Marianne Zandersen

1
ERE4 Valuation (11) Benefit Transfer

• Marianne Zandersen
• FNU

2
Valuation 11 Benefit Transfer

• Background What is Benefit Transfer (BT)? What
  values are transferred? What methods are behind
  the data transferred?
• Use When are BT applied? Why perform BT at all?
• Requirements What is needed for a good BT?
• Categories of BT and examples
• Are BTs valid?

3
Definitions
transfers existing estimates of non-market
values to a new policy site which is different
from the study site for which the values were
originally estimated - simply the application of
secondary data to a new policy site (Boyle
Bergstrom, 1992)
The river where an existing study was conducted
is termed the study site and the river under
consideration for water quality improvements the
policy site - the estimated benefits are
transferred from the study site to the policy
site (Desvouges et al. 1992)
4
Definitions

• A policy-makers choice between a new assessment
  study or an extrapolation of benefit assessments
  done elsewhere
• A limit-setting process where benefit transfer
  infer the plausible range of values and limits on
  willingness to pay for a site with no benefit
  assessment
• No forecast or prediction but measures the
  effects of a predefined set of conditions being
  changed
• The application of monetary values obtained for a
  specific non-market goods analysis to an
  alternative or secondary policy decision setting
• The application of a data set that was developed
  for one particular use to a quite distinct
  alternative application

5
Coverage
Total Economic Value of Nature Goods Services
Direct Use Values
Indirect Use Values
Option Values
Bequest Values
Existence Values
6
Types of BT
Value Transfer
Function Transfer
Single point transfer
Measure of central tendency
Single point transfer adm. approved
Benefit/ Demand Function
Meta-Analysis Function
Adapt function to policy site
Use estimate at policy site
Use tailored estimate at policy site
7
Coverage

• Data used in BT are based on following methods
• Travel cost method (numerous BTs)
• Hedonic price method (numerous BTs)
• Random utility method (only a few of BTs to date)
• Contingent valuation and choice experiments
  (fairly new in BT)
• Note that
• BT 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 BT
• Why conduct BT when we are not even sure about
  the original study methods?

8
Why Benefit Transfer?

• Some types of benefit estimates are subject to
  less controversy than others
• 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
• Monetary values are needed in policy-making,
  project assessment, litigation cases, natural
  resource damage assessments..
• ..applied valuation studies are expensive and
  time consuming...
• ...but funding limitations often preclude an
  original study..
• and timely benefit estimates are often crucial in
  policy making.

9
Why Benefit Transfer?

• 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 needed in policy-making,
  project assessment, litigation cases..

10
Use of BT

• Cost Benefit Analysis of new projects and policy
  initiatives (e.g. value of environmental features
  of farmland, UK SEAs EIAs)
• Environmental Regulations (e.g. Water Framework
  Directive WATECO)
• Calculation of compensation payments in pollution
  accidents (e.g. oil spill cases, hazardous waste
  CERCLA)
• General Equilibrium Models (e.g. climate policy
  benefit estimates, ECOBICE)

11
Use of BT

• Analysis of the impact of environmental
  regulations on benefits and costs
• Litigation cases from natural resources damages
  (CERCLA) allows government to sue for the
  damages to public national resources from
  disposition of toxic contaminants and oil spills
  for type A events (too small to warrant
  full-scale studies of their own)
• Project development and improvements (marine
  sportfishing in Alaska Department of Wildlife,
  economic benefits of state programs to preserve
  farmland)

12
Types of BT

• Single point transfer adm. approved Based on
  Expert Opinion/Judgement.
• Unit-day value approach is one example.
• Used by the U.S. Water Resources Council from
  the 60s to the 90s, initially because of a lack
  of existing studies
• Some measures of value have been consulted, but
  fundamentally, the approach is based on
  reasonableness
• Expert judgement was used to develop an
  approximation of the average WTP for recreation
  activities
• An estimate, adjusted for characteristics of
  the study site was selected from a range of
  updated values approuved by federal guidelines
• No scientific basis for the adjustment.

