Which of the following is not predicted by the hedonic theory of wages, all else constant?

Statistical Methods for Modeling Land Use

From a human economic perspective, one of the most pertinent factors in land-use decisions is land value. Determinants of land markets have been extensively studied to quantify individual preferences for property attributes. One commonly employed method, hedonic analysis, uses the sale price of a parcel as a proxy for measuring individual willingness to pay for structural and locational property variations. A multiple regression approach is used to estimate the property price based on a suite of attributes. Table 1 provides a list of common covariates grouped into three categories: structural, neighborhood, and environmental. Equation [1] illustrates a general specification of a hedonic model structure:

[1]Pi=β1+β2(structural)+β3(neighborhood)+β4(environmental)+εi

where Pi = price, β1 = an intercept variable, βn = vectors of coefficients for the model variables, and εi = standard regression error term.

Table 1. Commonly used covariates for hedonic analysis

A multiple regression approach has also been used to estimate the demand for quantities of particular land-use types. The area of interest is divided into a set of discrete zones (e.g., block group, traffic analysis zone), and an array of predictor variables is used to estimate the area of land use by type. The dependent variable in this case, the area of each type of land use, is estimated with some combination of social, economic, and environmental variables specified in much the same way as the hedonic price model discussed above. While these approaches may be useful for explaining consumer preferences and historical land-use patterns, they have limited predictive ability to capture the nonlinear and emergent behavioral patterns that characterize urban transformation (e.g., edge cities).

Discrete choice analysis models the probability of a specific outcome from the set of alternatives that constrain a decision-making process. This technique commonly employs a binomial or multinomial statistical model, such as logit (see eqn [2]) or probit, where decisions are represented by a maximum likelihood or random utility function that is defined by independent variables whose values are estimated based on the actual decisions of individuals when faced with similar choices. Choice sets can be continuous or discrete and can occur as a single decision or within a nested structure. An example of the former would be representing a development event binomially, while the latter would be a choice between development and no development, followed by a nested choice within the development branch between commercial or residential.

The logistic regression or logit function is given by

[2]EY =expβ0+β1X1+⋯βp −1Xp−11+expβ0+β 1X1+⋯βp−1Xp−1

where E(Y) = the expected response value, β0 = an intercept variable, and βn = coefficient representing estimated odds ratios for variable Xn holding all else constant.

While these statistical approaches could serve as stand-alone models of land-use change, they are more commonly employed as components of more complex models. For example, the California Urban Futures (CUF) model uses a regression approach to calibrate the land-use change submodel against historical land-use changes using population and employment growth, proximity to employment, commercial and industrial land uses, and spatial and topographic factors as independent variables. Another example is UrbanSim, where hedonic analysis is used to estimate the land price model and a multinomial logit model is used to estimate probabilities that households and firms will occupy a particular parcel. Finally, discrete choice models are commonly employed in travel demand, cellular automata (CAs), and agent-based models, which are discussed below, and are particularly used to calibrate parameters from empirical data.

4.4.2.5 Qualities of Lakes Associated with Property Value

Humans derive substantial pleasure from simply seeing lakes, from shorelines, and shorelines from lakes. Studies of landscape preferences consistently reveal that humans prefer landscapes that contain water features compared to similar landscapes that lack water features (Kaplan and Kaplan, 1989). This aesthetic value translates directly to economic value. Hedonic analysis of home prices in the Twin Cities found that a shoreline location increased property value by $111 000 and even being within 200 ft of a lake added $61 000 to a property's value (Moscovitch, 2007). In a study of several cities in the Netherlands, ‘overlooking water’ added 8–10% to the value of a house, and a ‘garden facing water’ added 28% (Luttik, 2000). Lansford and Jones (1995) concluded that recreational and aesthetic values reflected by hedonic analysis – amounted to 15% of the total property value within 2000 ft of Lake Travis. Some wetlands also increase the value of nearby property. Doss and Taff (1996) reported that property values increased with proximity to open water wetlands but not forested wetlands.

