Chapter 6. Value transfer
Value transfers are the bedrock of practical policy analysis in that only infrequently are policy analysts afforded the luxury of designing and implementing original studies. Thus, in such instances, analysts must fall back on the information that can be gleaned from past studies in order to estimate monetary values for some current policy or project proposal. Whether this is a defensible short-cut is the object of value transfer tests, which provide important guidance about situations in which value transfer can be carried out confidently and when practitioners should proceed with more caution. The general lesson is that there are possibly significant trade-offs between simplicity and accuracy of the resulting transfer. Thus, a competent application of transfer methods demands informed judgement and expertise and sometimes, according to more demanding commentators, as advanced technical skills as those required for original research. This is something of a paradox then for surely the point of transfer exercises is to make routine valuation more straightforward and widely used. Another development which may help in this respect is valuation databases (such as EVRI) and “look-up tables” (lists of average values and ranges for various categories of environmental goods and services). These are important facilitators of valuation uptake in policy formulation, although in turn these do require good guidance on practice and use.
6.1. Introduction
Advances in methods to value environmental goods and services (and non-market commodities, more generally) have been a striking feature of modern cost-benefit analysis (CBA). Just as prominent has been the growing use of these methods to help inform policy and investment project choices across an increasing number of countries. Taking full advantage of this apparent willingness amongst decision-makers to employ the fruits of these advances in this practical way depends, in turn, on a number of further considerations.
An example is the need for a crucial ingredient: plenty of original valuation studies which can be applied to these nascent policy and project questions. However, these ingredients are costly and, as a result, may be in short supply. If so, then practitioners may have to be more enterprising in meeting the policy world’s demand for CBA. One example of this initiative is a greater reliance on value (or, more narrowly, benefits) transfer: that is, taking a unit value of a non-market good estimated in an original or primary study and using this estimate – perhaps after some adjustment – to value benefits or costs that arise when a new policy (or investment project) is implemented.
Value transfer is now the subject of a large literature. The reason is obvious. If value transfer were a valid procedure, then the need for costly and time-consuming original (or “primary”) studies of non-market values would be vastly reduced. In other words, the valuation process can make do with fewer ingredients. To stretch this culinary metaphor further, however, a number of other considerations have to borne in mind. The original ingredients need to be of sufficiently good quality for the resulting dish to be palatable and there needs to be a recipe guiding use of these ingredients. For these reasons, Rolfe et al. (2015) describe value transfer as “… superficially attractive …” (p. 4).
It is this, the validity of value transfer – rather than the abundance of good quality studies – that is the primary focus of much of this Chapter. At the risk of caricature, endeavour in value transfer often reflects two possibly opposing traditions. The first reflects a quest to make valuation as accessible as conceivable. An aspect of this, for example, is “look-up” values: standard values for (non-market) impacts routinely valued in the appraisal of policies or investment projects. The second starts with concern that poorly done value transfer may result in policy selection mistakes and a key ingredient in understanding this is tests of when the value transfer works and when it does not. Perhaps ironically, both traditions are understandable. There will surely be an issue about uptake if valuation is the preserve of the highly trained and specialised experts allowing value transfer but requiring it to be ever more sophisticated in its application. Equally, the cause of applying cost-benefit thinking to policy and investment is not helped by valuation estimates which are not sufficiently robust so as to be easily challenged.
This is because the validity and reliability of value transfers remains open to scrutiny, and – as various tests in the literature show – can give rise to inaccuracy of varying degrees of magnitude. Of course, conclusions about the accuracy of value transfers must be contain some degree of pragmatism. Put another way, some inaccuracy is almost inevitable and the finding that transfers are invalid might be based on criteria, which are too strict. As a practical issue, it could be that some greater degree of inaccuracy “does not matter”, although there is a legitimate debate to be had about what this really means.
Clearly then there is a balance to strike and practical considerations should not translate into an “anything goes” approach. The scrutiny of value transfer to date has also been important in showing how it seems to work better in some contexts and situations than in others. The reasons for this are becoming clearer as the empirical record grows along with the quality of the tests conducted. As a result, such findings can help to guide the use of value transfer by indicating when it can be relied upon and when more caution must be applied. This should allow better value transfer to be done as a result. An example here is the issue of spatial variability. For ecosystem services, location matters and transfers which do not account for this could be highly misleading. However, if these spatial considerations can be accommodated, then value transfer can be a useful and possibly powerful means of evaluating new policies and investment projects.
The holy grail of value transfer is a comprehensive database of studies or specific non-market values or, which can be taken “off the shelf” and applied to new policies and projects as needed. A number of examples of these databases now exist, with perhaps the EVRI inventory1 being the most prominent as well as the most longstanding. The establishment of “reference values” and “look-up” tables used by, for example, national government while not common is also a notable development. A critical question is whether these (welcome) developments are accompanied by sufficient guidance about how to transfer values in a valid and reliable way.