13
Types of BT

• Measurement by Proxy is another example single
  point transfers
• researchers use alternative values from related
  goods, often used in courtrooms in the U.S.
• EXAMPLES
• The proxy for the value of deer used were
  beef-prices (Lac Courte Oreilles Band of Lake
  Superior Chipewa Indians versus State of
  Wisconsin, 1988)
• The proxy for valuing natural resource damage
  was based on a ranking of different types of
  effects through the use of a weighting scheme
  e.g. for unquantifiable oil spills (Leschine
  Rubin, 1989)

14
Types of BT

• Researchers figured out in the 1960s that
  transferring the entire demand equation is a more
  conceptually sound approach than just
  transferring the result of the demand equation
  the average WTP
• Transferring the demand equation allows for
  adjustment to
• socio-economic characteristics of the
  population relevant to the policy site (income,
  sex, education, population density, tastes)
• characteristics of the policy site (quality,
  quantity, distance etc.)
• characteristics of substitute sites (quality,
  quantity, distance etc.)

15
Criteria

• Under the assumption that valid BT should produce
  no sign. Errors
• The change and level of non-market commodities
  need to be the same
• Population characteristics should be similar for
  the policy and study sites
• You can not switch welfare from WTP to WTA
• Additional recommendations
• Collect primary data at the policy site to update
  the benefit transfer. If comparisons are
  statistically similar, then benefit transfer is
  more likely to be valid
• This means conceptually that validity is ensured
  only if the functional form of the indirect
  utility function in two populations is the same,
  if vectors of prices and vectors of environmental
  quality are identical

16
Criteria

• Null-hypothesis of Equality
• Environmental goods are identical
  transfer errors should be zero
• H0 equality estimates of original study results
  estimates of transfer
• HAequality estimates of original study results
  estimates of transfer
• rejection of a null-hypothesis is interpreted as
  evidence against the validity of BT
• Requires that policy and study site models and
  results are be statistically equal
• Most classical test of BT validity
• Found by some to be a necessary but unsufficient
  condition
• Found by yet others to be erroneous

17
Criteria
Null-hypothesis of Equivalence

• Environmental goods are heterogenous
  transfer errors are inevitable, i.e. as we assume
  differences we are careful to state the opposite.
  We do this by testing that the differences are
  not larger than a predetermined level of
  difference
• Rejection of H0 hypothesis is interpreted as
  evidence in favour of the validity of BT
• Instead of assuming that policy and study site
  models and results are statistically equal, we
  should define an acceptable transfer error prior
  to the validity test and test it against the
  error level.

18
Differences between equality and equivalence test
Criteria
H0 Accepting identity
HA Reject non-equivalence
The classical test tends to accept transfers with
larger variance and standard errors of variance
but rejects transfers with smaller variance and
standard errors
Source Kristofersson and Navrud, 2005
19
Criteria
Thought Example of Equivalence versus equality
Tests

• An original estimate of WTP is
  100USD/household/year
• The benefit transfer estimate is
  110USD/household/year.
• Assume s.e. of the difference is
  4USD/household/year transfer error is 10.
• Under a classical test of equality, the transfer
  result would be rejected as valid at the 5 level
  of significance (point A)
• Under an equivalence test where 20 error margin
  is used, the transfer result would be accepted at
  the 5 level
• If, however, the s.e. is 8USD/household/year, the
  results are reversed (point B)

20
Criteria

• Some examples of validity tests
• Equal distribution test (Kolmogorov-Smirnov)
• Equality of means in the sample non-parametric
  test (Mann-Whitney)
• Equality of regression coefficients, sum of
  squared residuals (Wald)
• Equality of overall models, pooled and individual
  sample (Log-likelihood ratio)
• Equality of overall models (Log-likelihood ratio)

