Professional Documents
Culture Documents
Corresponding author.
E-mail address: meyerhoff@imup.tu-berlin.de (J. Meyerhoff).
Introduction
Forest ecosystems harbour most of the terrestrial biological diversity globally and,
therefore, the majority of animal and plant species that are becoming extinct come
from forest ecosystems (Secretariat of the CBD, 2002). Thus, forests are critically
important habitats in terms of the biological diversity they contain and the ecological
functions they serve. However, the threats to forest biodiversity differ very much
between various regions of the world. While, for instance, in developing countries
deforestation is a major threat to forest biodiversity, in Europe the area covered by
forests was increasing slightly in recent decades (MCPFE, 2003). One point of
concern with respect to biodiversity is that European forests are dominated by
relatively young even-aged stands of few tree species in a number of countries, as in
Germany. Therefore, the so-called nature-oriented silviculture is currently the main
trend in European forestry aiming, among other things, at the conservation and
enhancement of forest biodiversity. It is based on less-intensive management
methods favouring retention of trees and decaying wood, the establishment of
natural tree species and species mixtures, and the protection of small key biotopes
(EEA, 2007).
This raises the question to what extent nature-oriented silviculture should take
place. From an economic point of view, comparing the costs and benets arising
from this kind of silviculture could provide helpful information for decision making.
But although nature-oriented silviculture is an important topic in German forestry,
no study on the benets arising from it has been conducted to date (cf. Elsasser and
Meyerhoff, 2007). Moreover, only one study has investigated the non-market
benets of forest biodiversity in Germany. Ku pker et al. (2005) elicited
individuals willingness to pay for a forest biodiversity programme nationwide and
in Schleswig-Holstein, one of the federal states of Germany, using the contingent
valuation method. The study, from which results are reported here (Meyerhoff
et al., 2006), is the rst one in Germany that investigates to what extent people value
the changes in forest biodiversity of nature-based silviculture due to forest
conversion. In both the study regions, the Lu neburger Heide (LH) and the Solling
and Harz (SH) region, we used choice experiments as well as the contingent
valuation method.
The aims of the present paper are two-fold. First, we will present the results from
the choice experiments we employed in our study. The reason for this focus on choice
experiments is that the application of attribute-based methods to forest valuation is
relatively new (Holmes and Boyle, 2003). Second, we will discuss different variants of
calculating welfare measures from choice experiments for an environmental change.
To our knowledge, it has not been agreed in the literature to date whether the
alternative specic constant (ASC) has to be recognised or not when welfare
measures are calculated. Under certain conditions, the welfare measures can become
negative when the ASC is included in the calculation (Adamowicz et al., 1998). Thus,
excluding the ASC as it is done in several studies may be one way to respond to this
situation, but may entail other drawbacks that need to be investigated. The same
argument holds for approaches that conne the calculation of welfare measures to
ARTICLE IN PRESS
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 38
those respondents who are willing to pay, i.e., who at least once did not choose the
status quo alternative.
Methods and background
Choice experiments and welfare analysis
Choice experiments belong to the group of stated preference methods, i.e., they
establish a hypothetical market (e.g., in surveys) in order to value environmental
changes. In contrast to the contingent valuation method, choice experiments are
attribute based and ask respondents to make comparisons and to choose between
environmental alternatives characterised by a variety of attributes and the levels of
these. Therefore, in choice experiments the focus is on the attributes in addition to
overall changes in the provision of the public good in question. Typically,
respondents are offered multiple choices during the survey, with each choice
consisting of two alternative designs of the environmental change in question, say
programme A and B, and the option to choose. Often the latter is represented by the
status quo, i.e., a situation without additional environmental management. The
record of the choices among the alternatives is used to estimate the respondents
willingness to pay (WTP) by modelling the probability of an alternative being
chosen. Choice experiments are useful for multidimensional changes because they
provide a wide range of information on trade-offs among the attributes of the
environmental change in question. Varying the level of the attributes of each of the
alternatives makes it possible to measure the individuals willingness to substitute
one attribute for another. Given that one of the attributes is the monetary cost, it is
possible to estimate how much people are willing to pay to achieve more of an
attribute, i.e., the implicit price, as well as the willingness to pay to move away from
the status quo to a bundle of attributes that correspond to the policy outcomes that
are of interest.
