How Choice Experiment Design Affects Estimated Valuation of Use of Gestation Crates

This study utilises a split‐sample experimental approach where six different choice experiment designs that vary in the number of attributes are used to control for the effects of design dimensions and consumer inferences in assessing consumer demand. Our application is in the use of gestation stalls (crates) in pork production. Our findings suggest that effects of both the mixture of attributes included in the design and consumer inferences on the specific attributes presented in experiments influence estimated willingness‐to‐pay and consumer welfare. Implications are that when researchers limit their studies to the use of a unique choice experiment design, resulting willingness‐to‐pay estimates might not be robust to different specifications.     

Willingness to Pay for Rural Landscape Improvements: Combining Mixed Logit and Random‐Effects Models

This paper reports the findings from a discrete‐choice experiment designed to estimate the economic benefits associated with rural landscape improvements in Ireland. Using a mixed logit model, the panel nature of the dataset is exploited to retrieve willingness‐to‐pay values for every individual in the sample. This departs from customary approaches in which the willingness‐to‐pay estimates are normally expressed as measures of central tendency of an a priori distribution. Random‐effects models for panel data are subsequently used to identify the determinants of the individual‐specific willingness‐to‐pay estimates. In comparison with the standard methods used to incorporate individual‐specific variables into the analysis of discrete‐choice experiments, the analytical approach outlined in this paper is shown to add considerable explanatory power to the welfare estimates.     

The Impacts of Elicitation Context on Stated Preferences for Agricultural Landscapes

Statements of willingness to pay (WTP) have been shown to be dependent upon the framing of the hypothetical market. In this paper we investigate the effects of variations in the timing and location of choice experiment questions concerned with conservation of a UK national park, as research involving measurement of psychological well‐being suggests potential differences for the same individual dependent upon when and where preferences are elicited. We apply the choice experiment technique to the valuation of changes in upland agricultural and semi‐natural landscapes in the Peak District National Park in the UK, to investigate whether timing and location of elicitation (context) affects the value associated with changes in ecosystem services under different management regimes. Four treatments are employed – using the same sample of individuals answering the same choice scenarios – to measure WTP ex‐ante (off site), in situ (on site), and ex‐post at two different time intervals (off site). We show that our on‐site (in situ) treatment generates very different estimates of preferences than any of the off‐site treatments. That stated preferences associated with environmental goods are so context dependent may have implications for the use of stated preferences in policy analysis in terms of identifying how environmental policy is funded and the divergence in value attributed to sampling different populations.     

Value elicitation for multiple quantities of a quasi‐public good using open ended choice experiments and uniform price auctions

Choice experiments and experimental auctions have become popular mechanisms for estimating willingness to pay (WTP). However, these methods have primarily been used for estimating WTP for single units of goods. We analyze the results from experimental auctions and choice experiments in the context of multiple quantities of a quasi‐public good (animal welfare product). We show that the use of WTP values for a single unit of a product, a common practice in experimental valuation literature, can result in underestimation of aggregate demand. We use and compare open ended choice experiments (OECE), second price Vickrey auctions, and random Nth price auctions as mechanisms for valuing WTP. Our results also suggest that individual level demand estimates from OECE are less elastic than demand estimates from uniform price auctions.     

Transfers of landowner willingness to accept: A convergent validity and reliability test using choice experiments in two Canadian watersheds

We examined the reliability and validity of transferring estimates of marginal willingness to accept and compensating surplus. In doing so, we used data from two case studies applying choice experiments to elicit landowner preferences for incentive-based wetland conservation programs in two adjacent watersheds in Southern Ontario, Canada (Grand and Upper Thames Rivers in parallel in 2013). The choice experiment data were modeled in willingness to accept space using a generalized multinomial logit. Transfer reliability was investigated by calculating transfer errors, while validity was investigated by testing the equality of utility functions as well as by assessing the similarity of welfare estimates using traditional hypothesis tests and equivalence tests. The main findings are that transfers of willingness to accept are similar to existing transfers of willingness to pay in terms of validity and reliability. Additionally, a sensitivity analysis finds that including demographic variables in the choice model can lead to lower transfer validity though does not substantially affect reliability. Though further research is required, our results suggest that willingness to accept can be transferred as part of policy analyses.     