21
Issues around BT

• WHAT CAN GO WRONG?
• demand estimation misspecified model f(.),
  poor measurements of variables and data gathering
  problems
• benefit estimation aggregation errors,
  incorrect estimation of a population, incorrect
  structure for a demand error process (e.g.
  incorrect assumptions, incorrect site
  definitions, incorrect choice frameworks,
  counting trips for multiple purposes)
• Transfer Process uncertainty (use
  researchers judgement, not only hypothesis
  testing), use of transfer tests not robust
  enough, not enough variance in the population
  sample used for transfer

22
Examples

• Six studies performed the following test
  Estimate the value of something at two sites,
  predict the value of the one site from the
  observations of the other and compare
• Sport fishing 5-40 (distance, harvest, quality)
• Water quality 1-75 (costs, size, depth,
  accessibility, quality, use, income)
• Recreation 1-475 (costs, size, substitutes,
  population, age)
• Water quality 18-41 (bid, use, education, age,
  user)
• Fishing 1-34 (bid, period)
• Rafting 6-228 (flow, costs, intensity, reason
  for visit, home, income, sex, age, education)

23
Examples
Dutch Wetlands -1

• Brouwer Spaninks (1999) benefit transfer of
  meadow land in the Netherlands looking at
  agricultural wildlife management to measure use
  and non-use values of varying attributes in a
  spatial area
• Study 1 (1993) 500ha of peat meadow area in the
  province of Friesland (N. Holland) monthly and
  annual WTP 3 variations of questionnaires use
  of b/w descriptions 1 question 36 protest
  bidders locals and further afield 30 response
  
• Study 2 (1994) 15.660ha in South Holland annual
  WTP identical questionnaires, use of colour
  descriptions 3 questions locals and much
  further afield 30 response 32 protest
  bidders

24
Dutch Wetlands - 2
Examples
Study 1A is an outlier note the temporal
embedding effect if asked per month, people are
willing to pay more per year
Source Brouwer Spaninks, 1999
25
Dutch Wetlands - 3
Examples

• Study 1 did not show significant distance decay
  study 2 did.
• The results were regressed on sex, education,
  household composition, income, attitude,
  knowledge, and membership.
• Four Function transfers were carried out and
  tested for validity
• Study 1(all) Study 2
• Study 2 Study 1 (all)
• Study 1 (yearly) Study 2
• Study 2 Study 1 (yearly)
• Two Value transfers were made and tested for
  validity
• Study 1(all) Study 2
• Study 1 (yearly) Study 2

26
Dutch Wetlands - 4
Examples

• Only valid function transfer
• Only valid value transfer
• Point transfers appear to yield better results
• Median transfers cause significantly larger
  errors

27
Dutch Wetlands - 5
Examples

• Test of equality of benefit functions
• H0 is rejected for 3 out of four transfers (using
  Wald test)
• Transfer from sample 2 to 1(yearly) is accepted
  as valid (error-22.4)
• Test equality between individual and pooled
  sample coefficients
• H0 cannot be rejected in all 4 tests (using Wald
  test) ? equality of coefficients of individual
  and pooled models does not guarantee equality of
  coefficients of individual models
• Test of goodness of fit (comparison of sum of
  squared residuals)
• H0 is rejected in both cases (using LLR test) ?
  unlikely that the goodness of fit of the demand
  functions at the 2 sites is the same

28
Dutch Wetlands - 6
Examples

• Test of equality of means in the samples (Value
  Transfer)
• H0 is rejected for 3 out of four transfers (using
  Wald test)
• Transfer from sample 1(yearly) to 2 is accepted
  as valid (error -27)
• Test of equivalence of means in the samples
  (Value Transfer)
• The equivalence test cannot reject difference
  between the sample 1(yearly) and sample 2, even
  when testing based on a transfer error of 50 ?
  sample 1(yearly) is very small and high s.e. of
  mean WTP
• Test of normal distribution of stated WTP
  amounts
• All three samples were rejected as having Normal
  distributions (Kolmogorov-Smirnov test)
• Test of equal distributions (Kolmogorov-Smirnov
  test)
• H0 is not rejected for samples 1(yearly) and 2
• H0 is rejected for samples 1(yealy and monthly)
  and 2

