Final Report: Web-Based Methods for Valuing Wetland ServicesEPA Grant Number: R827922
Title: Web-Based Methods for Valuing Wetland Services
Investigators: Hoehn, John P. , Lupi, Frank , Kaplowitz, Michael D.
Institution: Michigan State University
EPA Project Officer: Chung, Serena
Project Period: October 1, 1999 through September 30, 2002 (Extended to November 30, 2003)
Project Amount: $227,758
RFA: Decision-Making and Valuation for Environmental Policy (1999) RFA Text | Recipients Lists
Research Category: Environmental Justice
Wetland ecosystems offer an important opportunity for developing ecosystem valuation methods. Wetlands are one of a small number of ecosystem types that are protected and managed under Federal and State regulations. The basic goal of Federal regulation is "no net loss" of wetlands (National Research Council, 2001). To avoid a net loss of wetland services, Federal and State regulations require mitigation to obtain permits for activities that may impair or destroy them. Mitigation raises the policy issue of determining what and how much should be done to offset the loss or impairment of a wetland.
The basic goal of the project was to develop a Web-based stated preference questionnaire and use it to examine the equivalency of drained and restored wetlands from the viewpoint of economics and human preferences. The specific objectives were to: (1) learn more about how the public values and perceives the attributes of freshwater wetland ecosystems; (2) evaluate the feasibility of a Web-based stated preference survey for valuing the attributes of a complex environmental and natural resource; (3) allow respondents to interact with salient information regarding the wetland characteristics and test whether this interaction affects estimated preferences; and (4) examine whether a Web-based experimental design that sequentially adjusts attribute levels during the survey improves the accuracy of estimated preferences for wetland characteristics.
The project found that it was feasible for individuals to make reasoned choices across wetland ecosystems using a Web-based questionnaire. However, the two questionnaires that were fielded in a large sample survey of the general public did not perform equally well. The primary questionnaire derived from a multiple method, iterative design process by the researchers proved much less prone to characteristic biases common in complex decisions, specifically the underweighting of gains and the overweighting of losses. The results are summarized below by the specific project objectives.
Baseline Knowledge about Wetland Ecosystems
Baseline knowledge focus groups were used to understand how ordinary citizens use, understand, and perceive wetlands and wetland services. The results showed that participants had direct experience with wetland habitats and were familiar with many types of plants and animals found in such habitats. The discussions identified gaps and misperceptions in participants' knowledge that might be filled in and corrected by questionnaire narrative. The discussions also showed that it was easy to misinform respondents by using certain terms or an unexplained photograph. These results provided a basis for developing draft questionnaires, while also underscoring the importance of a presurvey process for testing and evaluating questionnaire performance.
For the three focus groups to explore participants’ baseline knowledge about wetlands and wetland attributes, participants were drawn from the mid-Michigan region using randomly selected telephone numbers. Contacted adults were invited to participate in a discussion about important public issues on the Michigan State University campus. Focus group discussions were led by a trained moderator and followed specially prepared discussion guides. The script began with a discussion of environmental issues to determine whether wetland ecosystems might be a high priority concern. The script then nudged the discussion toward specific wetland topics, such as the kinds of activities that participants associated with wetlands. Participants were also asked to evaluate informational materials and photographs of wetland and nonwetland ecosystems. These sessions concluded with discussion of mitigation policy scenarios, such as how wetland restoration might make up for wetlands covered by highway construction.
Participants were generally knowledgeable about wetlands in terms of wildlife and plants. Many participants identified wetlands as habitats for birds, fish, beavers, muskrats, and mosquitoes. Some participants also discussed plants associated with wetlands. Additionally, focus group participants raised various wetland functions including flood prevention, water purification, and habitat for wildlife during their discussions. Respondents talked most easily and from personal experience about wetlands as habitats for plants and animals. Respondents offered specific examples of wildlife depending on wetlands, as well as simple remarks on the importance of wetlands as places for animals and plants. Many participants had some knowledge of other ecological functions of wetlands, but almost all respondents indicated an awareness of wetlands as habitats for desirable plants and animals.
The discussions also revealed participants' misperceptions about wetlands. Common misperceptions were "trees don't grow in wetlands" and "wetlands kill trees." These particular misperceptions are especially interesting because wooded wetlands make up more than two-thirds of Michigan's wetlands. The misperception that wetlands kill trees may seem justified to a casual observer because recent episodes of Dutch elm disease have left many dead trees in Michigan wetlands. Indeed, a number of participants were aware of the disease problem and mentioned it when reviewing wetland photographs.