1
In order to link actual choices with the theoretical construct utility, the
random utility framework is used. According to random utility theory the ith
respondent is assumed to obtain utility U
ij
from the jth alternative in choice set C. U
ij
is supposed to comprise a systematic component (V
ij
) and a random error
component (e
ij
):
U
ij
V
ij
ij
. (1)
Selection of alternative h by individual i over other alternatives implies
that the utility (U
ih
) of that alternative is greater than the utility of the other
alternatives j:
P
ih
ProbV
ih
ih
4V
ij
ij
; 8 h; j 2 C; jah. (2)
ARTICLE IN PRESS
1
For an introduction to choice experiments see, for instance, Holmes and Adamowicz (2003) or Stewart
and Kahn (2006). Comprehensive descriptions are provided by Louviere et al. (2000), Hensher et al. (2005)
and in the volume edited by Kanninen (2006).
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 39
Assuming that the error components are distributed independently and identically
(IID) and follow the Gumbel distribution, the probability that alternative h would be
chosen is calculated in the conditional logit model (CL) as
P
ih
expmV
ih
P
j2C
expmV
ij
, (3)
where m is a scale parameter which is commonly normalised to 1 for any one data set.
The systematic part of utility of the jth alternative is assumed to be a linear function
of attributes:
V
j
ASC b
1
X
1
b
2
X
2
b
n
X
n
, (4)
where X
n
represents the attributes and the ASC captures the inuence of unobserved
attributes on choice relative to specic alternatives (Train, 2003). The CL requires
the restrictive assumption that choices are independent of irrelevant alternatives
(IIA). One way to bypass this limitation is to allow for correlations among the error
terms within different subsets of alternatives by estimating a nested logit model
(NL). In this case IIA holds within each subset or nest. The probability of an
individual choosing the alternative h in branch r can be expressed in a NL by
P
hr
PhjrPr, (5)
P
hr
expV
hr
=a
r
expI
r
expa
r
I
r
P
R
k1
expa
k
I
k
" #
, (6)
with
I
r
log
X
H
r
i1
expV
ir
=a
r
" #
. (7)
In this model, P(r) is the probability of choosing branch r, P(h|r) is the probability
of choosing an alternative h conditional on choosing branch r; V
hr
is the indirect
utility of alternative h; the inclusive value coefcient a
r
measures substitutability
across alternatives; I
r
, known as the inclusive value, measures the expected maximum
utility from the alternatives associated with the rth class of alternatives; R is the
number of branches and H
r
is the number of alternatives in branch r (Kling and
Thompson, 1996; Train, 2003).
The implicit prices (also known as part-worth or marginal willingness to pay) for a
change in any attribute, everything else equal, can be estimated using the results of
the conditional as well as the NL model. In a linear model, they are given by
IP b
Attribute
=b
Money
, (8)
where b
Attribute
represents the coefcient of the corresponding non-monetary
attribute, and b
Money
represents the marginal utility of income. They enable some
understanding of the relative importance people place on the various attributes
(Bennett and Adamowicz, 2001). Moreover, in a state of the world model, the
ARTICLE IN PRESS
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 40
welfare change for a combination of changes in attributes is expressed as
CS 1=b
Money
V
0
V
1
, (9)
where CS is the compensating surplus welfare measure and V
0
and V
1
represent the
conditional indirect utility associated with the status quo (subscript 0) and the
changed situation (subscript 1).