Given the large premium on property values associated with proximity to lakes, there has been surprisingly little analysis of environmental qualities of lakes in relation to property value. The few studies that have been done have focused primarily on the relationship between water clarity and property value, using hedonic analysis (Steinnes, 1992; Michael et al., 1996; Boyle et al., 1998; Gibbs et al., 2002; Krysel et al., 2003). The effect of a loss of 1 m clarity varies among lakes and can range from a few percent to as much as 34% (Long lake, in Michael et al., 1996). Gibbs et al. also reported that clarity influenced purchasing decisions of 46% of survey respondents. Dodds et al. (2009) estimated annual US lakeshore property value losses because of eutrophication of $0.3–$2.8 billion per year. To our knowledge, there are no hedonic model studies to relate lake clarity to property values in urban areas, and few that relate property values to other aspects of lake aesthetics, although some studies have evaluated effects of invasive species. For example, Horsch and Lewis (2009) found property values dropped an average of 13% after Eurasian watermilfoil (M. spicatum) invasion in Wisconsin lakes.

Accounting Approaches and Basic Valuation Methods

The simplest approach to assessing ecosystem services for marine and terrestrial systems is a straightforward accounting of those services and their values. This approach is not one that explicitly models the ecological or human systems, but it has been used to assess marine ecosystem services (Beaumont et al., 2007, 2008). Generally, the approach begins by defining a particular geographic location and then identifying the set of ecosystem services flowing from that area. The assessment can then be as simple as compiling local information on each of these services for that region – e.g., using revenues from an area's fishery harvests as a measure of the economic value of that type of provisioning service.

A more common method that has been used for terrestrial systems is ecosystem-service mapping (Troy and Wilson, 2006). This method differentiates a particular area by land cover, biome, or some other set of ecologically based landscape or seascape type. Drawing from a standard set of ecosystem service categories (e.g., deGroot et al., 2002), each landscape type (e.g., forestland) is then linked to a set of services (e.g., recreation or carbon sequestration) believed to be provided by that type. The quantity of an ecosystem service produced is then assumed to be linearly related to the area of the landscape type.

From these starting points, local economic data, if they exist, can be used to transform the quantity of ecosystem services into economic values. Such data are not commonly available, however, and so the more typical approach to assigning economic value is to use what is known as benefit transfer. Benefit transfer is a method for taking economic data on benefits (or values in general) gathered from one site and applying the data to another site. This method is rarely the “first-best” choice for estimating economic values but the costs of gathering primary, site-specific data have made it a common practice for studies of ecosystem service value (e.g., see Rosenberger and Loomis (2001) and National Research Council (2005) for examples of recreational uses of natural sites). Using the assumption of a linear relation between ecosystem service quantities and landscape area, a “unit value” for a particular ecosystem service and landscape type can then be derived by taking an existing study's estimated value for a particular ecosystem service and dividing by the area of the studied site's landscape (e.g., recreation or carbon sequestration value per acre of forestland). These unit values can then be used to assess the value of ecosystem services at other sites for multiple services and landscape types if they are present.

This accounting approach was pioneered by Costanza and colleagues (1997), who estimated global values for 17 ecosystem services for 16 biomes. Their effort included five marine and coastal biomes: open ocean, estuaries, seagrass or algae beds, coral reefs, and ocean shelf. These five biomes accounted for almost two-thirds of the estimated annual $33 trillion value of Earth's services. Their use of benefit transfer was almost universal; the original data used in the study were gathered in one or more locations and then projected worldwide.

Accounting approaches such as Costanza et al. (1997) are simple and play the important role of raising the awareness of the values of traditionally undervalued, nonmarket ecosystem services. They should be used cautiously, however, with important qualifications in mind. First, this approach imposes linearity on the valuation of ecosystem services, which has the potential to produce biased estimates. Human pressures on ecosystems and the services they provide can result in impacts that respond nonlinearly to changes in the scale of the pressure or change discontinuously if a threshold is crossed (Groffman et al., 2006). The valuation of ecosystems services should therefore account for such nonlinearities if the scale under consideration is more than minimal (Barbier et al., 2008; Koch et al., 2009). The human values that apply to ecosystem services can also exhibit nonlinearities as the quantity or quality of the ecosystem service available changes (Bockstael et al., 2000; Toman, 1998). Again, if the scale of a service under consideration is large, projecting a value estimated for a small change has the potential to produce significant bias. And if more than one service is evaluated, the aggregate value of the group of services is likely to be nonlinear with respect to considering more and more services. If estimates of individual ecosystem service values are used in such an exercise, simply adding them together will again produce a biased estimate of the group's value. These problems grow more and more significant as the scale of the ecosystem service valuation exercise increases. Given that this approach is often used to produce national or even global estimates (Anielski and Wilson, 2009; Beaumont et al., 2007, 2008; Ingraham and Foster, 2008; Naidoo et al., 2008), the issue of nonlinearities is an important one.

