- Email: [email protected]
Forest certification and Swedish wood supply
Forest Policy and Economics 9 (2007) 452 – 463 www.elsevier.com/locate/forpol
Forest certification and Swedish wood supply Ljusk Ola Eriksson a,*, Ola Sallna¨s b,1, Go¨ran Sta˚hl a,2 a
Department of Forest Resource Management and Geomatics, SLU, SE-901 83 Umea˚, Sweden b Southern Swedish Forest Research Centre, P.O. Box 49, SLU, SE-230 53 Alnarp, Sweden
Received 19 April 2004; received in revised form 20 October 2005; accepted 8 November 2005
Abstract A number of measures have been introduced into Swedish forestry in order to satisfy demands on biodiversity and sustainability. Protection measures include set asides, areas with continuous cover forestry and retention trees on harvesting sites, to name but a few. Most of these practices will have implications for the total wood supply of the Swedish forests. In 1998, the Swedish standard for forest certification according to the Forest Stewardship Council (FSC) was approved. The standard adopts many detailed regulations regarding how forests should be managed. The aim of this study is to assess the likely effect of FSC certification on short- and long-term supply of roundwood in Sweden. A scenario, expressed as the distribution of the forest land base on production forests, areas with restrictions for management and reserves, is input into a simulation model in which forest owners are assumed to be guided by economic criteria when deciding on management. Four scenarios were created based on the data of the Swedish National Forest Inventory. For each scenario the short-term wood supply was assessed. Also, a number of long run simulations with different relative price levels are presented. The results indicate that full adoption of the FSC standard on the entire land base, compared with adherence only to the Forestry Act, could result in a substantial reduction in supply, or, conversely, could induce a price increase in case supplied quantities should be maintained at current levels. This is under the assumption that no compensating mechanisms, exogenous to the model, come into effect. Furthermore, a sustained price increase that would compensate for lost volumes today does not seem to prevail in the long run. In conclusion, the ongoing adaptation of Swedish forestry to the standards of the certification programs could have substantial effects on the timber supply and will probably influence the international competitiveness of the Swedish forest sector. Effects and tendencies like these should form integral parts of future analyses of wood balances and wood supply in a regional perspective. D 2006 Elsevier B.V. All rights reserved. Keywords: Long range projection; Forest sector; Price expectations; Markov model
* Corresponding author. Tel.: +46 90 7865840; fax: +46 90 778116. E-mail addresses: [email protected] (L.O. Eriksson), [email protected] (O. Sallna¨s), [email protected] (G. Sta˚hl). 1 Tel.: +46 4 415181; fax: +46 40 462325. 2 Tel.: +46 90 7865837; fax: +46 90 778116. 1389-9341/$ - see front matter D 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.forpol.2005.11.001
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
1. Introduction The Swedish standard for forest certification according to Forest Stewardship Council (FSC) principles and criteria was approved in 1998 (FSC council, 1998). Within a short period, all major forest companies in Sweden had adopted the standard and modified their management practices. The overall aim of the standard is to outline management principles that maintain the ecosystem’s productivity and biodiversity, secure local people’s livelihood, and promote long-term valuable wood production. In particular, the standard comprises many detailed regulations that determine how forests should be managed. For example, 5% of forest areas should be set aside for free development or be managed to promote biological values. Certain areas must be managed to promote deciduous forests, and boundary zones to streams, lakes, and non-productive land should be maintained, as should also a certain minimum number of trees (10 ha 1) during harvesting operations (FSC council, 1998). Compared with management practices of most companies before adoption of the FSC standard, the new management regimes imply lower levels of potential harvest. Some studies have quantified the effects of the new management on harvesting levels (e.g., Lundstro¨m et al., 1997; Skogsstyrelsen, 2000; Jacobsson, 2002). These studies show the range of harvesting reduction due to the FSC standard to be between 10% and 20%. Similar results are presented by Verkaik and Nabuurs (2000) in a study of the consequences of adopting bnature-orientatedQ management in Scandinavian forests, and by Nabuurs et al. (2001) analyzing the effects of a multi-purpose management scenario for European forests. However, in the latter studies, no attempts were made to apply the FSC standard as such. Long-term sustainable harvesting levels in Sweden before the adoption of the FSC criteria have been assessed to be slightly more than 100 million m3 per year (Skogsstyrelsen, 2000). The annual cutting levels during the last decade have been about 70–75 million m3 per year. There are several reasons for the relatively large difference between the two figures. First, many forests in Sweden are still relatively young, and thus not mature for final felling. Second, we are dealing with different concepts here. The level
