Valuation 11: Benefit Transfer and Meta-Analysis

1
Valuation 11 Benefit Transfer and Meta-Analysis

• Why benefit transfer?
• Different types of benefit transfer
• Value transfer
• Function transfer
• Validity and sources of error
• Examples

2
Last weeks we looked at

• Various methods to estimate the value of
  environmental goods and services not traded on
  markets
• Also, some empirical examples were shown
• It turned out that it is actually very hard to
  reliably estimate prices one needs many
  assumptions, good data, and smart statistics, and
  even then all sorts of things may go wrong

3
What is benefit transfer?

• Benefit transfer uses economic information
  captured at one place and time to make inferences
  about the economic value of environmental goods
  and services at another place and time
• Benefit transfer involves economic values that
  may be either positive (benefits) or negative
  (costs)

4
Why benefit transfer?

• Valuation is hard
• As a result, applied valuation studies are
  expensive
• A small hedonic pricing study, for instance,
  costs about a year of a PhD student that is,
  after the data have been collected and digitised
• Ditto for travel costs
• A contingent valuation study is more expensive
• Monetary values are also hard needed
• Wouldn't it, therefore, be nice if we could take
  the estimated values of case 1 and apply them to
  case 2?

5
History

• Environmental benefit transfer came into being
  only once the non-market literature itself grew
  large enough to allow comprehensive synthesis and
  cross-study comparison
• In 1973, the U.S. Water Resources Council began
  publishing unit day estimates for recreation
  activities relates to water projects
• In 1980, the U.S. Forest Service began publishing
  Resources Planning Act values for recreation (per
  person per day)
• First synthesis study mid to late 1980s
• Today, applied to issues involving values for
  recreation, water quality, fishing, air quality,
  wetlands, biodiversity

6
Types of BT
Value Transfer
Function Transfer
Meta-Analysis Function
Single point transfer
Measure of central tendency
Single point transfer adm. approved
Benefit/ Demand Function
Adapt function to policy site
Use estimate at policy site
Use tailored estimate at policy site
7
The basis for any benefit transfer analysis

• The basis is a formal literature review
• The original research studies are screened for
• relevance
• how well they correspond to the policy context
• the quality of research
• what kind of information is provided
• Advantages include a much larger sample of data,
  different analytical techniques, and different
  analysts
• The main disadvantage is that one typically only
  has access to published results, which are always
  incomplete

8
Value transfer Point estimates

• Point estimate transfer typically uses a single
  measure
• Example The U.S. Army Corps of Engineers was
  considering the removal of four dams on the Lower
  Snake River from its confluence with the Columbia
  River
• One of the benefits would be the restoration of
  spawning habitat for native salmon populations
  (by about 47,500 fish)
• The agency was interested in the passive use
  value based on per household annual WTP
• The literature search found four studies that
  provided values for salmon population

9
(No Transcript)
10
Other value transfer

• The measure of central tendency entails using a
  mean, median or other measures of central
  tendency based on all or a subset of original
  research outcomes
• Same example as above great disparity between
  average marginal passive use value (40,620) and
  median value (3,197)
• Administratively approved estimate transfer is
  the simplest approach
• Estimates are derived from empirical evidence in
  the literature, expert judgment, and political
  screening
• Problems Criteria for political screening are
  unknown, selection might be biased and only
  updated every so often

11
Function transfer

• Value transfer requires a strong similarity
  between the study site and the policy site is
  required
• Function transfers are generally considered to
  perform better than value transfers
• Unlike value transfers, they may be tailored to
  fit some of the characteristics of the policy
  site
• Function transfers entails the application of a
  statistical function that relates the summary
  statistics of original research to the specifics
  of the study site
• There are two types of function transfers
• Demand/Benefit function transfer
• Meta-regression analysis

12
Demand/benefit function transfer

• For benefit function transfer additional
  information is required
• We need a function that models the statistical
  relationship between the summary measures of
  interest and characteristics of the original site
• This function is then adjusted to specific
  characteristics of the policy site

13
Example

• Case study by VandenBerg, Poe and Powell (2001)
  estimates the benefits of improving groundwater
  quality used for drinking to a very safe level in
  12 towns in New York, Massachusetts, and
  Pennsylvania
• The authors use a contingent valuation survey
  with a payment card question format
• To perform the benefit function transfer all of
  the survey data except for one town are pooled
  and a WTP equation is estimated