The remainder of this Chapter is organised as follows. Section 6.2 provides a definition of value transfer and then goes on to outline the steps that a value transfer approach typically might take. It also looks at ways in which values (to be transferred) might be adjusted to “fit better” the characteristics (of the good and the affected population) that accompany a new policy. A brief review of what is known about how robust these transfers are is then offered in Section 6.3. Section 6.4 describes how one lesson of those tests has been used to guide better use of value transfer for the case of spatial variability. Section 6.5 discusses efforts to develop comprehensive databases of values for use in future transfers. Section 6.6 offers concluding remarks on issues such as best practice in the light of the preceding discussion.
6.2. Value transfer: Basic concepts and methods
6.2.1. Defining value transfer
Value or benefit transfer (VT or BT) concepts have been advanced in a number of articles over the past 25 years or so. Early developments include the pioneering contributions in the 1992 issue of Water Resources Research (Vol. 28, No. 3), which was dedicated specifically to BT. A definition of BT offered in that volume was: “…the transfer of existing estimates of non-market values to a new study which is different from the study for which the values were originally estimated” (Boyle and Bergstrom, 1992). Since then the number and quality of VT and BT studies have increased significantly. Another milestone was Desvousges, Johnson and Banzhaf (1998), one of the first major published studies of the validity of BT. That volume distinguished two basic definitions of BT, which still largely apply now.
The first definition is a broader concept based on the use of existing information designed for one specific context (original context) to address policy questions in another context (transfer context). These types of transfer studies are not limited to cost-benefit analysis (CBA) and related applications. They occur whenever analysts draw on past studies to predict effects of policies in another context. Put this way, value transfer – in some shape or form – is far more pervasive to policy analysis than many perhaps would fully realise.
The second definition is a narrower concept based on the use of values of a good estimated in one site (the “study site”) as a proxy for values of the (same) good in another site (the “policy site”). This is the type of VT most commonly used in CBA and thus it is this more specific definition that is the basis of this Chapter.
The application of this latter type of value transfer covers a remarkably wide range of goods. For example, the provision of a non-market good at a policy site might refer to a river at a particular geographical location (where study sites relate to rivers at different locations). However, relevant impacts at a site might also entail some change in a human health state change. A policy-site also might be a wholly different country to that where the study was originally conducted. That is, perhaps values are being transferred from countries, which are data-rich (i.e. the minority) to countries where there is a paucity of such information (i.e. the majority).
6.2.2. Transfer methods
An important point is that value transfer is not necessarily a passive or straightforward choice for analysts. Once value transfer has been selected as the assessment method (itself a choice requiring some reflection), then judgement and insight is required for all of the basic steps entailed in undertaking a VT exercise. For example, information needs to be obtained on baseline environmental quality and changes as well as relevant socio-economic data. In addition, original studies for transfer need to be identified. Published and unpublished (e.g. so-called “grey”) literature might be sought in this regard. It may be, however, that a database of past studies exists in which case consulting this source would seem an appropriate starting point. Later on, this Chapter describes efforts to construct databases of environmental valuation studies (see Section 6.5).
In general rule a transfer can be no more reliable than the original estimates upon which it is based. Given a lack of good quality original studies for many types of non-market values and the fact that even good studies typically have not been designed specifically for transfer applications, care must be taken here. Clearly, the analyst needs to have some criteria for judging the quality of studies if no “official” (or other) guidance exists.
Perhaps the most crucial stage is where existing estimates or models are selected and estimated effects are obtained for the policy site (e.g. per household benefits). This is the point at which the actual transfer occurs and implies choosing a particular transfer approach (see below). In addition, the population at the relevant policy site must be determined. Aggregation is achieved by multiplying per household values by population, the choice of which itself requires careful consideration.
One example of the problems of deciding the population over which to aggregate was the use of VT in the United Kingdom to guide decisions about withholding abstraction licences to water companies on the basis of alleviating low flow problems in English waterways. One of these decisions was overturned on the basis of a judicial review, which determined that those households previously ascribed non-use value for one river (the Kennet) by a factor of 75.2 This is a now rather dated example, but at the time many viewed it this decision as a serious blow to CBA (or at least its use in environmental decision-making in the United Kingdom) (Pearce, 1998). With hindsight, those fears have proven to be overblown; however, such episodes should not be forgotten as a cautionary tale relevant for today.
There are at least three different types of adjustment of increasing sophistication for the analyst to choose from. These options are reviewed in what follows.
Unadjusted (or Naïve) value transfer
The procedure here is to “borrow” an estimate of WTP in context σ (the study site) and apply it to context P (the policy site). The estimate is usually left unadjusted:
WTPS = WTPP.
A variety of unit values may be transferred; the most typical being mean or median measures. Mean values are readily compatible with CBA studies as they allow simple transformation to aggregate benefit estimates: e.g. multiply mean – average – WTP by the relevant affected population to calculate aggregate benefits.
