want scientific data not may be from all the do gooders R ed tide blooms were reported by ship's logs in the 18 hundreds. This is not something new ; it is just bett er observed and reported today! I think that much of our tax money is wasted on foolish things. Think Charlotte County and the wasted money on Murdock. I do not support new taxes... start using the $$$ we are paying now to pay for prog rams like stopping red tide.

PAGE 124

113 LIST OF REFERENCES Arrow K., Solow, R., Portney, P., Leamer, E., Radner, R. and Schuman, H., 1993. Repo rt of the NOAA panel on contingent valuation. Federal Register 58 : 4601 4614. Backer, L.C. 2009. Harmful Algae 8(4): 618 622 Carson, Richard T., N.E. Flores and N.F. Meade. Environmental and Resource Economics 19(2): 173 210. Dumas, C.F., R.T. Burrus, Jr., E Diener, J. Payne, and J. Rose. Journal of Park and Recreation Administration 25(2): 29 41. Fleming, L E., Bean, J A., Kirkpatrick, B Yung S C Pierce, R. Naar, J ., Nierenberg, K ., Backer, L .C., Wanner, A ., Reich, A ., Yue Z ., Watkins, S ., Henry, M., Zaias, J., Abraham, W M. Benson, J ., Cassedy, A ., Hollenbeck, J ., Kirkpatrick, G ., Clarke, T 2009. "Exposure and Effect Assessment of Aerosolized Red Tide Toxins (Brevetoxins) and Asthma." Environmental Health Perspectives 117 ( 7): 1095 1100. Genio, E.L. Jr., R.M. Rejesus, R.S. Pomeroy, A. initiated mangrove protection activities. Oceans and Coastal Management 50(10): 808 828. Habas E.J. and C.K. Gilbert 1974. The economic effects of the 1971 Florida red tide and the damage it presages for future occurrences Environmental Letters 6(2) : 139 147 Hoagland, P. and S. Scatasta, 2006: The economic effects of harmful algal blooms. In Ecology of Harmful Algae [E. Gr aneli and J. Turner (eds.)] Springer Verlag, Dordrecht, The Netherlands. Chap. 29. Ocean & Coastal Management 51: 4 2 0 429. Kirkpatrick, B. L.E. Fleming, L.C. Backer, J.A. Bean, R. Tamer, G. Kirkpatrick, T. Kane, A. Wanner, D. Dalpra, A. Reich and D. Baden Environmental exposures to Florida red tides: effects on emergency room respiratory diagnoses admissions H armful Algae : 526 533 Krinsky, I. and A.L. Robb. 1986. On Approximating the Statisti Review of Economics and Statistics 68:715 719. Larkin, S., and C. Adams Society and Natural Resources 20(9): 849 859.

PAGE 125

114 Larkin, S.L. and C.M. Adams. 2008. "Public Awareness and Knowledge of Red Tide Blooms." Journal of Extension 46(2): Article number 2COM2, pp. 12. Li Chuan Journal of Environmental Economics and Management 28(2): 256 269. Loomis, J.B., and A. Gonzalez to pay function for protecting acres Ecological Economics 25: 315 322. Morgan, K., S.L. Larkin, and C.M. Adams. 2010 "Red Tides and Participation in Marine based Activities: Estimating the Re sponse of Southwest Florida Residents." Harmful Algae 9(3): 333 341. Morgan, K.L., S.L. Larkin, and C.M. Adams. 2009. "Firm level economic effects of HABS: A tool for business loss assessment." Harmful Algae 8 (2) :212 218. Morgan, K.L., S.L. Larkin, and C.M Electronic Data Information Source (EDIS) FE711. Food and Resource Economics Department, University of Florida, Gainesville, FL Available at http://edis.ifas.ufl.edu/FE711 Land Economics 82(2): 174 188. Turnbull, B.W. 1976. The Empirical Distribution Function with Arbit rarily Grouped, Censored and Truncated Data. Journal of the Royal Statistics Society (38) : 290 295. contingent valuation: an open space survey and referendum in Corv Journal of Economic Behavior and Organization 51: 261 277.

PAGE 126

115 BIOGRAPHICAL SKETCH Kristen Lucas was born in Hollywood, Florida. She obtain ed her undergraduate degree in economics with a minor in business a dministration in 2007 at the Univer sity of Florida. After taking courses in environmental economics, she decided that she would like to focus her education on environmental and natural resource economics. She continued into her graduate education at the University of Florida in the Food and Resource Economics Department where she subsequently obtained her Master of Science.