Feedback from participants also indicated that some photographs had a potential to mislead respondents. While viewing panoramic photographs, dark areas were occasionally interpreted as burned areas, when in fact the dark areas were merely shadows. In close-up images of nonwetland plants, some respondents interpreted backgrounds as wetlands when, in fact, the backgrounds were out-of-focus nonwetland plants.
Participants were also sensitive to terms used in the mitigation scenarios. Participants had strong negative reactions to the idea that wetlands might be "created" as mitigation for destroyed wetlands. The latter term apparently evoked mental images of wetland soils and animals being "created" by human effort, images that simply lacked credibility for participants. A change of terms from "created" to "restored" obviated the difficulty. Participants had few objections to the idea that mitigation might "restore" a previously drained wetland.
Feasibility of Web-Based Questionnaires
Draft questionnaires were developed using the baseline information from the three initial focus groups, a subsequent meeting with a scientific advisory panel, and previous research. The draft questionnaires were then tested and evaluated using an iterative process of both focus group and individual interviews. The feedback from the evaluation focus groups and interviews allowed problems to be detected in the draft questionnaires and solutions to be tested. The process of developing the questionnaire was iterative in the sense that revisions were made based on previous focus groups and interviews, and then the revisions were tested again in subsequent focus groups or individual interviews.
The final Web-based questionnaires were administered in a large scale sample beginning in October and ending in December 2003. Respondents were invited to participate in the stated choice experiments using an e-mail invitation. The invitation described the experiment as a Michigan Citizens' Panel and provided a hyperlink to the questionnaire. Respondents each clicked on an individually encoded hyperlink to view the welcome page of the questionnaire. E-mail invitations sent to a commercially maintained Michigan panel of respondents resulted in 3,454 clicks on the welcome page of the Web-based questionnaire.
From the welcome page, 2,689 respondents began the first page of the questionnaire and completed one or more questions. A large number of individuals chose to leave one or more questions blank during the course of completing the questionnaire. The most common unanswered questions were the mitigation choice questions and the question regarding household income. Fully complete questionnaires numbered 1,060, or 31 percent of those who clicked on the welcome page. Usable questionnaires with at least one completed mitigation choice and complete demographic information numbered 1,373, or 40 percent of those visiting the welcome page. The total number of usable mitigation choices accompanied by usable demographic information was 6,714.
Information and Stated Preferences
Two different questionnaires were used in the large scale survey: a questionnaire using a tabular format for wetland information, and a questionnaire using a text format. The tabular format was designed to facilitate informed wetland mitigation decisions by ordinary respondents drawn from the general public. The tabular format placed the relevant wetland choice information in two adjacent columns, one for each wetland under consideration. Wetland habitats were described in four dimensions: (1) habitats for reptiles and amphibians; (2) habitats for song birds; (3) habitats for wading birds; and (4) habitats for wild flowers.
Each type of habitat was described with a rating of poor, good, or excellent. The questionnaire explained each of the ratings, which were based on what a visitor was likely to see during a visit to a wetland. A "poor" rating meant that the wetland habitat supported "these species in very small numbers...[so] a trained observer is unlikely to find any of these species." A "good" rating meant that the wetland habitat supported "these species in average numbers...[so] a casual observer is likely to see a few of these species." An "excellent" rating meant that the wetland habitat supported "these species in better than average numbers...[so] a casual observer is very likely to see a variety of these species."
Respondents found the tabular format easy to understand. The tabular format permitted rapid assimilation of the wetland features, encouraged feature-to-feature comparisons, and facilitated tradeoffs across different features. Pretest participants sometimes commented that the decisions were too easy.
The text format questionnaire replaced the tabular wetland information with two paragraphs of text, thereby creating a text format for the choice questions. The text format contained the same information as the tabular format and conformed with the rules for traditional willingness to pay scenario descriptions (Mitchell and Carson, 1989). However, the text format did not contain the structural and graphical devices of the tabular format for illustrating: (1) the five different dimensions of habitat quality; (2) the quality ratings of each wetland and each habitat dimension; and (3) tradeoffs across the two wetlands and their habitat qualities.