Forest biodiversity and attribute-based valuation methods
The application of attribute-based valuation methods (ABMs) to forest valuation
is relatively new (Holmes and Boyle, 2003). From their literature review which
comprises of eight ABM studies, Holmes and Boyle conclude that the general public
is willing to pay for changes in forest management and timber-harvesting operations
that reduce the biological and amenity impacts on forest ecosystems. This nding
was also conrmed by their own results which show that the general public in Maine,
USA, was willing to pay a considerable amount for changing timber-harvesting
practices. Table 1 summarises details of further ABM studies on forest ecosystems
and/or forest biodiversity which are not recognised in the review by Holmes and
Boyle. Lehtonen et al. (2003) investigated Finnish citizens valuations of forest
conservation programmes for southern Finland. In addition to the attributes such as
number of endangered species and biotopes at favourable levels, they included the
attributes information and education about environmental issues and the percentage
of forest area under conservation contracts. Xu et al. (2003) presented WTP values
for forest ecosystem management with respect to the three attributes: biodiversity,
aesthetics, and rural employment impacts in Washington State, USA. The attributes
and their levels were presented as results of management strategies dominated by
preservation, commercial interests or multiple-use management. The willingness to
pay for changes in levels of biodiversity protection under different conservation
programmes in the Coast Range of Oregon, USA, is estimated by Garber-Yonts
et al. (2004). In their study, biodiversity policy was presented as consisting of four
different conservation programmes: salmon and aquatic habitat conservation, forest
age class management, endangered species protection, and large-scale conservation
reserves. Watson et al. (2004) employed a choice experiment in the Robson Valley in
eastcentral British Columbia, Canada, to examine trade-offs inherent in conserving
forest biodiversity. Their attributes include not only conservation characteristics but
also recreation access. Horne et al. (2005) investigated preferences for forest
management at ve adjacent municipal recreation sites in Finland using a spatially
explicit choice experiment. The management alternatives they presented would result
in different levels of site-specic species richness and forest scenery. Bie nabe and
Hearne (2006) elicited the preferences of foreign tourists and Costa Ricans for
increased support for nature conservation and scenic beauty through a system of
payments for environmental services. Respondents were asked to choose between
spatially differentiated areas to receive the environmental service payments. Finally,
Nielsen et al. (2007) determined the recreational benets associated with nature-
based forest management practices. They presented respondents with choice cards
ARTICLE IN PRESS
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 41
A
R
T
I
C
L
E
I
N
P
R
E
S
S
Table 1. Choice experiments eliciting willingness to pay for forest biodiversity
Reference Region,
country
Attributes CE design Choice cards/sets per
respondent
Survey
mode/
useable
sample size
N who
always chose
status quo
(%)
Holmes and
Boyle, 2003
a
Maine, USA Forest road density, dead
trees after harvest, live
trees after harvest,
maximum size of harvest
area, available for
harvesting, width of
riparian buffers, slash
disposal, one-time tax
increase
Completely
randomised design
across individuals
b
Choice card with four
management
alternatives, no status
quo alternative on card;
option of not choosing
was included in a later
question
Mail survey/
278
No gure
given
Lehtonen
et al., 2003
South
Finland
Information and
education, conservation
contracts, conservation
areas, biotopes at
favourable levels of
conservation, number of
endangered species,
increases in annual income
tax 20032012
Randomised main
effects design
b
Eight choice sets, each
with current situation
and two alternatives
Mail survey/
602
14
Xu et al.,
2003
Washington
State, USA
Management strategy,
biodiversity, aesthetics,
additional costs, rural
forest job losses
Design takes into
account the utility
balance among
management plans
by selecting choice
sets from a set of
fractional factorial
design candidates
Four choice sets with
each time four
management plan
alternatives, not status
quo alternative
Mail survey/
1245
No status
quo
alternative
J
.