453
around 100 million m3 is what would be a sustainable harvest from the forests from a biological point of view, reflecting the potential yield. In this calculation, no restrictions of a technical, social or economic kind are taken into account. The recorded figure of 70– 75 million m3 reflects the real supply of wood, given the economic setting and restrictions on forestry of different kinds. In most of the above studies of the effect of the FSC standard on harvesting levels, the modeling approach has been to control management activities by a set of rules formulated to mimic certain forest management practices. These rules are applied even if a deeper analysis (which is typically not made) for some forestry activities would imply negative results from the economic point of view. Together this implies that these studies are dealing with the bpotentialQ discussed above. A different approach is to put forestry into an economic framework, allowing harvesting activities to be controlled largely by economic considerations. The result of such an analysis most likely would be a different supply of roundwood, leading to lower harvesting levels compared with the levels that correspond to the biological potential. This kind of analysis would provide a better assessment of the consequences on harvesting levels of new demands on forestry; e.g., those caused by FSC certification. The aim of this study is to analyze the implications of the FSC standard for the supply of roundwood in Sweden with respect to both the effects on current supply relationships and the long-term level of harvesting. Two different assumptions about the rigor of the FSC standard are analyzed. The study contrasts these two scenarios with a modified version of the environmental restrictions of the recent SKA99 study (Skogsstyrelsen, 2000) used as a baseline case. The constraints are less restrictive than in the SKA99 study and are intended to reflect forest management under the current Forestry Act, with no adherence to the particular stipulations of the FSC standard. Furthermore, an extreme case with no environmental considerations is included in order to assess the maximum economic potential. Whereas the Swedish Forestry Act establishes that environmental and production goals are equally important, it contains very few binding requirements of, e.g., leaving trees and stands of the kind stipulated by
454
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
the FSC certification standards. Thus, the environmental concerns in a scenario adopting the minimum requirements of the Forestry Act are not as high as in a scenario adopting the certification standards. The different levels of environmental objective form the basis for four scenarios. The scenarios differ with respect to the areas available for production or assigned for modified management and reserves. Except for the distribution of the forest area for permissible management, the economic and other conditions remain the same. The analysis uses a model of forest management decisions developed by Sallna¨s and Eriksson (1989). The model derives economically optimal harvesting regimes for all types of forest given certain assumptions regarding prices and costs.
2. Methods 2.1. Basic model concepts The results of this report rest on solutions to a forest management problem that confronts the forest owner. The outcome of different projections is the aggregate of the owners’ actions. We begin with the basic assumptions behind the management problem, then present the different models and solution procedures in the next section. The forest owner seeks to maximize monetary profits; i.e., the net present value of the forest holding. Three assumptions will underpin the representation of the management problem in the model. First, when calculating the optimal treatment for a certain piece of land at a certain time, we assume that the owner expects the current price level to persist, an assumption that should be consistent with efficient timber markets. Second, we assume that all forest owners encounter the same economic conditions. This means that the net revenue of a treatment of a certain piece of land, whether in establishment or harvesting, should be the same irrespective of ownership category. From this assumption, it also follows that financial markets must be efficient, as all owners would face the same interest rate. Third, it is assumed that decisions by the forest owner are not affected by considerations that span the whole forest estate, such as non-declining harvest constraints or objectives related to the composition of species and ages of the entire estate.