The virtue of this approach is clearly its simplicity and the ease with which it can be applied once suitable original studies have been identified. Of course, the flipside of this relative straightforwardness is that it fails to capture important differences between the characteristics of an original study site (or sites) and a new policy site. If these differences are significant determinants of WTP, then this transfer approach – which is sometimes more prescriptively known as a naïve transfer – will fail to reflect likely divergences in WTP at the study and policy sites.
Determinants of WTP that might differ between study and policy sites include:
-
The socio-economic and demographic characteristics of the relevant populations. This might include income, educational attainment and age.
-
The physical characteristics of the study and policy sites. This might include the environmental services that the good provides such as, in the case of a river, opportunities for recreation in general and angling in particular.
-
The proposed change in provision between the sites of the good to be valued. For example, the value of water quality improvements from studies involving small improvements may not apply to a policy involving a large change in quantity or quality (e.g. WTP and quantity may not have a straightforward linear relationship).
-
Differences in the “market” conditions applying to the sites. For example, variation in the availability of substitutes in the case of recreational resources such as rivers. Two otherwise identical rivers might be characterised by different levels of alternative recreational opportunities. Other things being equal (by assumption in this case), mean WTP to prevent a lowering of water quality at a river where there are few substitutes should be greater than WTP for avoiding the same quality loss at a river where there is an abundance of substitutes. The reason for this the former is a more scarce recreational resource than the latter.
-
Temporal changes. There may be changes in valuations over time, perhaps because of increasing incomes and/or decreasing availability of clean rivers.
As a general rule, there is little evidence that the conditions for accepting unadjusted value transfer hold in practice. Effectively, those conditions amount to saying that the various conditions listed above all do not hold, i.e. “sites” are effectively “identical” in all these characteristics (or that characteristics are not significant determinants of WTP, a conclusion which sits at odds with economic theory).
Value transfer with adjustment
A widely used formula for adjusted transfer is:
WTPP = WTPS (YP/YS)e,
where Y is income per capita, WTP is willingness-to-pay, and e is the income elasticity of WTP.3 This latter term is an estimate of how the WTP for the (non-market) good in question varies with changes in income). According to this expression, if e is assumed to be equal to one, then the ratio of WTP at sites S and P is equivalent to the ratio of per capita incomes at the two sites (i.e. WTPP/WTPS = YP/YS). In this example, values are simply adjusted upwards for projects affecting people with higher than average incomes and downwards for projects that affect people with lower than average incomes. As an example, Hamilton et al. (2014), based in turn on OECD (2014), transfer WTP for various health states (particularly mortality risks) using the ratio of incomes between two areas (and various assumptions about the income elasticity of WTP) in order to estimate the health burden of PM2.5 which is co‐produced by industrial processes along with carbon dioxide.
In the above commonly used adjustment, the only feature that is changed between the two sites is income per capita. The rationale for this is perhaps this is the most important factor determining in changes in WTP, as meta-studies such as OECD (2014) appear to find. Of course, to the extent that say income is not the sole determinant of WTP, then even this improvement may well fall short of approximating actual WTP at the study site. However, it is also possible to make a similar adjustment for, say, changes in age structure between the two sites, changes in population density, and so on. Making multiple changes of these kind amounts to transferring benefit functions and this last transfer approach is considered below.
Value function transfer
A more sophisticated approach is to transfer the benefit or value function from S and apply it to P. Thus, if it is known that WTP at the study site is a function of a range of physical features of the site and its use as well as the socio-economic (and demographic) characteristics of the population at the site, then this information itself can be used as part of the transfer. For example, if WTPS = f(A, B, C, Y) where A,B,C are additional and significant factors affecting WTP (in addition to Y) at site S, then WTPP can be estimated using the coefficients from this equation, but using the values of A, B, C, Y at site P: i.e.
WTPS = f(A, B, C, Y)
WTPS = a0 + a1A + a2B + a3C + a4Y,
where the terms ai refer to the coefficients which quantify the change in WTP as a result of a (marginal) change in that variable. For example, assume that WTP (simply) depends on the income, age and educational attainment of the population at the study site and that the analysts undertaking that study estimated the following relationship between WTP and these (explanatory) variables:
WTPS = 3 + 0.5YS - 0.3 AGES + 2.2 EDUCS
That is, WTPS increases with income and educational attainment but decreases with age as described. In this transfer approach, the entire benefit function would be transferred as follows:
Þ WTPP = 3 + 0.5YP - 0.3 AGEP + 2.2 EDUCP
As an example of the implications of this approach, if the population at the policy site is generally much older than that at the study site, then WTPP – other things being equal – will be lower than WTPS.
