| Front page | | Contents | | Previous | | Next |

Practical tools for value transfer in Denmark – guidelines and an example

4 A Practical Value Transfer Guide

• 4.1 Stock of Danish valuation studies
• 4.2 Practical guidelines

4.1 Stock of Danish valuation studies

Appendix A lists 17 Danish primary valuation studies on the priority environmental goods. None of them are older than 10 years, and the majority of the studies have been performed the last 5 years. Several new primary valuation studies will also be published in the next few years (see appendix B). This clearly shows the increased activity in the field of environmental valuation in Denmark. Table 1 shows how the available primary studies are distributed on the priority environmental goods, and the different valuation methods applied.

Table 1: Review of Danish Valuation studies on priority environmental goods (CV = Contingent Valuation, CE = Choice Experiment, CR = Contingent Ranking, TC = Travel Cost, HP = Hedonic Pricing). Since some primary studies value more than one environmental good and/or apply more than one valuation method the numbers do not add up to 17, which is total number of studies in Appendix A.

Table 1 shows that there are Danish valuation studies for all the priority environmental goods considered in this report. However, nearly half of these valuation studies are on forests, and particularly the recreational value of forests. For the other environmental goods listed there is only one study in each subcategory, which means that the basis for value transfer for these goods is very thin. It should also be noted that several valuation studies consider restoration (rather than preservation) projects, e.g. reforestation projects and restoration of wetlands.

Contingent Valuation and Choice Experiments dominate among the methods (8 applications of each mehod; sometimes in the same study), but also CR, HP and TC have been applied. Five HP studies have been performed, which is more than other European countries for this type of environmental goods.

Based on the above review of Danish valuation studies conducted to date, and the discussion on validity of spatial value transfer in chapter 2, the time does not seem ripe for establishing general unit values for the priority environmental goods. The empirical basis for setting such general values simply does not exist. More primary valuation studies are needed for all these environmental goods, with the possible exception of forest recreation. However, these values also seem to be too site specific to construct general unit values. However, value transfer is still defensible, but the value transfer practitioner should apply the practical guidelines outlined in chapter 4.2 below.

4.2 Practical guidelines

The guidelines below are based on the discussion in chapters 2 and 3, and the stepwise procedure outlined by Desvousges et al. (1998) (see chapter 3.3). A practical illustration on how to apply the guide is provided in chapter 5.

STEP 1 - Identify the change in the environmental good to be valued at policy site

(i)                 Type of environmental good

(ii)               Describe (expected) change in environmental quality

a) baseline level, b) magnitude and direction of change

  (gain vs.  loss; and c) prevention^[9] vs. restoration )

STEP 2 – Identify the affected population at the policy site

Desvousges et al. (1998) use this as the last step in their Value transfer guide. However, it is important to identify the size of the affected population at the policy site before we review the valuation literature and evaluate the relevance of selected studies. The transferred value should come from the same type of affected individuals in terms of spatial scale.

If we just want to establish the use value of some activity, the relevant, affected population is the recreationists. If we would like to estimate both use and non-use values, and the policy site is only of local importance (e.g. a small river or lake with many substitutes regionally), we should use only the population of the municipality. If there are few substitutes for the sites at the regional level (e.g. a forest area in Denmark), the population in several communities, or even the county population, should be used. If the good is of national importance, e.g. a national park, or the single site of a red-listed species in the country, the national population should be used.

For use values, the number of individual recreationists should be estimated (before and after the change), while for non-use values (or use and non-use values combined) the number of households should be the unit of aggregation at the relevant geographical scale (community, regional/county or national level).

STEP 3 - Conduct a literature review to identify relevant primary studies

Review first the primary Danish Valuation studies in Appendices A and B to see if there are studies in Denmark of the environmental good in question. Table 1 clearly shows that the probability of finding relevant Danish studies is highest for forest recreation while it is much lower for the other priority environmental goods. Since Hedonic Price (HP) functions could be potentially difficult to transfer, especially between countries, due to the fact that the results are strongly influenced by characteristics of the market (e.g. attractiveness and overall price level of the area), Danish HP studies should be used where applicable. Thus, relevant Danish HP studies do exist, and should be utilized, for soil quality, and selected landscape features. (The advantage of HP is that it is based on revealed actual behaviour in a market where the environmental good is incorporated, as opposed to CV, CE and CR which are stated preferences methods).