Second, the method requires rigid association of a particular set of ecosystem services to a particular landscape type; thus it does not allow assessment of how policies might change the flow of ecosystem services and affect their values, short of comparing values under wholesale destruction of the landscape type. Relying on landscape type to assign a fixed presence and quantity of services provides little information to evaluate how a policy will change ecosystem service values in any way other than a change in the landscape type. This “all or nothing” approach may make sense for some cases – for example, replacing forest with open agricultural fields – but not for others, as it is highly unlikely that any policy is capable of converting saltwater estuaries into upland forestland.

A third caution stems from the common use of benefit transfer for ecosystem service mapping exercises and other accounting approaches. The broad categories used in these types of exercises facilitate a comprehensive analysis of ecosystem service values for large geographic areas, but they also subject the analysis to a high likelihood of what is called generalization error (Plummer, 2009). As an example, consider the problem of estimating the value of providing “recreation” in a “marine area.” Recreation covers scores of possible market and nonmarket activities that take place in natural settings, and some may be present in one marine area but not in others. Also, the presence of human-built infrastructure and the accessibility of the marine area are important determinants of the economic value of the recreational activity. Because these factors can vary widely, using estimates of economic value from one site and applying them to another is likely to produce significant errors if the categories of “recreation” and “marine area” are treated at a high level of generalization.

Finally, existing estimates of economic values for ecosystem services are strongly influenced by the methodology employed to make the estimate. As the NRC report on valuing ecosystem services (2005) emphasizes, not all methodologies are equally good. Happily, several valuation approaches are accepted practice by economists, but the interpretation of the resulting values should be informed by the method used.

A final revealed preference method is useful for cultural ecosystem services such as recreation and other environmental experiences that take place outside any formal market. In this approach, the cost of engaging in the activity can be used to derive estimates of its economic value (Clawson, 1959; Knetsch, 1963). Similar to the assumptions for hedonic models, the recreation “good” can be viewed as a bundle of characteristics, some of which are the environmental features important to the recreational experience. If data are available for visits to multiple sites with varying levels of those features, one can then estimate the contribution of a particular feature to the demand for that recreation and from this estimate the feature's value (Morey, 1981).

Economic Valuation Methods

Much of the impetus to value ecosystem services comes from the US federal mandate to evaluate the environmental trade-offs present in any land-use or resource management policy decision. CBA was first used in the 1930s to justify budget allocations for US Army Corps of Engineers water resource development projects. In response to the tendency for CBA to count only benefits for which prices were readily available, Congress included in the National Environmental Policy Act of 1969 a stipulation that Environmental Impact Statements must also consider noneconomic values that might be impacted by given projects. Rather than encouraging policymakers to reach beyond economic valuation, this incentivized research into increasingly sophisticated valuation techniques that could capture the economic value of nonmarketed elements of nature. As attention has turned to the conservation of ecosystem services, these economic valuation tools have been applied to capture their value in environmental decision-making. Executive Order 12291, issued in 1981, required that CBA be undertaken for any new federal regulation, and required that, any regulation's benefits must outweigh the costs to society. This requirement is relaxed by Executive Order 13258, issued in 1996, which simply stated that the benefits of the regulations must justify the costs.

The monetary value of ecosystems can be measured in several ways; mainstream economists use the concept of ‘Total Economic Value’ to classify the values in ecosystem services. Methods used to find the value of ecosystem services may be classified into two categories. The first are known as revealed preference methods, and involve gathering data by observing the behavior of individuals in real-world situations involving interaction with ecosystem services. There are five main methods in this category:

Hedonic models: All other features being equal, price differentials between similar goods can be attributed to variance in the features of interest. In the case of ecosystem services, hedonic analysis has been used to observe costs for residential lots that are associated with differential access to aesthetic services provided by nearby wetlands. For identical housing lots, the difference in price should be due to the presence or absence of wetlands.