A still more ambitious approach is that of meta-analysis (e.g. Bateman et al., 2000). This is a statistical analysis of summary results of a (typically) large group of studies. The aim is to explain why different studies result in different mean (or median) estimates of WTP. At its simplest, a meta-analysis might take an average of existing estimates of WTP, provided the dispersion about the average is not found to be substantial, and use that average in policy site studies. Alternatively, average values might be weighted by the dispersion about the mean; the wider the dispersion, the lower the weight that an estimate would receive.
The results from past studies can also be analysed in such a way that variations in WTP found can be explained. This should enable better transfer of values since the analyst can learn about what WTP systematically depends on. In the meta-analysis case, whole functions are transferred rather than average values, but the functions do not come from a single study, but from collections of studies. As an illustration, assume that the following function is estimated using past valuation studies of wetland provision in a particular country:
WTP = a1 + a2 TYPE OF SITE + a3 SIZE OF CHANGE + a4 VISITORS + a5 NON-USERS + a6 INCOME + a7 ELICITATION FORMAT + a8 YEAR
This illustrative meta-analysis attempts to explain WTP with reference not only to the features of the wetland study sites (type, size of change in provision in the wetland as well as distinguishing between visitors and non-users) and socio-economic characteristics (income) but also process variables relating to the methods used in original studies (elicitation format in stated preference studies and so on) and the year in which the study was undertaken. Application of meta-analysis to the field of non-market valuation has expanded rapidly in recent years. Studies have taken place in respect of urban pollution, recreation, the ecological functions of wetlands, values of statistical life, noise and congestion.
Many commentators have concluded that, at least in theory, the more sophisticated the approach is the better, in terms of accuracy of the transfer. The rationale for this conclusion presumably being that there is little to commend VT if it is inaccurate and misleading. However, many have understandably also combined this aspiration for accuracy with some pragmatism about dismissing simplistic approaches altogether. On this view there is little to commend VT if it cannot be routinely applied. This latter point means that the appeal of VT is likely to be diminished if it is always and everywhere the preserve of the highly trained specialist. Meta studies such as OECD (2014) make clearer the situations in which simple approaches are justified and when they are not. However, tensions still exist as illustrated by the growing presence of sophisticated meta-functions for transferring values on the one hand and avowedly practical “look-up tables” (e.g. lists of average WTP values, and ranges, for ecosystem services) and valuation databases on the other.
6.3. How robust is value transfer?
Determining when value transfer is a robust procedure is clearly a crucial element in relying on it more and more for CBA. The responds to this challenge in VT studies broadly speaking has been two-fold. First, a growing number of studies has sought to ascertain the likely size of transfer errors and, importantly, understand when and where these errors are most likely to occur (as well be large). Second, actual practice has used these insights to improve transfers. The current section reviews the former development, while one illustration of the latter is explained in the section that follows.
A growing number of studies that have sought to test the validity of the value transfer. The basic idea underlying these validation tests is to carry out an original study at the policy site as well. The proposed value to be transferred can then be compared with the value that was obtained from the primary study. The overall merits of the transfer are clearly indicated by whether or not the transferred value and the primary estimate are similar judged on the basis of some (statistical or other) criterion or criteria.
The most prominent way of assessing this is with reference to convergent validity. That is, to what extent is there agreement or errors (a divergence or convergence) between WTP estimated at the study site and policy site? To measure how large this magnitude is – arising from a value transfer – each site in a VT test is, in turn, treated as the “target” or policy site of a transfer. That is, each is treated as the site for which a value estimate is needed. The transferred estimate is then compared with the own-study estimate for the target site, and the transfer error can be calculated as follows:
Brouwer et al. (2015) note that a virtue of discrete choice experiments (DCE) (see Chapter 5) is their valuation of marginal changes in individual attributes which comprise a policy change. In principle then this provides a solid basis for subsequent value transfers, especially where these attributes vary considerably between policy site and study site(s). To test this, these authors look at the transferability of values across countries. Specifically, the study covers Greece, Italy, Spain and Australia and uses a choice experiment (DCE). The focus is that all these countries are drought prone and the tests conducted look at the transferability of non-market values for water conservation. This refers to water as a good which, in turn, results in benefits enjoyed by domestic use of water by households as well as contributing to household well-being by enhancing ecosystems. The DCE attributes were: ecological status related to water flow; the probability of outdoor water use restrictions for households; and cost to a household in the form of its water bill.
A number of transfer approaches were conducted. This included transfers from single country to another single country (e.g. transferring values from Greece to Australia) as well as transfers of mean values from a pooled group of countries to a single country (e.g. transferring values from a pool consisting of Greece, Italy and Spain to Australia). Different statistical models to estimating attribute values were also used with an emphasis on using different models which could account for varying socioeconomic characteristics across these countries as well as preference heterogeneity in a relatively sophisticated way (i.e. a mixed logit model). As often seems to be the case with these tests, the results are both reassuring for pragmatists and disturbing for purists. The degree of transfer error is reduced considerably when pooling country data and adjusting for socioeconomic differences between policy and study site(s). However, unobserved preference heterogeneity is important too and that, by its very nature, cannot be so “easily” controlled for.