We use a stage-structured Markov model. These assumptions will then have the following implications for modeling profit maximizing by Swedish forest owners. The third assumption implies that the problem is separable. Given a set of prices, an optimal solution to the forest management problem is derived by selecting an optimal treatment for each piece of land separately (Johansson and Lo¨fgren, 1985). With a Markov model, each state of the model may be inspected one-by-one while still preserving optimality. The second assumption, together with the third, allows all forest belonging to the same forest state to be treated in the same (optimal) way. The first assumption implies that, for a given price level, the optimal decision will be the same for forests in a given state regardless of when it is made, as historic prices are irrelevant. One may note that, while the second and third assumptions are more technical in nature and primarily serve the purpose of simplifying the administration and running of the model, the constant price assumption could have fundamental consequences for the kind of results that emerge (Sallna¨s and Eriksson, 1989). 2.2. Forest model and solution procedure The solution procedures are built on the dynamics of the forest model (Sallna¨s, 1990). The Markov forest model is given by X xtþ1 ¼ ytik Tijk 8 ja I ð1Þ j ik
xti ¼
X
ytik
8 ia I
ð2Þ
k
where xti is the area residing in state i in period t, y ikt is the area in state i subject to treatment k in period t, and Tijk is the probability for an area residing in state i in period t to be found in state j in period t + 1 when subject to treatment k. Thus, (1) gives the transition of areas between states between period t and t + 1 as a function of the treatment of the forest and (2) ensures that the whole forest area is allocated a treatment (no action is also regarded as a treatment). The total set of states is I and the stage duration is 5 years. The forest area is divided into three broad treatment groups depending on what kinds of treatment are allowed. The amount that belongs to each
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
group depends on what scenario is analyzed (see Section 3.2). Within each treatment group, the forest area is differentiated between established forest (forest with an average height above 6 m) and young forest (including bare land). The state space of established forest is defined by geographical region, owner category (non-industrial or other), site quality, species composition, age, volume, thinning state (thinned or not thinned in the previous period) and comprises a total of 51,840 states. The state space of young forest is delineated by geographical region, owner category, site quality and age, and is composed of 384 states. The forest management problem in period t consists of maximizing the net present value; i.e., ! X X max nik þ d Tijk ej ytik ð3Þ jaI
ik
455
state of the forest in terms of the number of hectares in each state of the Markov model, x1i . 2. Set the economic conditions in terms of, for instance, discount rate and timber prices, and compute net revenues, n ik , and expected values, e i . 3. Make a projection for a number of 5-year periods by (i) solving the management problem for period 1 with Eqs. (3)–(5), (ii) projecting the forest into period 2 with Eq. (1), and then repeating steps (i) and (ii) for the number of periods that the projection should encompass. Once the projection has been made for all treatment groups, the result is consolidated to cover the whole forest area. By varying the amount of area distributed on different treatment groups or by changing the economic parameters, different developments may be created.
subject to xti ¼
X
ytik
8ia I
ð4Þ
k
ytik a Ki
3. Data 3.1. Regions
8 ia I
ð5Þ
where n ik is the net revenue of treatment k in forest belonging to state i, d is the one-period discount factor and e i is the expected value of forest and land of state i. Eq. (4) is equivalent to (2), and (5) ensures that the treatment of each state i belongs to the set of permissible treatments K i . The expected value of each class i, e i , is derived by solving the following set of equations. ( ei ¼ maxkaKi nik þ
X
) dTijk ei
8 ia I
ð6Þ
j
Thus, the expected values are derived under an infinite time horizon, assuming optimal treatments and constant prices. System (6) is solved with a successive approximation algorithm (Denardo, 1982). The first step in using the model is to distribute sample plot data of the forest on the three different treatment groups depending on scenario. Then, for each treatment group, the following steps are completed. 1. From the sample plots, compute the area residing in each state; i.e., obtain a description of the initial