Stated Attribute Non‐attendance in Successive Choice Experiments

Attribute non‐attendance in choice experiments affects willingness‐to‐pay (WTP) estimates and therefore the validity of the method. A recent strand of literature uses attenuated estimates of marginal utilities of ignored attributes. Following this approach, we propose a generalisation of the mixed logit model, whereby the distribution of marginal utility coefficients of a stated non‐attender has a potentially lower mean and lower variance than those of a stated attender. Model comparison shows that our shrinkage approach fits the data better and produces more reliable WTP estimates. We further find that while reliability of stated attribute non‐attendance increases in successive choice experiments, it does not increase when respondents report having ignored the same attribute twice.     

Design efficiency for non‐market valuation with choice modelling: how to measure it,what to report and why*

We review the basic principles for the evaluation of design efficiency in discrete choice modelling with a focus on efficiency of WTP estimates from the multinomial logit model. The discussion is developed under the realistic assumption that researchers can plausibly define a prior belief on the range of values for the utility coefficients. D‐, A‐, B‐, S‐ and C‐errors are compared as measures of design performance in applied studies and their rationale is discussed. An empirical example based on the generation and comparison of fifteen separate designs from a common set of assumptions illustrates the relevant considerations to the context of non‐market valuation, with particular emphasis placed on C‐efficiency. Conclusions are drawn for the practice of reporting in non‐market valuation and for future work on design research.     

Information processing strategies and framing effects in developing country choice experiments: results from rice farmers in India

This study explores the effect of various information processing strategies specifically related to attribute nonattendance in stated choice experiments. Our approach includes two forms of stated choice task nonattendance, each derived from separately framed questions asked immediately following the completion of each choice task. We also evaluate a measure of inferred serial nonattendance based on the posterior distributions of random coefficient estimates. We find that choice task nonattendance question framing statistically impacts marginal utility coefficients and, to a lesser degree, willingness‐to‐pay estimates. While direct questions addressing attribute attendance or nonattendance affect these estimates, inferred indicators of serial nonattendance suggest that many attributes are not likely ignored as often as respondents may indicate. Further research is needed to assess inferred versus stated approaches to modeling respondent information processing strategies.     

Consumer sorting and hedonic valuation of wine attributes: exploiting data from a field experiment

This article uses a novel experimental approach to measure consumer willingness to pay (WTP) for wine attributes. We invited customers of a local supermarket who had selected a bottle of wine to purchase to participate in a valuation experiment. Integrating their original wine choice into the experiment, each participant evaluated six alternative wines, generating a rich set of data on willingness to pay and consumer characteristics. The data from the experiment allow us to compare standard shelf price‐based wine attribute valuation estimates with estimates using WTP data and an increasing amount of information about individual consumers. The full model employs individual fixed effects to estimate WTP parameters without bias from consumer sorting or supply side influences. Our WTP estimates for wine attributes differ markedly from previous attribute value estimates. Consumers in our sample display clear and stable preferences for wine varieties, but less clear preferences for appellations. Our results suggest caution is needed in using market prices to estimate parameters of the consumer valuation function for product attributes.     

Using choice experiments to estimate consumer valuation: the role of experimental design and attribute information loads

With fixed dimensionality of choice experiments (CEs), previous simulation results show that D‐optimal design with correct a priori information generates more accurate valuation. In the absence of a priori information, random designs and designs incorporate attribute interactions result in more precise valuation estimates. In this article, Monte Carlo simulations demonstrate that the performances of different design strategies are affected by attribute information loads in CEs. Consumer valuation estimates in simulation settings vary with the number of attributes.     

The value of green walls to urban biodiversity

Despite increased interest in the implementation of green walls in urban areas and the recognised benefits of monetary valuation of ecosystem services, no studies have been undertaken to estimate the economic value of biodiversity they provide. The valuation of natural resources allows policy makers to justify resource allocation. Using the Southampton, UK, as a case study, this paper estimates the public’s perceived value of green walls to urban biodiversity, in the form of their willingness to pay (WTP). Estimates were derived using a random parameter model that accounted for socio-economic and attitudinal determinants of choice, using choice experiment data. Three green infrastructure policies were tested; two green wall designs (‘living wall’ and ‘green façade’) and an ‘alternative green policy’; and compared against ‘no green policy’. Results indicated a WTP associated with green infrastructure that increases biodiversity. Attitudinal characteristics such as knowledge of biodiversity and aesthetic opinion were significant, providing an indication of identifiable preferences between green policies and green wall designs. A higher level of utility was associated with the living wall, followed by the green façade. In both cases, the value of the green wall policies exceeds the estimated investment cost; so our results suggest that implementation would provide net economic benefits.     