M
e
y
e
r
h
o
f
f
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
F
o
r
e
s
t
E
c
o
n
o
m
i
c
s
1
5
(
2
0
0
9
)
3
7
5
8
4
2
A
R
T
I
C
L
E
I
N
P
R
E
S
S
that optimise the
estimation of the
MNL model
b
Garber-
Yonts et al.,
2004
Oregon
State, USA
Salmon habitat,
endangered species
protection, forest age
management, biodiversity
reserves and the price a
household would have to
pay
Not clearly
specied, SAS
macros provided
by Kuhfeld were
used
Four choice sets each
with a status quo and
two alternatives
Mail survey/
1090
No gure
given but the
information
that authors
found
resistance to
any change
in
conservation
policy
Watson et
al., 2004
British
Columbia,
Canada
Protected areas in percent
of total region, age of
stands, recreation access,
biodiversity levels, changes
in taxes
Orthogonal main
effects design
b
Seven choice sets, each
with two alternatives
and the current
situation
Mail survey/
1003
18
Horne et al.,
2005
Helsinki
area, Finland
Species richness at each
site, average species
richness, variance of
species richness, scenery at
each site, change in
municipal taxes
Main effects
design
b
Six choice sets, each
with two forest
management
alternatives and the
current situation
On-site
interviews/
431
21
Bie nabe and
Hearne, 2006
Nationwide,
Costa Rica
Number of conservation-
focused zones, number of
scenic beauty/access-
focused zones, payment
through airport taxes
(tourists) or municipal
taxes (Costa Ricans)
Efcient choice
design based on
D-optimality;
computerized
search strategy
adopted from
Kuhfeld
b
Four choice sets, each
with one option
corresponding to
increased Payments for
Environmental Services
(PES) with a focus on
accessibility, one option
corresponding to
Face-to-face/
240 foreign
tourists, 473
Costa Ricans
No gure
given
J
.
M
e
y
e
r
h
o
f
f
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
F
o
r
e
s
t
E
c
o
n
o
m
i
c
s
1
5
(
2
0
0
9
)
3
7
5
8
4
3
A
R
T
I
C
L
E
I
N
P
R
E
S
S
Table 1. (continued )
Reference Region,
country
Attributes CE design Choice cards/sets per
respondent
Survey
mode/
useable
sample size
N who
always chose
status quo
(%)
increased PES with a
focus on conservation,
and the status quo
Horne, 2006 Nationwide,
Finland
Initiator of the contract,
restrictions on forest use,
compensation/ha/year,
duration of contract,
cancellation policy
No details given Six choice sets, each
with two contract
alternatives and the
status quo
Mail survey/
1240
33
Nielsen et al.,
2007
Nationwide,
Denmark
Species composition, tree
height structure, standing
and fallen dead trees,
increase in annual tax
payment per household
SAS macros
provided by
Kuhfeld were used
Six choice sets, each
with two alternatives
visualized by
illustrations, no status
quo option
Mail survey/
548
No gure
given
Notes: For the WTP values the reader is requested to consult the original publications because of the broad range of values that can be calculated based on the
estimation results.
a
This study also used contingent ranking, but the details reported relate to the choice experiment.
b
The description of the CE design is taken almost literally from the publication.
J
.
M
e
y
e
r
h
o
f
f
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
F
o
r
e
s
t
E
c
o
n
o
m
i
c
s
1
5
(
2
0
0
9
)
3
7
5
8
4
4
that show illustrations for the different levels of the attributes species composition,
height structure, and stand and fallen dead trees.
The studies published subsequently to the literature review by Holmes and Boyle
(2003) in general support the earlier nding that the general public is willing to pay
for protection and enhancement of forest ecosystems. In all studies attributes
representing enrichments of biodiversity, for example, number of species or
percentage of habitat in which species are protected, have a signicant and positive
effect on individuals WTP. However, in some studies in which a status quo option
was offered on the choice cards a certain amount of respondents always chose this
status quo option (Table 1), indicating that they are not willing to pay for nature-
oriented silviculture. The study by Horne (2006) differs from the other studies as it
examines the factors that affect the acceptability of biodiversity conservation
contracts among private forest owners in Finland, and the amount of compensation
needed to ensure that the forest owners are at least as well off as before the contract.
Treatment of the ASC when calculating welfare measures
Eq. (4) indicates that the utility may also depend on the value of the ASC.