Calculations are done for the four balance regions as defined in SKA99 (Skogsstyrelsen, 2000), here denoted by BR1–BR4. The regions correspond well to the regions of the growth model (Sallna¨s, 1990). Certain functions, such as those for reduction for bark, are differentiated on Northern and Southern Sweden. BR1 and BR2 are here classified as Northern Sweden, and BR3 and BR4 as Southern Sweden. Other data, such as minimum final felling age, need to be referred to by even smaller regions. Here, BR1 will be represented by the county of Va¨sterbotten, BR2 by Dalarna, BR3 by Va¨rmland, and BR4 by Jo¨nko¨ping. 3.2. Scenarios The forest owners are constrained in the choice of treatments by the endowment of forest areas in different treatment groups. There are three such groups: production forest, forest with nature-oriented modified management, and reserves (this corresponds to the groups bTraditionell sko¨tselQ, bNaturanpassad sko¨tselQ and bIngen avverkningQ in SKA99 (Skogsstyrelsen, 2000), respectively). A given distribution of the Swedish forest area among the treatment groups is
456
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
here termed a scenario. The distribution of the forest among treatment groups for a given scenario is static; i.e., the initial distribution of treatment groups for a given scenario remains constant throughout the entire projection period. Production forests correspond to forests where both thinning and final felling are allowed. Minimum final felling age is set according to the previous Forestry Act, which for normal site conditions is about 80 years. This gives permissible ages about 10 years higher than in current regulations, which is motivated by our impression that many forest owners, including large forest companies, owing to environmental concerns, economic considerations or a need to ration forest for final felling apply rules that are stricter than the current implementation of the Forestry Act. The model allows thinning only for age classes that are not permissible for final felling. Final felling is in the form of clear felling. The treatment group with modified management differs from the first by allowing only thinning. Thinning is permissible for all established forest irrespective of age. Forests set aside as reserves have neither thinning nor final felling. All treatment groups share the same growth model given by Eq. (1). A data set consisting of 21,301 sample plots of the National Forest Inventory (NFI) from 1996, 1997 and 1998 was prepared. The area is 22.4 million ha, which corresponds to the productive forest area in Sweden, reduced by 1 million ha of existing and planned reserves with legal protection. The planned reserves are based on an estimate made in the SKA99 study of the likely area to be protected up to about 2010 (Skogsstyrelsen, 2000); in the projections, they remain protected over the entire planning horizon. The areas associated with the NFI plots on treatment groups for a particular scenario were distributed as follows. The first step established the area for each region and owner category that should belong to a particular treatment group. Categories should, although they have identical economic conditions, be kept separate since they are differentiated in the growth model. The next step derived the distribution for each region and category of the forest area on values for two categorical sample plot variables describing the importance of the plot for nature conservation. One variable is more closely related to the legal status and usage of the plot, and
will below be denoted S, and the other to the state of the tree layer, and is below denoted F. Both variables have values ranging from 1 to 6 where 1 signifies the highest priority for retention and the 6 signifies no specific values. A particular combination of the categorical variables determined what treatment group, or groups in boundary cases, the area represented by a sample plot should belong to such that the target area established in the first step was met. The procedure was somewhat complicated due to the definitions associated with the variables. The delineation was done as follows for each region and category for reserves: all plots with F = {1,2} and all plots with S = 1 were allocated to reserves. Then, as many plots as needed with S = 2, in increasing order of variable F, was set aside. The procedure was then followed for S = 3, etc. With a few, and in area insignificant, exceptions, the following procedure was followed on the remaining plots for each region and category for modified management: as many plots as needed with S = 2, in increasing order of variable F, was allocated to modified management. The procedure was then followed for S = 3. Then, plots were allocated in order of minimum sum S + F. Ties–S,F = (4,4) has the same sum as S,F = (5,3)–were broken by giving priority, first, to the plot with minimum value on min(S,F), and second, on taking plots with a higher value on the F variable before taking plots with this value on the S variable. After this, the remaining plots were placed in the production forest group (the detailed results for each region