The next step is to search the EVRI, ENVALUE and ValueBase ^SWE databases to identify similar studies from the other Nordic countries. This recommendation is based on value transfer validity tests (chapter 2.4) showing that studies closer spatially tend to have lower transfer errors. Studies closest in time should be selected for the same reason. However, one should note that this evidence is not conclusive. If there are no or only very few primary Nordic studies of the environmental good in question, or the valued change in the quality of the environmental good is outside the range considered at the policy site, the same databases and other bibliographies (e.g. the UK valuation studies list) should be searched for relevant studies. Meta-analyses (including also North American studies) could also be consulted, bearing in mind the limitations for value transfer of meta analyses with a broad scope (i.e. too large variation in definition of the environmental good). Thus, in practice, only meta-analyses for well defined recreational activities seem to produce meta-functions without methodological variables dominating the WTP function (see chapter 2.2), and with meta-functions explaining a sufficiently large part of the variation in WTP.^[12]

Often, the databases do not have all the data needed for the relevance of the study site to be evaluated, and the full study report should be found.

STEP 4 – Assessing the relevance and quality of study site values for transfer

Here, the quality of the relevant valuation studies is assessed in terms of scientific soundness and richness of information. Desvousges et al. (1998) identify the following criteria for assessing the quality and relevance of candidate studies for transfer:

i) Scientific soundness - The transfer estimates are only as good as the methodology and assumptions employed in the original studies

-                 Sound data collection procedures (for Stated Preference surveys this means either personal interviews, or mail/internet surveys with high response rate (>50 %), and questionnaires based on results from focus groups and pre-tests to test wording and scenarios

-                 Sound empirical methodology (i.e. large sample size; adhere to “best practice”-guidelines guidelines for e.g. Stated Preference surveys - see chapter 3.3 and appendix J)

-                 Consistency with scientific or economic theory (e.g. links exist between endpoints of dose-response functions and the unit used for valuation, statistical techniques employed should be sound; and CV, CR, CE, HP and TC functions should include variables predicted from economic theory to influence valuation)

ii) Relevance - the original studies should be similar and applicable to the “new” context

-                 Magnitude of change in environmental quality should be similar

-                 Baseline level of environmental quality should be similar

-                 Affected eco-system services and environmental goods should be similar

-                 The affected sites should be similar when relevant (e.g. when assessing recreational values)

-                 Duration and timing of the impact should be similar

-                 Socio-economic characteristics of the affected population should be similar

-                 Property rights, culture, institutional setting should be similar

 iii) Richness in detail – the original studies should provide a detailed dataset and accompanying information

-                 Identify full specification of the original valuation equations, including precise definitions and units of measurements of all variables, as well as their mean values

-                 Explanation of how substitutes (and complementary) goods were treated

-                 Data on participation rates and extent of aggregation employed

-                 Provision of standard errors and other statistical measures of dispersion.

All three criteria and their components are equally important for assessing the relevance and quality of the study.

STEP 5 –  Select and summarize the data available from the study site(s)

For our priority environmental good we will, with the possible exception of forest recreation, at the most have only a couple of relevant Danish primary valuation study. Even when we extend the scope to the Nordic and European valuation studies, we would frequently have only a few relevant study sites to transfer from. Then, the selection of a “best” value estimate is not very difficult. The problem arises, when several relevant studies are available, as is the case for forest recreation in Denmark (see table 1). Although we could still try to select the “best study”, this approach would ignore potentially valuable information contained in the studies neglected. Several parallel approaches should be applied, and the results from these should be used to present a range of values:

i)                    Search the studies to provide low and high estimates, which can define a lower and upper bound for the transferred estimate, respectively.

Collect data on the mean estimate and standard error, and specific spatial transfer errors if available (if not use the general transfer errors of + 25 - 40 %; see chapter 2.4). Consult relevant meta-analyses (see the table below for some examples) to see if the scope of these is narrow enough to provide relevant information about the estimate to be transferred. The scope could be too wide to produce reliable estimates if the meta-analysis consists of studies which vary a lot in terms of methodology, and the environmental good considered.

Compare the magnitude of the value from the meta-analyses, when methodological parameters in the meta-function is set according to the best practice guidelines and a context corresponding to the policy site. Methodological variables in meta-analyses (of CV studies) that reflect best practice guidelines include survey mode (preferable in-person interviews or mail surveys with high response rates), studies not older than about 10 years; i.e. conducted after the NOAA panel guidelines to CV (Arrow et al. 1993) (year of study often used as a proxy variable for quality in some meta-analyses), similar as possible in magnitude and direction of change, substitutes, characteristics of the population; and a realistic and fair payment vehicle (not voluntary contribution without a provision point mechanism, and not payment vehicles that create a large degree of protest behaviour).