However, welfare measure calculations for environmental changes differ with respect
to the inclusion of the ASC. Among many other studies, Rolfe et al. (2000), Bennett
et al. (2001), and Birol et al. (2006) included the value of the ASC when calculating
the welfare measure without reporting unexpected results, i.e., negative values of the
measure. Moreover, Birol et al. (2006) explicitly point out that it is necessary to
include the ASC in order to estimate overall WTP. Mogas et al. (2005) present two
welfare measures from a choice experiment about afforestation, one including the
ASC and the other excluding it. The welfare measure that includes the ASC is higher
but both are positive. On the other hand, Adamowicz et al. (1998) report that when
they included the ASC their linear CE specication produced a negative welfare
measure for the proposed environmental change. The ASC equalled one when the
status quo option was not chosen and had a negative sign indicating that
respondents are not in favour of moving away from the status quo. The authors
consider the signicant and negative ASC to be a form of status quo bias or
endowment effect and suggest as possible explanations for respondents choices,
inter alia, mistrust in the providing organisation, complexity in the choice task or
protest against the survey. When Adamowicz et al. (1998) excluded the ASC, the
welfare measure was positive.
Among the studies shown in Table 1, Xu et al. (2003), Bie nabe and Hearne (2006)
and Nielsen et al. (2007) only present marginal willingness to pay values. Lehtonen et
al. (2003) do not take the ASC into account when calculating the welfare measures
for their forest management strategies. Garber-Yonts et al. (2004) report that when
they take the ASC into account in welfare calculation it partially offset the estimated
benets of changing conservation levels. The ASC indicates the status quo and is
signicant and positive. Watson et al. (2004) rst of all excluded all respondents who
had always chosen the status quo option (18% of the sample) from their choice
model. But even in this case a change from the status quo was still negative for many
ARTICLE IN PRESS
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 45
respondents. When they calculate welfare measures for various management
scenarios, inclusion of the ASC results in all cases in negative gures, i.e., the costs
to move away from the status quo are higher than the benets from the biodiversity
conservation measures. Horne et al. (2005) report a signicant and positive ASC,
representing the current situation. The compensating variation measure of their new
management scenario indicates a loss for the whole sample when the ASC is
recognised. Therefore, they conclude that any change in management would need to
bring large benets to compensate for the negative impact of moving away from the
current situation. In the study by Horne (2006), the ASC also indicates preferences
for no additional conservation. Calculating the welfare measure based on an
estimation using all data (respondents) results in a negative measure. Accordingly,
forest owners would have to be compensated for biodiversity conservation services.
In contrast, calculating the welfare measure for the same contract but based
on an estimation that excluded all those respondents who had always chosen the
status quo resulted in a positive welfare measure. Leaving out those who never chose
an alternative to the status quo changes not only the magnitude of the welfare
measure, but also the sign indicating whether people would have to be compensated
or not.
Forest conversion and biodiversity in Lower Saxony
Study area and selection of biodiversity attributes
Approximately one quarter of Lower Saxony, Germany, is covered by forests
(1.1 million hectares). Of this, 32% is owned by the state of Lower Saxony and 46%
is privately owned. The remaining forests are owned by communities and cloisters.
The LH, one of our study regions, is located in the relatively humid north-western
part of Germany. Due to historic land uses, large parts of the landscape are covered
with heath and, at present, with pine monocultures. The other region is the area of
the SH. Both the Solling and Harz are part of the mountain ranges in the south of
Lower Saxony. There are naturally occurring beech forests on nutrient-poor and
acidic sandy soils. However, historical land use such as intensive forest grazing and
timber use led to widespread devastation at the end of the 18th century. Thus, the SH
area was reforested mainly with Norway spruce, which still covers large areas of the
mountain ranges.
As a response to the domination of coniferous trees, in 1991 the government of
Lower Saxony introduced the forest strategy programme LO
WE (Langfristige
O
WE CE LO
WE
Habitat for endangered
and protected species
(HAB)
Medium, high High Low, medium,
high
Medium
Species diversity (SPE) Medium, high Medium Medium, high Medium
Forest stand structure
(FSS)
Low, medium,
high
High Low, medium,
high
Medium
Landscape diversity
(LCD)
Low, medium,
high
Medium Low, medium,
high
Medium
Contribution to forest
conversion fund (h)
0
a
, 5, 10, 20,
35, 50, 75
0
a
, 5, 10, 20,
35, 50, 75
WE forest conversion
programme with and without ASC
Lu neburger Heide Solling and Harz
CL NL CL NL
All respondents
With ASC 35.03 (52.4717.59) 121.03 (237.194.86) 32.59 (45.8119.37) 44.44 (73.7715.10)
Without ASC 31.30 (13.2349.36) 32.39 (18.4046.37) 14.40 (6.9821.83) 12.62 (5.0620.18)
Conned to those who at least once chose programme A or B
With ASC 45.86 (37.6755.30) 52.4 (41.0268.39) 36.92 (30.0344.27) 36.83 (27.1653.75)
Note: The 95% condence intervals are given in parentheses and were calculated using the Krinsky and
Robb (1986) procedure with 1000 draws.