Kaul et al. (2013) provides a test of transfer errors using a relatively comprehensive meta-study of more than 30 past studies, comprising in total more than 1 000 estimates of transfer error (although mostly drawn from the United States and Europe). As a result their paper provides influential findings on what critical empirical insights can be gleaned given this stock of past studies. A number of findings emerge. The possible ranges of error are extremely large indeed. That is, for a typical study, the error can vary from just a few% to an order of magnitude of that amount (and sometimes even more). Controlling for extreme outliers, however, (which reduces the sample to 925 VT tests), the average transfer error is about 40%.
A number of further identifiable things also appear to contribute to differences in errors. More sophisticated approaches (based on benefit function transfers) outperformed simpler approaches (based on largely unadjusted value transfers) in terms of reducing the likely error range, although pooling estimates also helps reduce error. Geographical proximity between policy and study sites reduces transfer error. In addition, transfer errors are smaller for policies involving changes in environmental quantities than for those involving changes in environmental quality.
These findings are important in that, as the authors suggest, they help provide guidance as to when practitioners should be more cautious about using VT. This does not necessarily mean that VT should be avoided. It may be the only analytical option for valuing policy or project changes, after all. However, what may be appropriate is greater care and use of sensitivity analysis, and so on. Evaluating policy changes when environmental quality is the issue is a case in point.
In making sense of these findings, it still remains important to ask how much transfer error policy makers (or analysts) should be willing to expose themselves to in order to inform better policy advice. One interpretation is that whether these (and other) margins of error should be considered “large” or “too large” might depend on the use of the results. For some project and policy applications, it is probably acceptable for errors of the magnitude suggested in Figure 6.1. Indeed, Ready et al. (2004) argue that, as a practical matter, relative to other sources of uncertainty in a policy analysis, the scale of error that they find is probably acceptable. Any uncertainty of the final results can be dealt with through sensitivity analysis.
Source: Brookshire (1992).
There is a legitimate discussion to be had regarding how much accuracy is required. An early but valuable contribution to frame this thinking is Brookshire (1992). Figure 6.1 indicates that if the objective of a value transfer study is to gain more knowledge about some value at a policy site or provide an initial assessment of the value of policy options (i.e. scoping/screening), then it may be that a relatively low level of accuracy is acceptable. Once the analyst moves towards undertaking a transfer study to inform an actual policy decision or natural resource damage assessment compensation litigation, then a greater degree of accuracy is arguably desirable. In such cases, presumably, either compelling evidence for the validity of value transfer needs to exist or an original valuation study may be warranted.
6.4. Value transfer and spatial variability4
Tests of the validity of VT as well as meta-studies of those tests (such as Kaul et al., 2013) make clear that geographical similarity tends to reduce possible errors. In other words, transfers where this condition of “similarity” does not hold needs to be done with extra care. Critically, spatial variability needs to be considered when performing a value transfer. Some of the issues can be illustrated with regards to standard values on per hectare ecosystem services provided by broad habitat types (such as uplands, urban green space, and so on). The possible problems are several in naïvely estimating total value as the product of this representative unit value and (say the change of) total ecosystem area of a particular type.
One example is Barbier et al. (2008), which focuses on the possibly non-linear relationship between ecosystem extent and the functions and so services that it provides. Using the example of Thailand’s mangroves in attenuating wave damage from more commonly experienced storm events, spatial heterogeneity arises because proximity (of mangroves) to shorelines is a critical determinant of the degree to which this function is provided: that is, it diminishes the further the ecosystem is (inland) from the shore. Taking explicit account of this heterogeneity is needed as a more defensible basis for aggregation. This is also required for more accurate policy analysis. Put another way, what Barbier et al., show is the (estimated) marginal value of mangrove area in their study area in Thailand is declining. Clearly, taking account of such non-linearity is important for more robust transfers.
One of the largest ecosystem service value transfer exercises conducted to date involved the core of the economic analysis underpinning the UK National Ecosystem Assessment (UK-NEA, 2011). Value functions were estimated for multiple ecosystem services, including the provisioning value of agricultural food production, the regulating services of the environment as a store for greenhouse gases (GHGs) and the so-called cultural services of both rural and urban nature recreation. The approach taken followed Bateman et al. (2011), with value functions simplified to focus upon the main determinants of value, so as to provide a degree of generality to subsequent general. The functions were also constructed in a unified way linking each to the others. As an illustration, if provisioning values are increased as a result of agricultural intensification, this intensification also might translate into an increase in GHG emissions and deterioration of rural recreation opportunities.