Using Geographically Weighted Choice Models to Account for the Spatial Heterogeneity of Preferences

In this paper, we investigate the use of geographically weighted choice models for modelling spatially clustered preferences. We argue that this is a useful way of generating highly‐detailed spatial maps of willingness to pay for environmental conservation, given the costs of collecting data. The data used in this study come from a discrete choice experiment survey of public preferences for the implementation of a new national forest management and protection programme in Poland. We combine these with high‐resolution spatial data related to local forest characteristics. Using locally estimated discrete choice models we obtain location‐specific estimates of willingness to pay (WTP). Variation in these estimates is explained by characteristics of the forests close to where respondents live. These results are compared with those obtained from a more typical, two stage procedure which uses Bayesian posterior means of the mixed logit model random parameters to calculate location‐specific estimates of WTP. We find that there are indeed strong spatial patterns to the benefits of changes to the management to national forests. People living in areas with more species‐rich forests and those living nearer bigger areas of mixed forests have significantly different WTP values than those living in other locations. This kind of information potentially enables a better distributional analysis of the gains and losses from changes to natural resource management, and better targeting of investments in forest quality.     

Comparing Serial,and Choice Task Stated and Inferred Attribute Non‐Attendance Methods in Food Choice Experiments

A number of choice experiment (CE) studies have shown that survey respondents employ heuristics such as attribute non‐attendance (ANA) while evaluating food products. This paper addresses a set of related methodological questions using empirical consumer data from a CE on poultry meat with sustainability labels. First, it assesses whether there are differences in terms of marginal willingness to pay estimates between the two most common ways of collecting stated ANA (serial and choice task level). Second, it validates the self‐reported ANA behaviour across both approaches. Third, it explores the concordance of stated methods with that of the inferred method. Results show that WTP estimates from serial‐level data differ from those from choice task‐level data. Also, self‐reported measures on choice task ANA are found to be more congruent with model estimates than those for serial ANA, as well as with inferred ANA.     

Buyers’ perceptions of importance and willingness‐to‐pay for certain attributes of source and production verified bred heifers

This research reports buyers’ perceptions of and willingness‐to‐pay for replacement heifers produced through a rigorous, third‐party verified production protocol. Survey respondents attended and registered to purchase heifers at sanctioned Missouri Show‐Me‐Select Replacement Heifer Program^® sales between 1997 and 2002. Responses indicate that pen uniformity, artificially inseminated to calving ease bull, synchronized calving, and heifer size are perceived as important, and their willingness‐to‐pay for these characteristics is economically significant. Though prior research suggests willingness‐to‐pay estimates particularly for inexperienced consumers may be biased, we find little difference between inexperienced and experienced buyers and also little difference from hedonic estimates of heifer characteristics’ value.     

Stated and Inferred Attribute Attendance Models: A Comparison with Environmental Choice Experiments

There is increasing evidence that respondents to choice experiment surveys do not consider all attributes presented in the choice sets. Not accounting for this ‘attribute non‐attendance’ leads to biased parameter estimates, and hence biased estimates of willingness to pay. Various methods exist to account for non‐attendance in the analysis of choice data, with limited agreement as to which method is ‘best’. This paper compares modelling approaches that can account for non‐attendance, based on stated and inferred attribute non‐attendance. Respondents' stated non‐attendance is incorporated in the specification of multinomial and mixed logit models. Inference of non‐attendance is based on equality constrained latent class models. Results show that model fit is significantly improved when attribute non‐attendance is taken into account, and that welfare estimates are lower when incorporating non‐attendance. The inference based on equality constrained latent class models provides the best model fit. There is little concordance between stated and inferred non‐attendance, suggesting that respondents may not answer attendance statements truthfully.     

