1。Transaction price and location of  dwellings;

2。Property characteristics, such as number and size of rooms, and lot  size;

3。Neighbourhood characteristics and accessibility, such as crime rates, distance to shopping centres, and availability of public transportation; and

4。Environmental  characteristics  in question。

Once the data are collected, the next step is to statistically estimate the hedonic function that relates property values to these characteristics by means of multiple regression analysis, which measures the portion (i。e。 the partial derivative) of the property price that is attributable to each characteristic。 Hence, the value of the environmental asset can be estimated by the partial derivative, which is the implicit marginal attribute price, ceteris paribus [62]。 It is then possible to trace back the inverse demand function based on the implicit marginal price function, with respect to certain environmental quality changes。

Although the principal concept is simple, there are serious difficulties and limitations with putting the method into practice。 First, extensive cross-section data, time series data or a mixture of both should be used。 Gathering a suitable data set can be time consuming, expensive and, in some cases, impossible, since data are not always available。 Second, given that a  property value function depends on many variables, in order to measure their impact on the price, they must all be included in the regression model。 Nevertheless, in many cases, due to data unavailability, some explanatory variables are omitted。 Butler [68] argues that a model containing a few variables would have a significant effect on the results。 On the other hand, when a hedonistic function contains a large number of variables, multi- collinearity may undermine the results [6,69]。 The form of the hedonic model cannot be specified on purely theoretical grounds and must be determined empirically [9,62,70] and the results are highly sensitive to model specification and level of disaggregation [6,70]。 The temporal stability of hedonistic value equations should also be considered [70], especially when housing markets receive significant shocks over a given time period。 Problems also originate from market distortions, which exist in many countries。 For example, in the UK, the public sector owns a substantial part of housing stock and the rents are politically influenced [6]。 Finally, hedonistic studies cannot be easily applied when potential changes in environmental quality are investi- gated and are mainly focused on ex post   analyses。

3。5。 Benefit transfer method

In general, original environmental valuation studies are expansive and time consuming and, in the majority of cases, resources in terms of time and money  are barely available。 The latter  has  resulted  in  adopting the use of the ‘‘benefit transfer (BT)’’ method, which refers  to  the  application  of  the  results  obtained from

a particular case to another area。 The site of the original research is usually called ‘‘study site’’, while the site to which the benefit estimate is transferred is called ‘‘policy site’’。 This process is a cost-effective way to evaluate the environmental effects of projects when original research is not possible due to budget constrains and/or time limitations。 In any case, primary research is the ‘‘first- best’’ strategy。 However, when primary research cannot be carried out, BT provides useful information for decision making [71], especially in cases where a high degree of precision is not  critical。

Several necessary conditions should be met to perform effective and efficient BTs [71,72], including: studies transferred must be based on adequate data, sound economic methods and correct empirical techni- ques [73] and the environmental resource and the change in quality reflected in policy should be similar to the study sites。 Several factors may also affect the reliability and validity of this method [71,74], including the quality of the original study, the different research methods that may have been used for data collection, the different statistical methods that may have been applied for esti- mating models, and the issue of temporality or stability of data over time, since the existing studies occurred at different points in  time。