14
OLS regression, dependent variable is mean
WTP/HH/year, n667
15
Validity of benefit transfer

• A number of studies performed the following test
  Estimate the value of something at two sites, and
  predict the value of the one site from the
  observations of the other
• Resources or activities include
• Sport fishing (distance, harvest, quality)
• Recreation (costs, size, substitutes, population,
  age)
• Rafting (flow, costs, intensity, reason for
  visit, home, income, sex, age, education)
• Water quality (costs, size, depth, accessibility,
  quality, use, income)
• Water quality (bid, use, education, age, user)

16
(No Transcript)
17
Sources of error

• Generalisation error
• When estimates from study sites are adapted to
  represent different policy sites
• The error is inversely related to the degree of
  correspondence between the sites
• Measurement error
• Occurs when researchers decision affect the
  accuracy of the transferability of values
• Methodological choices, non-reporting of study
  characteristics
• Publication selection error
• There is a preference for publishing
  statistically significant results that conform
  theoretical expectations
• Authorship effect

18
Meta-regression analysis

• A drawback of the demand function transfer is the
  assumption that the statistical relationship
  between the dependent and independent variables
  are the same for both study as well as policy
  site
• Also, demand function transfer relies mostly on a
  single study
• Meta-regression analysis summarises and syntheses
  outcomes from several studies
• Meta-analysis not only provides a rigorous
  synthesis of the literature, it is also able to
  explain the variation found across empirical
  studies and to identify outlier studies,
  knowledge gaps, and priors for further analysis
• Meta-analysis is a technique that originates in
  medical science

19
Meta-regression analysis (2)

• Some of the variation in value estimates may be
  due to identifiable characteristics among the
  different studies themselves
• valuation method, survey mode, geographic
  location, time etc.
• These characteristics are not explanatory
  variables in the original studies as they are
  constant
• Advantage of the method it does not prejudge
  research findings on the basis of the original
  studys quality (input) , while it avoids a
  differential subjective weighting of studies in
  the interpretation of findings (output side)
• However, the publication selection error still
  applies

20
Example Meta-regression analysis

• Brouwer, Langford, Bateman and Turner (1999) is
  one of the early applications of meta-regression
  analysis
• Analysis covers 30 studies of the WTP per person
  for wetland preservation in temperate climate
  zones in developed economies in North America and
  Europe
• In total, 103 data points as some studies
  reported more than one observation (split sample)
• Wetlands were made comparable by looking at their
  types and functions flood control, water supply,
  water purification, and nature/recreation
• Each observation was associated with one or more
  functions

21
Wetlands/Brouwer

• Additional explanatory variables included
• payment vehicle (tax or other)
• elicitation format (open-ended or other)
• value type (use value, non-use value or both)
• country (USA or Canada, Europe)
• Quality was measured by response rate, but as an
  explanatory variable
• All studies are CVM studies
• In the regression analysis wetland size and
  income were excluded due to data availability

22
Regression results
GLS model specification
Parameter Estimate Standard error
Payment vehicle (tax) 1.576 0.362
Elicitation (open-ended) -0.376 0.183
Country (North America) 1.629 0.363
Response rate 30-50 -1.722 0.451
Response rate gt50 -1.461 0.450
Flood control 1.134 0.456
Water supply 0.441 0.479
Purification 0.659 0.327
significant at 0.001 significant at 0.01
significant at 0.05 R2 0.38
23
Wetlands/Brouwer (2)

• On average, people are willing to pay
  93/person/year for wetland preservation
• Note that the median is only 51
• Taxes attract higher contributions
• Open ended questions lead to smaller answers
• North Americans are willing to pay more
• Higher response rates imply lower values
• Less differences between functionality of wetlands

24
Conclusion

• Benefit transfer can perform no better than the
  quality of original studies
• The underlying questions of accuracy and
  appropriateness of non-market methods are not
  solved in benefit transfer
• However, some type of benefit estimates are
  subject to less controversy
• Benefit transfers are defensible as long as they
  are based on organised research agenda and seek
  to expand knowledge
• Theres a great deal of pragmatism in
  policy-decision making not all decisions
  require the same level of accuracy