STEP 6 – Transfer value estimate from study site(s) to policy site

a) Determine the transfer unit

The recommended units of transfer for use and non-use values are:

i) Use value:    

For recreation: Consumer surplus per activity day^[13]

For other types of use, e.g. groundwater or surface water for drinking:

WTP/household/year

For recreation consumer surplus per year (or per visit) per visitor could also be used, but then the average number of activity days (or visits) per year should be the same at the study and policy sites.

ii) Non-use value: WTP/household/year^[14]

Total WTP per ton of pollutant (i.e. aggregated WTP across affected households divided by the total number of tons of a pollutant causing the change in environmental good we are valuing; determined through the DFA , figure 1) could also be used, but this assumes a constant marginal WTP, and the same size of the affected population at the study and policy sites.

The use of total WTP per ha of ecosystem or landscape type assumes both the same size of the affected population and that the value pr. ha. is constant. However, empirical evidence shows that WTP does not increase proportionally with the number of ha of ecosystems or landscape types, or distance to off-shore windmills in km (see e.g. Ladenburg et al (2005)). Since SP surveys clearly show that WTP per unit of area varies widely, we should caution against converting households´ stated mean WTP for a discrete change in environmental quality to marginal values like WTP pr km or ha per household. However, this unit is ”better” than total WTP per km or ha, because in the latter case one also has to assume similar population density at the policy and study sites).

b) Determine the transfer method for spatial transfer

If the policy site is considered to be very close to the study sites either in Denmark, the other Nordic countries or other European countries) in all respects, unit value transfer can be used. If we have got several equally suitable study sites to transfer from they should all be evaluated, and the transferred values calculated to form a value range.

For unit transfers between countries, differences in currency, income and cost of living between countries can be corrected for by using Purchase Power Parity (PPP) corrected exchange rates; see e.g. http://www.oecd.org/dataoecd/61/56/1876133.xls. Within a country we could also use unit value transfer with a correction for differences in income level, using equation (1) in chapter 2.2 and an income elasticity of WTP lower than 1. Based on the discussion in chapter 2.4, these estimates should be presented with error bounds of + 40 %. However, if the sites are very similar, or the primary study was designed with transfer to sites such as the policy site in mind, an error bound of  + 25 % could be used. If the study and policy sites are not quite close, unit transfer could still be used, but arguments for over- and underestimation in the transfer should be listed and the unit value should be presented with error bounds of + 100 % (based on the observed large variation in individual estimates observed in validity tests; see chapter 2.4).

Function transfer can be used if value functions have sufficient explanatory power^[15] and contain variables for which data is readily available at the policy site. Most often the ”best” model is based on variables where new surveys have to be conducted at the policy site to collect data. Then one could just as well perform a full-blown primary valuation study. If models are constructed based on variables for which there exist data at the study site, they very often have low explanatory power. In general, WTP functions based on Stated Preference surveys (especially Contingent Valuation) have much lower explanatory power than functions based on TC and HP studies. Thus, it could be more relevant to use function transfer transferring estimates from these Revealed Preference methods.^[16]

If relevant meta-analyses are identified (see previous step), estimates from these could also be used in a comparison of several transfer methods. Sensitivity analysis could be performed to see how much the transferred value estimate could vary. The constructed upper and lower values should be used to bound the transferred estimate. However, all meta-analyses to date seem to be dominated by the methodological choices of the primary studies they consider. Thus, until we get enough primary valuation studies using the same methodology, estimates from meta-analyses would be less reliable than unit value transfers (and value function transfers from a single study site).

To conclude, unit value transfer is recommended as the simplest and most transparent way of transfer both within and between countries. This transfer method has in general also been found to be just as reliable as the more complex procedures of value function transfers and meta-analysis.  This is mainly due to the low explanatory power of willingness-to-pay (WTP) functions of Stated Preference studies, and the fact that methodological choice, rather than the characteristics of the site and affected populations, has a large explanatory power in meta-analyses. Generally speaking, error bounds of + 25 - 40 % should be used if the study and policy sites are very similar (which we should strive for) . If there is less similarity between study and policy sites, error bounds of + 100 % should be used.

c) Determine the transfer method for temporal transfer

The value estimate should be adjusted from the time of data collection to current e.g. 2005-DKK using the Consumer Price Index (CPI) for Denmark (see chapter 2.10, and CPI for Denmark and equation to be used for the conversion in appendix H). If we transfer values from a study site outside Denmark, we first convert to DKK, in the year of data collection, using PPP corrected exchange rates in the year of data collection, and then use the Danish CPI to update to current-DKK.

However, environmental goods could also increase more or less in value than the

goods the CPI is based on. However, there is no general rule for adjustments of preferences for environmental goods over time.

STEP  7 - Calculating total benefits or costs

For non-use values, mean WTP/household/year is multiplied by the total number of affected households to derive the annual benefit or cost. If WTP at the study site is stated as annual WTP for e.g. 5 or 10 years, the total benefits or costs should be calculated as the Present Value (PV) over that same period. On the other hand, if WTP is stated as one-time amounts the amounts must be viewed as a present value (of all benefits from the environmental good in question).