Previous research suggests that the structural differences between the two choice elicitation formats should lead to differences in the decisions made by respondents. Viscusi and Magat (1987) found that tabular formats had a greater impact on risk avoidance behavior and willingness to pay than text formats. Psychological research stresses that cognitive constraints lead to characteristic biases when dealing with complicated decisions (Kahneman, 2003). One characteristic bias from cognitive complexity is that respondents overweight losses and underweight gains when faced with information that exceeds their ability to understand and assimilate (McFadden, 2000). The literature provided a basis for a quantitative and testable hypothesis about the performance of the tabular and text format questionnaires. If the tabular format was successful in encouraging reasoned, balanced decisions, respondents should weight wetland gains and losses in a more balanced way relative to those respondents who used the text version of the questionnaire.
The stated choice data collected in the Web-based survey was used to estimate choice equations using both the tabular and text data. The choice equations were used to derive mitigation equivalency functions. These equivalency functions, in turn, provided the amount of restored wetland acreage that would compensate the median respondent for the loss of a drained wetland of a given size. The independent variables in the equivalency functions were drained wetland acreage, degree of public access to the restored wetland, type of wetland, the changes in habitat qualities of the restored wetland relative to the drained wetland, and the demographic characteristics of respondents.
Mean levels of income, education, age, and gender were similar for respondents to both the tabular and text questionnaires. One exception was the age of respondents; the text data set contained about 8 percent more respondents over 65 years of age. The mean income level for respondents to both versions was about the same as the 2002 Census mean for the State of Michigan. Respondents to the questionnaires were somewhat more schooled with some college study and were more likely to be female and over age 65.
The choice data were used to estimate two mitigation equivalency functions, which indicate the amount of restored wetland that compensates for a drained wetland, one using the tabular data and one using the text data. The coefficients for the tabular equation were almost all of the anticipated signs and statistically different from zero at the 95-percent level. Interestingly, the tabular equation suggested that restored acreage was worth less than acreage of drained wetland, ceteris paribus. The loss of an acre of the drained wetland was compensated by 1.4 acres of restored wetland, so the loss of a 20-acre wetland would require 28 acres of restored wetland, given no change in other wetland features.
When wetland features are different in the restored wetland, the mitigation equivalency equation increases or decreases the restored wetland acreage that is needed to compensate for the loss of the drained wetland. If the restored wetland has poorer quality features than the drained wetland, the mitigation equation requires more restored acreage to compensate for the loss. If the restored wetland has higher quality features, less restoration is required as compensation. For example, 6 additional acres of restored wetland was required to make up for the loss of public access to the restored wetland. A change in wetland type required 5 additional acres in restoration as compensation.
Changes in habitat quality for amphibians and reptiles, wading birds, song birds, and wild flowers also were associated with differences in the amount of required compensatory restoration. Reduction in habitat qualities from good to poor all required additional acreage to offset the loss in quality. Improvements in habitat quality relative to the drained wetland reduced the amount of restored acreage required for compensatory mitigation.
A mitigation equivalency function was also estimated for the data obtained with the text format questionnaire. The most noticeable feature of the text data mitigation equation was the relatively large size of the coefficients representing a loss of habitat quality in the restored wetland. Respondents who were randomly assigned to the text questionnaire required as much as four times more acreage compensation for losses than respondents who were randomly assigned to the tabular questionnaire. For improvements in habitat quality, the relationship between the two equivalency functions was just the reverse. Respondents to the text questionnaire underweighted gains in habitat quality relative to respondents who received the tabular questionnaire. Indeed, the text-equivalency function coefficients were not statistically different from zero for improvements in habitat quality from good to excellent in the text questionnaire.
Among other things, the results indicate that the text respondents were susceptible to decision-making biases frequently noted by psychologists: relative to the tabular format, text respondents tended to overweight losses and underweight gains. The tabular elicitation format appears relatively resistant to the effects of such biases. The tabular coefficients were relatively precise in statistical terms, and the differences between coefficients in the tabular equivalency function appeared reasonable and consistent with intuition. The iterative design process appeared successful in deriving a questionnaire instrument that supported balanced, reasoned decisions for rather complex mitigation choices.
Adaptive Experimental Designs
One fundamental issue in stated-preference research is selection of attribute levels of the characteristics of the goods presented to respondents. Ultimately, this is a question of experimental design. An experimental design is: (1) the set of characteristics and their attribute levels; (2) the bundling of these characteristics and attribute levels into composite goods (in our case, wetlands); and (3) the goods (wetlands) combinations presented to the respondents for their ranking and choice.