and category are given in Eriksson et al., 2002, Appendix 2). Four scenarios are studied: the NUL, BAS, FSC and FSC+ scenarios. The principles behind the distribution of areas into treatment groups for different scenarios are the following. The NUL scenario is included here as a reference in order to assess the maximum economic potential, and lacks practical significance in today’s forestry. Observe that this scenario, as the other scenarios, excludes areas of existing and until year 2010 planned reserves as was stated above. Whereas the BAS scenario is intended to include only environmental considerations induced by the Forestry Act, the SKA99 study reflects the environmental restrictions applied by Swedish forestry at the end of the 1990s. Subsequently, the SKA99 study
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
encompasses forest owners that already, to a smaller or larger extent, have implemented the FSC or some other certification scheme. The distribution of areas for the BAS scenario follows the SKA99 study (Skogsstyrelsen, 2000) as regards areas with natureoriented modified management. The areas designated as reserves in the BAS scenario follow what in the SKA99 study are set aside as patches and biotopes demanding care, amounting to about 2% for each. However, the BAS scenario omits the effect of retention trees in the SKA99 study, rendering this scenario less restrictive than the SKA99 study. The BAS scenario therefore represents forestry with no adherence to the particular stipulations of the FSC standard. The FSC scenario is intended to reflect a situation where the FSC certification standard is applied to all forests in Sweden. The FSC standard prescribes that at least 5% of the productive forest area should be set aside (2% in the SKA99 study). A further 3.5% constitutes smaller patches demanding care that should not be harvested (2% in the SKA99 study). The effect of retention trees left at the harvesting sites reduces the available area by an estimated 4.5%. This produces a total restricted area of 13%. The area with modified management remains essentially the same as in the BAS scenario, but is somewhat rearranged owing to the increased area requirements of reserves. The FSC+ scenario means that another 4% and 5%, compared with the FSC
457
Table 2 The distribution of the area of production forest on different site quality classes for different scenarios (%) Scenario
NUL BAS FSC FSC+
Site quality classa 1
2
3
4
11 10 10 10
26 25 25 24
29 30 30 31
33 35 35 35
Class is defined in m3 ha 1 year 1 , potential mean increment, for all classes except classes 1 and 2 (the least productive sites) in region 1 (Northern Sweden) where the definition is based on altitude (see Sallna¨s, 1990; Appendix 1). a
scenario, are added to the areas of modified management and reserves respectively. Following these principles and procedures, the ensuing distribution of areas is as shown in Table 1. We can note a minute tendency for the site quality composition of production forest to improve as areas are transferred from production forest to forest with modified management and reserves (Table 2). 3.3. Economic data Prices of timber and pulpwood were calculated based on averages for the years 1996–1998 for pine and spruce, whereas broadleaves only qualified as pulpwood. Pulpwood prices were collected from the Official Statistics of Sweden (Skogsstyrelsen, 1999a). Price lists for timber are for BR1 from Norra Skogsa¨garna,
Table 1 The distribution of the forest area into treatment groups (million hectares) Scenario NUL
BAS
FSC
FSC+
Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area
(%)
(%)
(%)
(%)
Production forest
Modified management
Reserve area
Sum
14.53 7.86 22.39 100 13.08 7.32 20.40 91 11.68 6.64 18.32 82 10.37 5.98 16.36 73
0 0 0 0 0.71 0.33 1.04 5 0.75 0.38 1.13 5 1.40 0.66 2.06 9
0 0 0 0 0.73 0.21 0.94 4 2.10 0.85 2.95 13 2.76 1.21 3.97 18
14.53 7.86 22.39 100 14.52 7.86 22.38 100 14.53 7.87 22.40 100 14.53 7.85 22.39 100
458
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
Table 3 Annual harvest during the first 10 years of simulation (million m3) at different discount rates under the BAS scenario Discount rate (%) 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 61.6 66.0 67.3 71.1 72.6 76.3 81.1 83.5 87.3 91.5 96.6