Figure 6.2 illustrates findings from the UK-NEA analysis of rural recreation benefits arising from a change of land use from conventional farming towards multipurpose, open-access, woodland (see also Bateman et al., 2003). The distribution obtained by transferring a recreational value function across the entirety of Wales reflects various factors, including the distribution of population and the availability and quality of the road network. Such spatially disaggregated outputs allow decision makers to target resources in a more efficient manner. These advantages were quickly realised by UK policy makers and the lessons of the UK NEA were explicitly incorporated in the UK Natural Environment White Paper (Defra, 2011), published in the aftermath of the NEA report.
Source: Adapted from UK-NEA (2011).
As an example of these transfer exercise outputs, Bateman et al. (2011) estimate that, in the United Kingdom, ecosystem services help contribute to 3 billion outdoor recreational visits annually with the social value of the output created by these trips likely to be more than GBP 10 billion. Importantly, location (of these sites) matters a great deal and not surprisingly, the aggregate picture is only part of the story. A specific and moderate sized nature recreation site, for example, might generate values of between GBP 1 000 and GBP 65 000 per annum depending solely on where it is located. A critical determinant of this range is perhaps not surprisingly proximity to significant conurbations. Put another way, woodlands in the “wrong” place (i.e. relatively far from potential visiting populations) are unlikely to give rise to such high social values (other things being equal), an insight of particular importance if policy makers are contemplating new investments in these nature sites.
6.5. Value transfer databases and guidelines
Without a readily accessible stock of value studies any VT exercise may be hampered by the daunting task of collecting past studies. Even this assumes that there is an abundance of original studies in the first place waiting to be collated in this way. This assumption may be optimistic. Surveys of VT studies and practice, such as Johnston and Rosenberger (2010) and Johnston et al. (2015), typically make important points about problems here. This includes the geographical skew in studies (e.g. mostly from North America and Western Europe). It also includes observations about the nature of research endeavour in the space of environmental valuation, which typically prizes academic novelty (i.e. generating new knowledge) over generating more empirically replicable but high quality data. This is one example of where progress at the CBA frontier may not serving policy needs as effectively it might. While Johnston and Rosenberger (2010) rightly reprove the research community for this bias, the question as to whether policy makers have sufficiently incentivised researcher direction also seems important. Loomis (2015) notes emerging evidence of exceptions to this trend in damage assessment of oil spills along US coastlines.
It has been long claimed that it is necessary to establish databases of valuation studies which is accessible for the researcher who intends to conduct benefit transfer. Indeed, one practical example of this is long-established. International collaboration between Environment Canada, the US EPA and the UK Ministry with environmental responsibilities (DEFRA) has resulted in the development of a substantial library of benefit estimates: the EVRI system.
Value transfer databases and manuals, in general, are a welcome development in the literature, as those analysts who have spent time searching for values no doubt would testify. There are caveats of course. There is still the need for expert judgement and analysis in selecting and adjusting values. In principle, the database provides information on the likely quality of the studies, although how this evaluation might work in practice is less clear at this point in time. That the analyst’s job is made much easier and more defensible as the findings of previous valuation studies are systematically distilled and organised, this is a welcome addition to the VT “tool-kit”.
A variant of the VT database are look-up values: “official” non-market values for benefit categories that practitioners charged with the task of appraising policies and investment projects, on behalf of decision-makers, should use when the need arises. An example of this for Germany (specifically, the German Federal Environment Agency) is Schwermer et al. (2014). This contains a range of unit values some of which are illustrated in Table 6.1 for air pollutants.
While its relative ease of calculation means that the approach can be widely practised, many analysts might balk at the potential over-simplicity, without accompanying guidance on how VT should be done in a robust way. So much depends on how the data are being used as well as resulting summary values are based on an abundance of good quality evidence. While there are some variation in the table depending on emission source and whether emissions occur in urban and rural areas (especially for road transport), to the extent that values such as those in Table 6.1 simply are being “pulled off-the-shelf” and applied unadjusted then the questions that arise are what degree of accuracy is being sacrificed. The general lesson is that benefit transfer database approaches are to be welcomed but it would be worthwhile allying these efforts to the establishment of widely agreed and authoritative protocols as to what is best practice with regards to using catalogued values. Schwermer et al. (2012), in a separate but accompanying document, provides information here, making the point for example that unit values such as those in the table, provide the basis to “… only permit a rough calculation of possible damage due to air pollutant emissions” (p. 22).
One conclusion of Rolfe et al. (2015) is the absence of such guidelines more generally, or at least the absence of general agreement on this. However, one example for the United Kingdom is eftec (2009), which provides the basis for conducting VT in official CBA. That is, it augments Defra guidance on valuing ecosystem services, which is in turn is an extension of the CBA guidelines published by HM Treasury (i.e. the so-called Green Book, HM Treasury, 2018). This advice sets out eight steps in all for conducting a VT for policy or investment project appraisal. Some of these steps involve general points about environmental CBA: for example, define the policy change, define the affected population at the outset of the analysis. Other steps are more specific to the VT task and involve asking a series of questions about the quality of primary studies to be used for the transfer as well as about the differences that might exist between study and policy sites. The emphasis is on practical demonstration on how these differences might be taken into account when conducting the transfer and how the sensitivity of findings might be tested.