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 55
Similar results have been reported by other authors (Watson et al., 2004; Horne,
2006). This raises the question of which of the three welfare measures is appropriate?
While the rst approach, i.e., including the ASC, may result in an underestimation of
the benets from biodiversity enrichment, the other two approaches could result in
an overestimation. Including the ASC may not be justied because not all
respondents who always chose the status quo would require compensation for
moving away from the current situation. From discussions during the focus groups,
we got the impression that many people are not willing to pay because they simply
do not care or have other priorities than enriching forest biodiversity. But they
would not suffer any loss if biodiversity is enriched according to the other
respondents willingness to pay. Therefore, we decided not to include negative prices
in the choice design which would have made it possible to measure respondents
willingness to accept. Moreover, implementing the LO
kologischen Waldumbaus in den Regionen Solling und Lu neburger Heide. Reihe B. Forschungszen-
trum Waldo kosysteme der Universita t Go ttingen, Go ttingen.
Ministerial conference on the protection of forests in Europe (MCPFE), 2003. State of Europes Forests
2003. The MCPFE Report on Sustainable Forest Management in Europe. Austria.
Mogas, J., Riera, P., Bennett, J., 2005. Accounting for afforestation externalities: a comparison of
contingent valuation and choice modelling. European Environment 15, 4458.
Niedersa chsische Landesregierung, 1991. Niedersa chsisches Programm zur langfristigen o kologischen
Waldentwicklung in den Landesforsten (Programme for Long-term Ecological Forest Development in
the Lower Saxony State Forests). Hannover.
Nielsen, A.B., Olsen, S.B., Lundhede, T., 2007. An economic valuation of the recreational benets
associated with nature-based forest management practices. Landscape and Urban Planning 80, 6371.
Rolfe, J., Bennett, J., Louviere, J., 2000. Choice modelling and its potential application to tropical
rainforest preservation. Ecological Economics 35, 289302.
Secretariat of the Convention on Biological Diversity (CBD), 2002. Review of the status and trends of, and
major threats to, the forest biological diversity. CBD technical Series no. 7. Montreal.
Stewart, S., Kahn, J.R., 2006. An introduction to choice modeling for non-market valuation. In: Alberini,
A., Kahn, J.R. (Eds.), Handbook on Contingent Valuation. Edward Elgar, Cheltenham.
Train, K.E., 2003. Discrete Choice Methods with Simulation. Cambridge University Press, Cambridge.
Xu, W., Lippke, B.R., Perez-Garcia, J., 2003. Valuing biodiversity, aesthetics, and job losses associated
with ecosystem management using stated preferences. Forest Science 49, 247257.
Watson, D.O., McFarlane, B.L., Haener, M.K., 2004. Human dimensions of biodiversity conservation in
the interior forests of British Columbia. BC Journal of Ecosystems and Management 4, 120.
Zerbe, S., 2002. Restoration of natural broad-leaved woodland in Central Europe on sites with coniferous
forest plantations. Forest Ecology and Management 167, 2742.
Zerbe, S., Kreyer, D., 2007. Inuence of different forest conversion strategies on pine (Pinus sylvestris L)
stands a case study on permanent plots in NE Germany. European Journal of Forest Research 126,
291301.
Zerbe, S., Kempa, D., Xinrong, L., 2007. Managing biological diversity in forests by applying different
development objectives. Archiv fu r Naturschutz und Landschaftsforschung Ma rz, 326.
ARTICLE IN PRESS
J. Meyerhoff et al. / Journal of Forest Economics 15 (2009) 3758 58