Travel cost: This approach tallies the costs in travel, fuel, and time that individuals expend to consume ecosystem services.

Averting behavior: The inverse of travel cost analysis is the assessment of the costs that individuals incur to avoid undesirable outcomes caused by the degradation of ecosystem services.

Replacement cost: An analysis of the expenditures required to perform the service provided by ecosystems.

Production function: This method is borrowed from standard analyses of industrial production. Assuming that ecosystem services provide inputs to certain kinds of commodity production (known as factor inputs), changes in ecosystem services can be linked to changes in the costs and supply of the final commodity. In traditional economic analysis, where an entire industrial production process is well understood and priced in noncontroversial ways, production function analysis is a powerful tool. To be used in the valuation of ecosystems services, it is crucial that underlying economic and ecological relationships be understood as well as, say, the factor inputs in steel production are understood in alloy manufacture. This is often not the case where ecological functions are concerned.

The second category of monetary valuation methods is known as stated preference methods, in which data are collected through posing questions to individuals about hypothetical valuation scenarios, usually in survey form. They are usually applied where there is no market behavior to observe that bears on the ecosystem service. Since this is usually the case, stated preference models are extremely common in assigning price value to ecosystem services. There are two principal methods in this category:

Contingent valuation (CV): CV methods include willingness-to-pay (WTP) and willingness-to-accept (WTA) scenarios. In a WTP analysis, surveyors ask individuals what they would be willing to pay to experience or consume the ecosystem service in question. In a WTA analysis, surveyors ask individuals what they would be willing to accept as compensation for being prevented from experiencing or consuming or conserving a decrease or increase in the provision of an ecosystem service in question.

Conjoint valuation: This method involves the elicitation of monetary values for key components of an ecosystem, presenting many scenarios that vary in small ways.

A final source of ecosystem service value estimates is a technique known as benefits transfer. In this widespread, but controversial, method, researchers use data on ecosystem service values generated at one location to apply to many other locations. Thus, the value of flood abatement services defined through a CV survey in Arizona will be transferred to inform a policy debate on the value of wetlands in Oregon or Tunisia. It is perhaps the most broadly used method of asserting the monetary value of ecosystem services, but also subject to sharp criticism for poor calibration and validation.

There is a general understanding, even among enthusiasts for ecosystem services valuation, that economic valuation does not provide transparent policy prescriptions, and that it may not be possible to attach monetary value to all features of the environment. But, however many caveats are attached to ecosystem service valuation, the monetary value attached to an ecosystem function is often transmitted to the public or policy realm without qualification. For all the sophistication of these varied methods, the results can be frustrating for policymakers and ecosystem service providers (ESPs) on the ground. Estimates of theoretical value are often considerably inflated over what is realistically tradable by the owners or produces of ecosystem services. The disconnect between economic theory and policy practice in the ecosystem services literature is often jarring.

4 Discussion

The quantitative analysis of the bibliometric analysis between 2000 and 2020, shows that the LCC is by far the most used methodology in the economic assessment of nature-based solutions. CBA stands in second place; however, these methods have difficulties in monetizing external (environmental, social) impacts. The use of EEA models, even though limited, is increasing recently. This demonstrates a growing interest for a more integrated (economics and environmental) approach. Other types of approaches including probabilistic analysis, hedonic price method, empirical monitoring data, multiple linear regression models, inexact two-stage stochastic programming, and travel cost are less frequently used than those discussed earlier.

The results of the analysis of the economic impact assessment tools reveal that there is a wide range of methodologies and tools which are currently suggested to evaluate the impact of NBS in the water sector. Table 4 summarises the most widely applied economic indicators of impact assessment of NBS on water systems, which were discussed in this study. It is worth noting that indicators such as NPV or BCR can be considered as universal and common indicators shared in many studies and could be considered as Base parameters in a Systems Approach. Whilst indicators like the value of wood production, construction, and maintenance cost in the economic assessment of green infrastructure (Liquete et al., 2016) are considered as case-based parameters.