6.6. Concluding remarks
Transfer studies are the bedrock of practical policy analysis in that only infrequently are policy analysts afforded the luxury of designing and implementing original studies. Thus, in such instances, analysts must fall back on the information that can be gleaned from past studies. Almost inevitably, VT introduces subjectivity and greater uncertainty into appraisals in that analysts must make a number of additional assumptions and judgements to those contained in original studies. Of course, this comment should be kept in context as the same could be said of almost any modelling exercise. The key question is whether the added subjectivity and uncertainty surrounding the transfer is acceptable and whether the transfer is still informative.
The discussion in this Chapter suggests that despite the central role played in public decision-making, transfer studies need to avoid employing simplistic methods for interpreting, summarising and integrating available information. The reason for this is the danger that there are likely to be significant trade-offs between simplicity and accuracy of the resulting transfer. Thus, a competent application of transfer methods demands informed judgement and expertise and sometimes, according to more demanding commentators, as advanced technical skills as those required for original research. Yet, the simplicity versus accuracy dilemma may only be part of the story given that a number of influential studies have cast doubt on whether more sophisticated approaches always yield more precise transfer values. Even so, it is unlikely – as well as undesirable – that reliable transfer exercises will ever be a purely mechanical procedure. Indeed, some experience shows that treating this process in this way can have risky implications for the wider regard in which cost-benefit approaches are held.
Certain conditions probably have to be met for a valid benefit transfer to take place. Surprisingly perhaps there are fewer generally accepted protocols in this regard (although see eftec, 2009 as one example here). However, there are a number of widely cited pieces of the puzzle with regards to what might constitute best practice in benefit transfer.
The studies included in the analysis must themselves be sound. Initial but crucial steps of any transfer are very much a matter of carefully scrutinising the accuracy and quality of the original studies. This in itself requires considerable judgement although the consolidation of information in the developing EVRI database, along with any assessment of the quality of each study within the system, makes this particular task less problematic. There is a need for parallel efforts to establish (official) protocols for best practice in value transfer as regards the “correct” procedures for, say, selecting and adjusting study site values. It is only in this way can the value of databases be fully and sensibly realised.
In conducting a value transfer, the study and policy sites must be similar in terms of affected population and population characteristics. If not then differences in population, and their implications for WTP values, need to be taken into account. Just as importantly, the change in the provision of the good being valued at the two sites also should be similar. This particular consideration raises many issues including that of whether the context in which a good is being provided is an important determinant of WTP. At some level, dissimilarity is the norm (e.g. the unique ecosystem habitats or the spatial pattern of substitutes around a site are unique). However, it is the degree to which this dissimilarity affects values which is the crucial point.
Tests of benefit transfer essentially have attempted to evaluate whether apparently similar goods can actually be characterised as such in reality. One reading of the results of these tests is that the validity and accuracy of benefit transfer can be questioned. Another interpretation is that those tests themselves provide important guidance about in what situations value transfer can be carried out confidently and when practitioners should proceed with more caution and scrutiny.
References
Atkinson, G., I.J. Bateman and S. Mourato (2012), “Recent advances in the valuation of ecosystem services and biodiversity”, Oxford Review of Economic Policy, Vol. 28/1, pp. 22-47, https://doi.org/10.1093/oxrep/grs007.
Barbier, E.B. et al. (2008) “Coastal ecosystem-based management with nonlinear ecological functions and Values”, Science, Vol. 319, pp. 321-323, https://doi.org/10.1126/science.1150349.
Bateman, I.J. et al. (2011a), “Economic analysis for ecosystems assessments”, Environmental and Resource Economics, Vol. 48, pp. 177-218, https://doi.org/10.1007/s10640-010-9418-x.
Bateman, I.J. et al. (2011b), “Economic values from ecosystems”, in The UK National Ecosystem Assessment Technical Report, UK National Ecosystem Assessment, UNEP-WCMC, Cambridge, also available from http://uknea.unep-wcmc.org/.
Bateman, I.J. et al. (2011c), “Making benefit transfers work: Deriving and testing principles for value transfers for similar and dissimilar sites using a case study of the non-market benefits of water quality improvements across Europe”, Environmental and Resource Economics, Vol. 50/3, pp. 356-387, https://doi.org/10.1007/s10640-011-9476-8.
Bateman, I.J. et al. (2000), Benefits Transfer in Theory and Practice: A Review and Some New Studies, Centre for Social and Economic Research on the Global Environment (CSERGE) and School of Environmental Sciences, University of East Anglia, Norwich.