The project implemented an extension of Kanninen's D-optimal experimental design with updating (Kanninen, 1998; Kanninen, 2002). D-optimal designs are experimental designs that maximize the precision of parameter estimates. Updating seeks to improve precision by iteratively adjusting the attribute levels contained in a given design. Our research extended Kanninen's approach in three novel ways. First, we implemented the approach in a Web-based format. A computer-based survey such as our Web survey proved particularly amenable to the updating of attribute levels. Second, we used a nonprice continuous variable, wetland acreage, as the balancing variable for the updating. Third, we used a model estimated on pilot data to choose the starting values for the continuous attribute (acreage).
Though the data are still being analyzed, several conclusions are evident from the initial analysis:
• It was feasible and straightforward to do the updating. The cost of updating was very low when done in a Web-based survey. The cost involved a small amount of programming and research time to review the trends in the response probabilities and adjust the attribute levels accordingly. In our case, this was easily accomplished between waves of the Web survey (the waves were done to avoid crashing our server, but served the dual purpose of providing a convenient opportunity to do the updating).
• The starting values were very good. The starting values were based on the model derived from the pilot data (Lupi, Kaplowitz, and Hoehn, 2002). Using the model-based starting values for each scenario's starting attribute values may have reduced the gains from updating, but clearly seemed worth it to get good starting values from the beginning.
• The final response probabilities were fairly close to the targets for all of the scenarios and in some cases were quite close. Thus, even with a good starting value for the acreage differences, the attribute levels could be improved upon by monitoring and updating them relative to a fixed design.
• Doing the updating required simplifying the design so that our qualitative habitat categories were reduced to two levels. We consider this a serious drawback. Published D-optimal design plans for more complex noncontinuous attributes would be beneficial.
• By creating fairly severe contrasts between the habitat levels, the updating approach necessitated, in several cases, a reduction of the acreage of the mitigated wetland so that it was smaller than the drained wetland. This was needed to offset the gains in habitat so that we could get the desired share of individuals voting for and against a scenario. This process increased the likelihood that some scenarios appeared counterfactual to some respondents.
• Experimental design alternatives were also examined in related research with wetland choice experiments implemented using mailed questionnaires. In these experiments, the fully randomized design was about 80 percent of the efficiency of the main effects design. It also enabled a straightforward estimation of all possible interaction effects without the need for any a priori selection of interaction effects to be included in the design. Further, the effect on the utility of moving from poor to good habitat was much larger than the effect on the utility of moving from good to excellent habitat. This apparently nonlinear effect of the qualitative habitat levels could not be identified if one were to strictly follow the Kanninen design approach, which involves selecting only the extreme levels of each attribute (Lupi, Kaplowitz, and Hoehn, 2003).
The project identified and applied a process for developing Web-based methods for valuing changes in ecosystem services using stated-choice methods. A Web-based questionnaire was developed using an iterative design process that began by assessing the public's baseline knowledge about wetland ecosystems. Questionnaire prototypes were then developed based on sound science but presented using language and concepts familiar to the general public. Questionnaire prototypes were tested and refined in successive stages using focus groups and individual interviews. The final choice scenarios were based on the mitigation problem typically faced by Federal and State authorities: determining the type and amount of mitigation that offsets the loss of a destroyed or drained common wetland.
The examination of the general public's baseline knowledge showed that a significant portion of Michigan respondents are aware of wetland ecosystems and care about some subset of wetland functions and services. Many respondents had direct experience with wetlands, usually through some type of leisure activity such as hunting, camping, walking, or other form of recreation. Respondents seemed aware of the role of wetlands in providing habitats for small animals, reptiles, amphibians, birds, and wild flowers. Respondents saw a direct relationship between wetlands and the availability of habitat for valued species. Respondents seemed less informed about other possible wetland functions such as flood control and nutrient cycling, especially in terms of how the impairment or destruction of a particular wetland might alter these functions.
The final Web-based stated-choice questionnaire built on respondents' baseline knowledge, filled gaps in respondents' knowledge with scientific information, and provided information to offset respondents' potential misperceptions. The stated-choice section of the questionnaire was specifically designed to present complex choices in a way that supported reasoned and informed decisions. Descriptions for drained and restored wetlands were arrayed in a tabular format with a short list of key wetland characteristics listed by row and the levels of these characteristics arrayed in adjacent columns for the drained and restored wetlands. Empirical testing showed that the tabular format reduced or eliminated characteristic biases in complex nonmarket decisions.