Table 4. List of economic indicators for sustainability assessment in water systems.

The study reveals that despite the existence of a wide range of methods and tools for economic impact evaluation of NBS and Water Management Systems, there is still scope for a Systems Approach in establishing standards and protocols for a global economic impact assessment for modern Sustainability and Circularity initiatives. The use of haphazard case-based indicators alone may not necessarily convince policymakers and investors to shift from existing solutions.

The LCC and CBA are the most applied approaches for the economic evaluation of NBS. Although the results of this study revealed that both the LCC and a CBA can be used within a sustainability assessment framework, there are some key parameters, that should be taken into account. The dissimilarity between LCC, and CBA is the first point. While LCC can be classified as product-related evaluation, CBA focuses more on programmes or policies (Ness et al., 2007; Rorarius, 2007). The timespan is a second dissimilarity factor. LCC concentrates on the economic life cycles of the target products, while CBA first focuses on the lifetime of a specific project and then the lifetime of the target products will be considered (Hoogmartens et al., 2014). The third dissimilarity of LCC and CBA is about their usage for comparison purposes. LCC is a comparative assessment method that evaluate goods, while CBA is usually used for an independent evaluation of projects. For example, the estimation of the NPV give a clear result for CBA analysis of the system without requiring a comparison with other alternative systems. Therefore, the integrated application of LCC and CBA is important for a sustainability assessment study. The paper contributes to the understanding of contradictory assessment outcomes and the harmonisation of assessment methodologies by clarifying the main aspects and the relationships between the various methods and tools.

Additionally, in order to estimate the economic impact of a water system in a circularity and sustainability assessment, environmental and social costs and benefits should be taken into account along with economic cost and benefit in the quantification of externalities principles. The environmental (actual impacts, e.g., increase or loss of biodiversity, and potential impacts, e.g., global warming potential) and social (health, job creation) aspects should be translated to a monetary value in order to be considered in an evaluation. Fig. 4 presents the data and indicators required to obtain a comprehensive economic impact evaluation of moving toward a circular water system following a systemic approach. This latter considers the economic, environmental, and social costs and benefits at the same time. The estimation of the monetary value of environmental and social impacts should be obtained through some pricing methods including shadow pricing and hedonic pricing. An integrated method of life cycle cost-benefit analysis (LCCB) can be employed to analyse the collected data and present the results to investors and decision-makers.

4.2 Outcomes of Evaluation

Our evaluation of methods found that each of the recreation accounting techniques tested require some degree of trade-off amongst the conceptual fit, information cost, and sensitivity evaluation criteria we have investigated. This was particularly true for production-based methods. Table 4 summarises the outcomes of the evaluation process.

Table 4. Summary of the methods evaluation

*Negative sign indicates high sensitivity to exogenous factors.

**In many cases, sensitivity to assumptions is not an insurmountable obstacle to accounting, but it may require further standardisation and harmonisation to be built in to SEEA EEA revisions. Sensitivity to exogenous factors inherent in some of the methods is more problematic.

^Consumption based methods are consistent with other SNA processes as discussed in Section 4 and we argue that consumption based methods are suitable for SEEA-EA.

We found that, from a conceptual perspective, the RUM method clearly allocates value to ecosystem and infrastructure but fails to account for tourism sector contributions. The method meets the need for spatial differences as it captures service provision per individual park. Modelling effort associated with this technique is substantial and requires a statewide travel cost analysis with random utility modelling. In NSW, we have a similar large-scale hedonic analysis of the amenity-type values of greenspaces currently underway. We estimate that together these two pieces of work (which would both be required to account for recreation and tourism services as described below) would take approximately two years to complete for the 700+ protected areas in the NSW National Parks Estate. This approach clearly cannot meet an annual accounting schedule to inform real-time decision-making. Such large empirical studies could be carried out at inter-annual intervals of 5 to 10 years instead, when it is reasonable to expect structural variables to have changed. For intervening annual accounting periods, parameters determining accounting prices are kept fixed. But to our knowledge, this type of all-parks study has only been undertaken to date in NSW, the U.K and the U.S.A., an indication that other jurisdictions may find the information cost to be prohibitive.