Boyle, K.J. and J.C. Bergstrom (1992), “Benefits transfer studies: Myths, pragmatism and idealism”, Water Resources Research, Vol. 28/3, pp. 657-663.
Brookshire, D.S. (1992), “Issues regarding benefits transfer”, Paper presented at the Association of Environmental and Resource Economists Workshop, Utah, June 1992.
Brouwer, R. et al. (2015), “Improving value transfer through socio-economic adjustments in a multi-country Choice Experiment of Water Conservation Alternatives”, Australian Journal of Agricultural and Resource Economics, Vol. 59, pp. 458-478, https://doi.org/10.1111/1467-8489.12099.
Defra (Department for Environment, Food and Rural Affairs) (2011), The Natural Choice: Securing the Value of Nature, Cm 8082, The Stationery Office, London, www.gov.uk/government/uploads/system/uploads/attachment_data/file/228842/8082.pdf.
Desvousges, W., F.R. Johnson H.S. and Banzaf (1998), Environmental Policy Analysis with Limited Information: Principles and Applications of the Transfer Method, Cheltenham: Edward Elgar.
Eftec (2009), Valuing Environmental Impacts: Practical Guidelines for the Use of Value Transfer in Policy and Project Appraisal, report for Defra, eftec Ltd., London, www.gov.uk/government/uploads/system/uploads/attachment_data/file/182376/vt-guidelines.pdf.
Hamilton, K. et al. (2014), Multiple Benefits from Climate Mitigation: Assessing the Evidence, report for New Climate Economy.
HM Treasury (2018), The Green Book: Central Government Guidance on Appraisal and Evaluation, HM Treasury, London, www.gov.uk/government/uploads/system/uploads/attachment_data/file/220541/green_book_complete.pdf.
Johnston, R. and R. Rosenberger (2010), “Methods, trends and controversies in contemporary benefit Transfer”, Journal of Economic Surveys, Vol. 24/3, pp. 479-510, https://doi.org/10.1111/j.1467-6419. 2009.00592.x.
Johnston, R. et al. (eds.) (2015), Benefit Transfer of Environmental and Resource Values: A Guide for Researchers and Practitioners, Springer, Dortrecht.
Kaul, S. et al. (2013) “What can we learn from benefit transfer errors? Evidence from 20 years of research on convergent validity”, Journal of Environmental Economics and Management, Vol. 66, pp. 90-104, https://doi.org/10.1016/j.jeem.2013.03.001.
Loomis, J. (2015), “The Use of Benefit Transfer in the US”, in Johnston, R. et al. (eds.) Benefit Transfer of Environmental and Resource Values: A Guide for Researchers and Practitioners, Springer, Dortrecht.
OECD (2014), The Cost of Air Pollution: Health Impacts of Road Transport, OECD Publishing, https://doi.org/10.1787/9789264210448-en.
Ready, R. et al. (2004), “Benefit transfer in Europe: How reliable are transfers between countries?”, Environmental and Resource Economics, Vol. 29, pp. 67-82, https://doi.org/10.1023/B:EARE.0000035441. 37039.8a.
Rolfe, J. et al. (2015), “Introduction: Benefit Transfer of Environmental and Resource Values”, in Johnston, R. et al. (eds.) Benefit Transfer of Environmental and Resource Values: A Guide for Researchers and Practitioners, Springer, Dortrecht.
Rosenberger, R.S. and J.B. Loomis (2003), “Benefits transfer”, in Champ, P.A., K.J. Boyle and T.C. Brown (eds.), A Primer on Nonmarket Valuation, Kluwer Academic Publishers, Dordrecht.
Roy, R. and N. Braathen (2017), “The Rising Cost of Ambient Air Pollution thus far in the 21st Century: Results from the BRIICS and the OECD Countries”, OECD Environment Working Papers, No. 124, OECD Publishing, Paris, https://doi.org/10.1787/d1b2b844-en.
Schwermer et al. (2014), Economic Valuation Methods: Annex B to “Economic Valuation of Environmental Damage – Methodological Convention 2.0 for Estimates of Environmental Costs, Unwelt Bundesamt/German Federal Environment Agency, Dessau.
Schwermer, S. et al. (2012), Economic Valuation Methods: Annex A to “Economic Valuation of Environmental Damage – Methodological Convention 2.0 for Estimates of Environmental Costs, Unwelt Bundesamt/German Federal Environment Agency, Dessau.
UK-NEA (UK National Ecosystem Assessment) (2011), The UK National Ecosystem Assessment: Technical Report. UNEP-WCMC, Cambridge, UK.
Notes
← 1. See www.evri.ca/en/splashify-splash.
← 2. From 7.5 million people to just 100 000.
← 3. This is the approach applied in e.g. OECD (2014) and in Roy and Braathen (2017).
← 4. This section is adapted from Atkinson et al. (2012).