The general equation for calculating the present value of the benefits PV (B) is:

                                                   (6)

where B_t is the total benefits in year t, T is the time horizon  (for the stated WTP amounts) and r is the social discount rate (r = 0.03 (3% p.a.) is the social discount rate currently used by Miljøstyrelsen.  Benefits and the discount rate are stated in real terms, i.e. 2005 DKK and the discount rate is a real rate of return (i.e. corrected for inflation, and not a nominal rate)).

If the time horizon is not stated in the WTP question in SP surveys, we must assume that this is an annual payment over an infinite time horizon, i.e. t → ∞ . In this case, and if the annual benefits B_t are the same each year, equation (6) can be simplified to:

PV (B)  =  B_t / r                                                                               (7)

Annual benefits B_t are equal to aggregated WTP over the affected population (WTP_tot), which can be calculated as:

WTP_tot = n x WTP_{i                                                                                                                                                                (8)}

where n = number of affected households, and WTP_i =  mean Willingness – To –Pay for household i. Since WTP per household varies between different parts of the affected population (e.g. with distance from the site, whether users and/or non-users are considered etc.), the estimates from the study site(s) should be based on the same type of affected population as at the policy site. If this is not possible, distance decay in WTP (e.g. percentage reduction in WTP pr km increased distance from the environmental good) could be assumed, based on empirical evidence from relevant study sites (if such evidence does exist and suggests this).

If we calculate use values, we just substitute households with individual recreationists in equation (8) and use estimates for Consumer surplus per activity day times the increase or decrease in number of activity days to calculate total use value of the project. For uses other than recreation, e.g. use of groundwater and surface water for drinking, values are often elicited on a household basis, and the same procedure as for non-use values can be employed.

When aggregating damages and costs of environmental goods, we also need to consider whether these goods are independent (meaning we can just add them up), or substitutes or complementaries. In the first case we would overestimate aggregated damage or benefits, while in the latter case we would underestimate.

Finally, when performing a Cost-benefit analysis of a new project of policy, the estimated PV of benefits (costs) should be compared with the corresponding PV of costs (benefits). The effect on total annual benefits (costs) due to an expected general transfer error of 25 - 40 % (see chapter 2.4) should be calculated in order to see if this reduces the PV of benefits (increases the costs) to a critical level, i.e. the PV of net benefits becomes negative (from positive). If this is so, the transfer errors are large enough to change the outcome of our CBA, and we should try to increase the accuracy of the transferred estimate (either by conducting a full primary study or calibrating the transferred value by conducting a small scale primary study).

When there is a need for estimates of environmental goods for policy purposes, a CBA of conducting a new environmental valuation study should be performed in order to determine whether the costs of a new primary study is worth the benefits in terms of lower probability of making the wrong decision. These decision rules could be used as a rough test of whether value transfer has acceptable transfer errors, e.g. in its most frequent policy use (Cost-benefit analysis).

Footnotes

^[9] A distinction should be made between prevention (which preserves the original/undisturbed environmental good) and restoration. Hasler et al.  2002 (see appendix A) find that people put a higher value on keeping the original (i.e. prevention) than restoration (i.e. in this case purification of polluted ground water).

^[10] See Chapter 2.2.

^[11] This general classification of biodiversity stems from a CE of restoring Åmose marshland  (Lundhede et al. 2005). CE and CR often use very general descriptions of the environmental good. These results could provide some general values for this aspect of the good in question, and could be used in transfers to similar type of ecosystems in need of an estimate for a similar qualitative increase in biodiversity (preferably from the same reference level). However, these values should not be used to characterize or value biodiversity of other types of ecosystems nor a similar type ecosystem where we need to value a very detailed and quantitatively specified change in biodiversity.

^[12] Roughly said to be a higher adjusted R² than 0.5, i.e. explaining more than 50 % of the variation in value.

^[13] An activity day is defined as one individual performing recreation for a shorter or longer period during one day.

^[14] Some studies of use and non-use values have asked for individual WTP (e.g. Lundhede et al. 2005 for biodiversity of a marsh area). However, we view the household as the smallest “economic” unit for none-use values of these priority environmental goods. Multiplying individual WTP with the mean number of adults per household would tend to overestimate household WTP. Therefore, we have conservatively assumed that the reported individual WTP is equivalent to household WTP.

^[15] Roughly said to be having a higher adjusted R² than 0.5, i.e. explaining more than 50 % of the variation in value

^[16] This does, however, not mean that we should concentrate on RP studies when we perform new primary studies, as only SP methods are capable of valuing non-use values and future changes in environmental quality.

| Front page | | Contents | | Previous | | Next | | Top |

Version 1.0 December 2007, © Danish Environmental Protection Agency