The information required for the Activity-Based and the Costs of Production Methods is already routinely and regularly collected in NSW, so the additional information cost would be minimal. But from a conceptual perspective, both these methods fail to identify the ecosystem contribution to tourism sectors (Activity-based) or to capture those contributions altogether (Costs of Production). The Simulated Exchange Value is based on concepts of supply and demand and benefits from a good conceptual fit with standard economic theory (although it is unclear what the equilibrium quantity represents), but again, the conceptual fit with Logic Chain 3 is problematic as the method does not account for the contributions of tourism sectors.

Compared to the Activity-Based and Costs of Production Methods, we consider the Resource Rent to be a poorer conceptual fit for nature-based recreation; it captures the ecosystem service flowing to supporting sectors rather than the main beneficiaries, and the monetary value of the service is obtained from business operating surpluses even though the service was obtained for free (Horlings et al 2019b). This method does not allow spatial differentiation as the information required for tourism sectors is not available at fine regional scale and statewide industry averages were used. An additional technical issue is that resource rent (and RUM) approaches may yield negative values for ecosystem contribution to recreation services, when clearly a benefit was obtained by visitors. Fenichel and Obst (2019) point out that negative rents may provide a useful indicator of problems with the management of the asset, but in the case of recreation services in NSW national parks, the level of investment in the asset is mostly driven by conservation objectives and available budget rather than being informed by visitation (i.e. these methods are sensitive to external factors rather than to asset condition).

Similarly to the Resource Rent method, we found the Similar Markets provide a poor conceptual fit as ecosystem, infrastructure and tourism sector inputs to the service are not clearly distinguished. Moreover, it is less likely that changes in the natural condition of the ecosystem lead to changes in service flows, as the economic inputs (accommodation, guided activities) appear to be the main driver of private supply and use. There is currently insufficient information to support spatial differences in ecosystem service supply between regions. The main advantage of this method is that it is relatively simple to implement with limited information required.

From a conceptual perspective, the final consumption approach differs from production-based methods by attributing all of the service value to the ecosystem, accounting for the tourism and infrastructure contributions only indirectly through additional sector output. Since the method is based on actual expenditure, it is not sensitive to exogenous factors, it is spatially explicit and the information cost is the same as the Activity-based, Simulated Exchange and Resource Rent methods. With the Final Consumption Method, contributions from infrastructure and tourism sectors are only accounted for in the main national accounts and not readily apparent in ecosystem accounts. The Mixed Contributions Method provides the additional advantage of clearly isolating all three inputs described in Logic Chain 5.

5.6 Behaviour

Eventually, all above described constructs will constitute aspects of consumers’ purchasing behaviour regarding wine with sustainability characteristics. As none of the reviewed articles analysed purchase behaviour in real market scenarios (with real market transactions), most of the studies implemented one of the following experimental designs to get more realistic preferences and WTP values: choice experiment, conjoint analysis, best worst scaling, different methods of auctions and contingent valuation (see Table A.1: method). Two studies implied the hedonic price method to observe consumers’ value for product characteristics indirectly through market supply data. Results of experimental and hedonic analysis are close antecedents of purchase behaviour, although they are still of a hypothetical nature. Studies applying incentive compatible scenarios with real wines sold and real money paid (Ay et al., 2014; Bazoche et al., 2008, 2015; Grebitus et al., 2013; Schmit et al., 2013; Vecchio, 2013) deliver consumer preferences and/or a WTP which can be considered consumers’ actual purchase behaviour (revealed preference). Because it is difficult to compare WTP values across studies from different countries applying different methods on wines with different sustainability characteristics, we did not report concrete WTP values.

