Making a convincing case for ecosystem services ultimately depends on our ability to develop reliable estimates of the actual financial value of these services- a task that is challenged by several methodological hurdles. The contingent valuation (CV) method for non-market valuation of ecosystem services typically involves presenting the respondent with a choice to pay for a program intended to improve future outcomes, such as a program to place parcels into conservation easement, or a program to manage an invasive species. However, deducing the value of the good (or bad) at the core of the program—the welfare gain generated by a parcel of conserved land, for instance, or the loss incurred by a species invasion—often is not possible because respondent preferences are conflated with their expectations about future environmental outcomes in the absence of the program. For example, a respondent may believe that a parcel of land is highly unlikely to be developed, and therefore may be unwilling to pay for a conservation easement program despite the high value they place on the parcel’s open space character.
Provencher et al. (in review) demonstrate mathematically the conflation of expectations and preferences, and argue that, it is unlikely to generate estimates that are robust to the specification of the willingness to pay (WTP) function. They show that additional survey data that explicitly elicits respondents’ expectations can resolve this problem. These formal arguments were applied to the case of preventing invasion of Eurasian Watermilfoil (Myriophyllum spicatum), an invasive aquatic plant that is present in many lakes in the northern U.S. and Canada and a possible threat to many more. The valuation exercise was applied to households with shoreline property on lakes without milfoil, and home owners were asked questions about a referendum on a program to prevent a milfoil invasion. Data for the analysis were generated from a web and mail survey of a sample of lakeshore property owners in the Northern Highlands Lake District. Sample property owners were initially surveyed in the summer of 2005, and then again in the early fall 2008. Overall, 2955 households were contacted in the 2008 survey, with 1565 (53%) providing usable responses.
The annual welfare loss from an invasion was determined to be approximately $1800 per shoreline property. Importantly, the results illustrate the consequence of conflating expectations and preferences, as estimates of the unadjusted annual willingness to pay for a program to prevent milfoil invasions was approximately $570, significantly less than the estimated annual welfare loss of an actual invasion. The disentangling methodology is supported by results of a different CV question asked of households on lakes that already have milfoil. The willingness to pay of these respondents, queried as to how they would vote on a program to control the existing milfoil on their lake, was much greater than willingness to pay for a program to prevent milfoil, but similar to the estimated loss from an actual invasion, which accounts for expectations.
In addition to eliciting explicit values for respondent expectations about the future of the environmental good in the absence of an intervention, the survey of lakeshore property owners also asked respondents to report the probability that they would vote “yes” on the referendum were it to actually occur. This contrasts with typical binary yes/no options, or less common likert-scale type response options or follow-up uncertainty questions that require strong assumptions necessary to map imprecise expressions of uncertainty into probability space. Provencher et al. develop an econometric model that explicitly incorporates respondent probabilities in the estimation of the willingness to pay function.
The validity of the stated preference CV approach to estimate the cost of a milfoil invasion was assessed by comparing it to results from a revealed preference hedonic price analysis. Horsch and Lewis (2009) used market sales data for shoreline properties in Vilas County, and accounted for the possibility that unobservables correlated with a lake’s vulnerability to a milfoil invasion might also affect the market value of its shoreline properties. The analysis examined over 1800 shoreline property sales on 172 lakes in Vilas County over the 10-year horizon 1997-2006, using two models. Table 1 presents results of the comparison. The presumption in the table is that the present value of milfoil loss is capitalized in shoreline property values. The per-property reduction in property value due to a milfoil invasion is estimated at $28,294 to $32,087, depending on model specification. Both of these values are close to the CV estimates of a welfare loss (net present value) of $23,614 to $30,550, revealing strong convergence between these two distinct approaches.
Comparison of Estimated Present Value of Welfare Loss from a Milfoil Invasion
Present Value ($)
Confidence Interval ($)
Milfoil Prevention Contingent Valuation:
With income (estimated annual loss=$1373)
Without income (estimated annual loss=$2106)
Milfoil Control Contingent Valuation:
With income (estimated annual loss=$1521):
Without income (estimated annual loss=$1226):
Horsch and Lewis Hedonic Study (2009):
Linear fixed effects
Nonlinear fixed effects
Provencher, B., D.J. Lewis, and K. Anderson. In review. Disentangling preferences and expectations in stated preference analysis with respondent uncertainty: The case of invasive species prevention.