Spatial and Scope Effects: Valuations of Coastal Management Practices

We explore how values for environmental management in the Kimberley region of Australia respond to changes in spatial scale and attribute scope. A discrete choice experiment was conducted that included the impacts of management on marine no‐take areas, Aboriginal rangers, recreational facilities, and coastal development. A split sample single‐site design was used to estimate values for the Kimberley region as a whole, and for two separate smaller sub‐regions, allowing us to test for spatial preference heterogeneity. Management outcomes for different regions were displayed on a map to show respondents explicitly where outcomes would occur. We show that willingness to pay results are similar between the two smaller sub‐regions, and that willingness to pay for the attributes increased when management occurred at the larger geographical scale. However, respondents were somewhat insensitive to changes in the scope of the two cardinal attributes: area of no‐take and number of rangers. We discuss the implications of this spatial and scope insensitivity for choice experiment research.     

Scale and taste heterogeneity for forest biodiversity: Models of serial nonparticipation and their effects

Serial non-participation is a response behavior that is frequently found in stated choice experiments. One form of serial non-participation is that a varying number of respondents chooses always the zero-price or status quo alternative. The approaches used in the literature to deal with this problem vary from excluding those respondents to using applying latent class models to endogenously allow for different preference structures. However, latent class models also allow to assign respondents to a known class. In this paper we compare specifications of latent class models with and without restrictions. Additionally, we control for differences in the error variance across respondents by applying a scale-extended latent class model. The comparisons are designed to show whether respondents are allocated differently among classes and whether willingness to pay estimates are affected significantly. The data are from a choice experiment regarding management actions to enhance forest biodiversity. The data set is characterized by a high percentage of respondents who were not willing to pay; almost 50% of the respondents always chose the zero price alternative locating this sample rather at the top of the range of respondents who are not willing to pay.     

Predicting versus testing: a conditional cross‐forecasting accuracy measure for hypothetical bias*

A measure of hypothetical bias, or the divergence between stated and revealed preferences, based on conditional cross‐forecasting accuracy is suggested, based on out‐of‐sample prediction accuracy when estimates from stated preference data are used in place of those from actual choices, and vice versa. We describe an application of this measure to assess hypothetical bias in the context of an inquiry into people’s willingness to pay to avoid canola oil produced from genetically modified plants. The analysis suggests the presence of groupwise hypothetical bias in these choice data.     

Demand for Crop Insurance in Developing Countries: New Evidence from India

Determining farmers’ real demand for crop insurance is difficult, especially in developing countries, where there is a lack of formal financial sector integration and a high reliance on informal risk mitigation options. We provide some new estimates of farmers’ willingness‐to‐pay for insurance in the context of a large‐scale subsidised programme in India. We conducted a discrete choice experiment with agricultural households across four states in India, enabling us to estimate preferences for specific insurance policy attributes such as coverage period, method of loss assessment, timing of indemnity payments and the cost of insurance. Our results suggest that farmers do value crop insurance under certain conditions and some are willing to pay a premium for such coverage in excess of the subsidised rates they are currently required to pay under this programme. In particular, farmers value the assurances that they will receive timely payouts when they incur losses, and may not have a strong preference for the method with which losses are assessed. On the other hand, farmers are quite sensitive to coverage periods. Our baseline assessment shows that when optimised to farmer requirements, there can be a sizeable demand for crop insurance by developing country farmers.     

Managing the energy trilemma of reliability,affordability and renewables: Assessing consumer demands with discrete choice experiments

In response to the looming climate crisis, many countries are adopting technologies to reduce the accumulation of greenhouse gases. However, national energy policies are often multiobjective and resolution deeply divisive. The result is a policy trilemma between the energy mix and the trade-offs with other policy objectives, including cost and reliability. Utilising a discrete choice experiment (DCE), the objective of this study is to explore Australian household preferences for alternative electricity contracts containing features reflecting changes in future energy policy. The first set of features include investments in renewable generation and community-based energy storage. The second set of features reflect demand-side management policies, including installing smart meters and consumption limits being imposed on households during peak demand. Two versions of the DCE were developed to obtain both willingness to pay and willingness to accept estimates for the same features. In line with the literature, differences in the two sets of estimates were observed, with the willingness to accept estimates being statistically larger for some features. These dollar value measures can be used to support public policy decision-making – the choice of which depending on the context of the policy problem being considered.