Wiedmann et al. (2014) identified higher willingness to recommend organic wine as well as higher WTP for organic wine compared to conventional wine through an open-ended question. The studies by Pagliarini et al. (2013) and Brugarolas et al. (2010) basically confirmed this result. Most of the participants stated, when directly questioned, that they would be willing to pay a premium for organic wine (Pagliarini et al., 2013) and a large percentage of consumers would pay more for an organic wine in a study applying contingent valuation (Brugarolas et al., 2010). Furthermore, the WTP for organic wine was significantly higher compared to conventional non-local wine in a French study applying the auction mechanism. For local wine, the organic premium was even higher (Ay et al., 2014). Moreover Kwong et al. (2011) found that organic wines command significantly higher prices in a hedonic price analysis on all wines produced in Ontario (Canada) and released for sale. They concluded that Canadian consumers care about viticulture techniques used in the cultivation of the grapes. In addition, organic wine was preferred over conventional wine in a conjoint study on Italian (Chiodo et al., 2011) and Spanish consumers (Bernabéu et al., 2008). However, in a conjoint study on Swiss consumers, most consumers preferred conventional wine over organic wine (Mann et al., 2012). Furthermore, Bazoche et al. (2008) found that the probability to buy and the WTP for Bordeaux wine was not significantly influenced by the organic label in a study applying an auction mechanism. For other environmental labels, no significant result or even a negative relation to purchase probability and willingness to buy was found. Bazoche et al. (2015) also found, in a choice experiment, no positive effect of the organic label on the purchase frequency, but a significant influence of the biodynamic label. Basically the same was revealed in a hedonic price analysis of Delmas and Grant (2008) on Californian wines. Eco-labels had a negative impact on prices, while a price premium for the eco-certification was detected. The result was attributed to consumers’ negative connotation with organic wine. Regarding organic wines without added sulphites, the majority of Italian respondents stated that they would not pay a premium. Instead, they would pay higher prices for naturalness and designation of origin (D’Amico et al., 2016). Conflicting results regarding WTP and preferences for organic wine may be explained to a certain extent by different study countries and varying research designs and methods.

Spanish consumers were willing to pay a price premium for sustainable wine in a study applying contingent valuation (Sellers, 2016) and for three different sustainable wines (environmental, social and ethical) in an Italian study employing an auction mechanism (Vecchio, 2013). Additionally, most of the respondents indicated an intention to pay more for an environmentally sustainable wine in a study on New Zealand consumers where respondents were asked directly if they were willing to pay a higher price for a sustainably produced wine (Forbes et al., 2009). For environmentally friendly wine, US consumers’ WTP was higher compared to regular Colorado wine, but fairly low in a study implementing contingent valuation (Loureiro, 2003). To compare consumers’ WTP for different sustainability logos, Mueller Loose and Remaud (2013) conducted a cross-national study applying a discrete choice experiment. The environmental corporate responsibility label showed a higher marginal WTP than the label related to social responsibility. The organic food logo received the highest implicit valuation by consumers on average, and in each analysed country; the highest WTP values were observed for France, followed by Germany.

Grebitus et al. (2013) found that German consumers were willing to pay more for local wine in a study adopting an auction mechanism: average WTP decreased when the distance of transportation, indicated through a food miles label, increased. Furthermore, Burgundian consumers’ WTP for local wine (Marsannay, a GI from Burgundy) was higher compared to the non-local wine (Vacqueyras, a GI from the Rhône Valley) (Ay et al., 2014). Castilla-La Mancha consumers showed the highest preference for wine from their local area followed by other wines from Castilla-La Mancha, and the lowest preference for national wine (Bernabéu et al., 2008). In contrast, Swiss wine consumers preferred wines from France over wines from Switzerland. They ascribed this to France’s longer tradition for high quality wine and concluded that wine purchasing decisions are made with a clear focus on quality (Mann et al., 2012). Furthermore, Ay et al. (2014) showed that living closer to a vineyard (participants lived 3–11 km away from the nearest vineyard) decreased the WTP for local, organic or conventional wine. This result could be due to the short supply chain facing consumers, the available social networks, and the presence of least-cost alternatives if they buy their wines directly from the closely located producers. However, focusing on organic premiums, participants living far from vineyards showed a smaller premium for organic wine than those living close to a vineyard. Basically, the same result was found by Brugarolas et al. (2010). They found that WTP for organic wine was higher for Spanish consumers from regions where wine production shares a larger proportion of the regional economy. Hence, for local wine purchases the demographic characteristic ‘place of residence’ is an important parameter.

3.2 Benefits