The research also demonstrated that stated-choice experiments with complex ecosystems are feasible for the general public. Careful research on baseline knowledge and systematic pretesting appear essential for obtaining reasonable, unbiased stated-choice results. The tabular questionnaire format that resulted from a multiple-method iterative design procedure performed well. The research also used a simple text-based choice question instrument, typical of some contingent valuation studies, to test for associated asymmetric biases based on recent psychological and economic research (Kahneman, 2003; McFadden, 2000). Analysis of the data from the two formats revealed that losses in ecosystem quality were overweighted and gains in quality were underweighted by respondents using the text-based choice instrument relative to results estimated from the data of respondents using the tabular format. Thus, while ecosystem choices may be complex enough to strain respondents' decision-making capacities, systematic questionnaire development can arrive at formats that reduce or eliminate the impact of characteristic biases on estimated values.
The results demonstrate that wetland qualities and services are valued by members of the general public. Wetland habitats for small animals, birds, and special plants were of interest and value to respondents. Respondents had direct experience with the latter types of wetland habitats and saw them as directly impacted by mitigation activities. The importance of habitat quality emerged consistently at all stages of the research including initial focus groups, pretest survey data analysis, mail survey results, and Web survey results. This finding is similar to other recent research on wetland ecosystems (Azevedo, Herriges, and King, 2000; Stevens, Benin, and Larson, 1995; Swallow, et al., 1998).
Two aspects of the research need to be kept in mind when interpreting the results. First, respondents to both the qualitative and quantitative research were drawn from residents of Michigan. Michigan’s climate is characteristic of the humid north-central portion of the United States. Wetlands are a common landscape feature, so Michigan residents may have more experience with wetlands than those in other parts of the United States, especially those living in arid regions. Second, while the study provides estimates of how to adjust mitigation ratios to account for differences in habitat quality, it should be considered a first step. The wetlands considered here were common types, which are regularly subject to permit actions in Michigan. The study results do not apply to wetlands, habitats, or rare species. Likewise, in the wetland choices studied here, respondents were explicitly asked to hold other functions of wetlands constant. Future research may wish to address the effect that wetland functions, beyond habitat, have on mitigation ratios that will provide in-kind compensation for the loss of an existing wetland.
Azevedo C, Herriges JA, King CL. Iowa wetlands: perceptions and values. Presented at the Center for Agricultural and Rural Development, Iowa State University, Ames, IA, March 2000.
Kahneman D. Maps of bounded rationality: psychology for behavioral economics. American Economic Review 2003;93(5):1449-1475.
Kanninen BJ. Optimal experimental design for binary choice experiments. Presented at the Hubert H. Humphrey Institute of Public Affairs, University of Minnesota, June 1998.
Kanninen BJ. Optimal designs for multinomial choice experiments. Journal of Marketing Research 2002;39(2):214-227.
McFadden D. Rationality for economists? Journal of Risk and Uncertainty 1999;19(1):73-105.
Mitchell RC, Carson RT. Using Surveys To Value Public Goods: The Contingent Valuation Method. Washington, DC: Resources for the Future Press, 1989.
National Research Council. Compensating for Wetland Losses Under the Clean Water Act. Washington, DC: National Academy Press, 2001.
Stevens TH, Benin S, Larson JS. Public attitudes and economic values for wetland preservation in New England. Wetlands 1995;15(3):226-231.
Swallow S, Spencer M, Miller C, Paton P, Deegen R, Whinstanley L, Shogren J. Methods and applications for ecosystem valuation: A collage. In: Kanninen B, ed. Proceedings of the First Workshop in the Environmental Policy and Economics Workshop Series, Washington, DC, October 28, 1998, pp. 17-19.
Journal Articles on this Report : 3 Displayed | Download in RIS Format
|Other project views:||All 27 publications||5 publications in selected types||All 3 journal articles|
||Hoehn JP, Lupi F, Kaplowitz MD. Untying a Lancastrian bundle: valuing ecosystems and ecosystem services for wetland mitigation. Journal of Environmental Management 2003;68(3):263-272.||
||Kaplowitz MD, Hoehn JP. Do focus groups and individual interviews reveal the same information for natural resource valuation? Ecological Economics 2001;36(2):237-247.||
||Lupi F, Kaplowitz MD, Hoehn JP. The economic equivalency of drained and restored wetlands in Michigan. American Journal of Agricultural Economics 2002;84(5):1355-1361.||