29
Costanzas Transfer - 1
Examples

• An attempt to estimate the total current economic
  value of ecosystem services, using
• Literature survey and selection of non-market
  valuation of eco-system goods and services (99
  studies), ranging from
• 17 defined services and goods for different
  biomes
• Quantification of the total ha per biome
• conversion of sample values to ha/year by biome
  and service followed by a simple extention to
  total ha per biome
• Obtain results of average USD33trillion per year
  of total yearly ecosystem services to human
  society

30
Costanzas Transfer - 2
31
Costanzas Transfer - 3
Examples

• Main problems with Costanzas transfer
• Conversion of per ha values to marginal values
• Budget constraint ignored
• Economic value of infinity
• Gross simplification of transferable
  environmental goods, e.g. value of grassland in
  Wisconsin transferred to global grassland area
• Gross simplification of values to population in
  policy sites
• Jeopardises the field of benefit transfers

32
Meta-Analysis

• Meta-analysis is a technique that originates in
  medical science
• One collects a number of quantitative studies,
  say on treatment X for disease Y
• One uses the results of these studies as
  observations in a regression, say the reduced
  mortality as the dependent variable and the dosis
  of the medicine, sex, progression of the disease
  and so on as explanatory variables

33
Meta-Analysis -2

• Meta-analysis is a formal literature review
• 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
• Meta-analysis not only provides a rigorous
  synthesis of the literature, it also identifies
  outlier studies, knowledge gaps, and prioritiess
  for further analysis

34
Wetlands/Brouwer

• 30 studies of the WTP per person for wetland
  preservation in North America and Europe
• 103 observations
• Wetlands were made comparable by looking at the
  their functions flood control, water supply,
  water purification, and nature/recreation
• Each observation was associated with one or more
  functions

35
Wetlands/Brouwer -2

• The study also served methodological purposes
• Additional explanatory variables included payment
  vehicle (tax or other) elicitation format
  (open-ended or other)
• Quality was measured by response rate, but as an
  explanatory variable
• All studies are CVM studies
• Wetland size and income were excluded

36
LnWTP R2 0.38
37
Wetlands/Brouwer -3

• 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
• Little differences between functionality of
  wetlands

38
Conclusion -1

• There is a very wide span of quality in BT from
  examples of no or little basis in economic theory
  to complex estimation and updating procedures in
  BT
• relatively few studies actually test the validity
  of BT
• Classical tests presume identity across studies,
  but in reality, the combination of sites and
  people is bound to be different
• Depending on the test (equality vs. Equivalence),
  you may obtain different evidence for valid
  benefit transfers
• As econometric estimation methods of benefits
  improve, reducing the standard errors of the
  estimates and narrowing the confidence intervals,
  classical tests of equivalence are more likely to
  reject validity

39
Conclusion -2

• Transfers based on demand functions are
  considered superior to transfers based on the
  results of the demand functions value transfers
  of mean WTP...
• ...but sometimes value transfers perform better
  than benefit function transfers
• Benefit transfers are here to stay, but...
• ...there is no blue-print for a perfect benefit
  transfer

40
Sources

• Brouwer,R. (2000) Environmental value transfer
  state of the art and future prospects Ecological
  economics 32 137-152
• Bouwer,R. Spaninks,F (1999) The validity of
  environmental benefits transfer further
  empirical testing. Environmental and Resource
  Economics 14 95-117
• Brouwer R., Langford I.H., Bateman I.J., Turner
  R.K. (1999) A meta-analysis of wetland contingent
  valuation studies. Regional Environmental Change
  1, 47-57.
• Costanza et al. (1997) The value of the worlds
  ecosystem services and natural capital. Nature,
  vol.387253-260
• Kristofersson,D., Navrud,S. (2005) Validity tests
  of benefit transfer are we performing the wrong
  tests? Environmental and Resource Economics
  30279-286
• Special Issue of Water Resources Research, Vol.
  28, no.3 dedicated to benefits transfers