Umea˚ district, for BR2 and BR3 from Mellanskog and the price lists applicable for Dalarna and Va¨rmland, respectively, and for BR4 from So¨dra Skogsa¨garna, Jo¨nko¨ping district. Timber prices of standard assortment are given over diameter for five quality classes for pine and four classes for spruce in the price lists. They are weighted together into one price list for each species and region by using statistics on the distribution on quality from three timber measurements associations (VMF, 1998), assuming that the distribution of volumes over diameter classes does not differ between qualities. Log conversion is calculated using the method of Ollas (1980) with the relation of volume over and under bark given by Na¨slund (1947). Species composition and dimension are of importance for the economic return from harvests. The growth model includes three broad species classes: more than 50% conifers with pine dominating, more than 50% conifers with spruce dominating, and more than 50% broadleaves. For a better estimation of the species distribution in individual states of the model, species distribution functions were developed from NFI data (see Eriksson et al., 2002, Appendix 4). Average diameter functions from Eriksson et al. (2002, Appendix 5) are used as a basis for the application of the conversion functions in individual states of the growth model. Costs for cleaning, scarification, planting (excluding plants) and precommercial thinning on a perhectare basis are from Skogsstyrelsen (1999b) for Northern and Southern Sweden, respectively. Plant cost is from Svenska Skogsplantor (1999), with the number of plants differentiated by region and site index. Harvesting costs for thinning and final felling were estimated with functions for harvester and forwarder from SLU (1989). In order to account for increased productivity since 1987, the input data on cost per hour was reduced such that the cost per cubic meter
was the same for a tree with 20 cm diameter at breast height as the cost given by Skogsstyrelsen (1999a). (For details on prices, see Appendix A). The cost of capital–i.e., the discount rate–was estimated using the method presented by Berck and Bible (1984). Assuming that the model is correct, a real after-tax discount rate that gives model results coinciding with observed behavior is determined. Running the model for the first two 5-year periods under the BAS scenario shows that a discount rate of 2.5% generates a harvest level that quite closely coincides with the current harvest level in Sweden (Table 3). This discount rate will be used in subsequent analyses.
4. Results Results are here presented as an aggregation of growth and harvest of two stages instead of the separate 5-year periods for ease of interpretation; there is a variation in harvests, up and down, for the first two 5-year periods that make the single 5-year period results less relevant (see also explanation to Fig. 2). The total annual harvest during the first 10 years of simulation in the BAS scenario is 76.3 million m3 (Table 4)). This level is 9% higher than that of scenario FSC and 15% higher than FSC+, while the NUL alternative would yield some 7% higher harvest level than that of the BAS scenario. These differences are rather well reflected in the results for the individual balance regions, except for BR3, where we notice only a marginal difference between the FSC and FSC+ scenarios. In Fig. 1, the average supply for the first 10 years is given for a gradient of relative timber prices. One implication of the results is that in order to reach the same short-term harvesting level in the FSC scenario Table 4 Annual harvest for the first 10 years, for different scenarios Scenario
BR1
BR2
BR3
BR4
Total
NUL BAS FSC FSC+
110 27.1 89 84
106 18.4 91 85
104 7.7 92 91
105 23.1 91 84
107 76.3 91 85
The figures for scenario BAS are given in million m3 , and for the other scenarios in percent of scenario BAS.
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
459
Fig. 1. Annual supply during first 10 years for different scenarios and different relative prices.
as in the BAS scenario with relative price 1.0, the price level has to be increased by some 17%. The corresponding figure for the FSC+ scenario is just below 40%. Price elasticities provide another perspective on the same phenomenon. Calculating over a 20% interval on both sides of relative price 1.0, the elasticities range from 0.55 to 0.51, with the more benvironmentalQ scenarios in the lower end. As expected, the elasticities decrease with an increased price level. In Fig. 2, the development over time of annual harvests is given for the different scenarios at a price level of 1.0. The curves follow each other consistently. The dip in the second period is found for all scenarios. This is probably due to an adjustment of the forest estate to the economic assumptions of the model; existing over-mature forest, from an economic
Fig. 3. Harvests over time for some scenarios at different price levels as a relative difference from the level of the BAS scenario at price level 1.0.
point of view, is harvested in the first period, leading to a reduced supply in the next period. During the rest of the century we find a rather steady increase in harvests. After a decrease for some periods, the harvests reach the former high level at the end of the simulations. The price effects on harvest levels illustrated in Fig. 1 are in the short term, in this case the first 10 years. In Fig. 3, the long-term harvest development for a number of selected combinations of scenario and price level are compared to that of the BAS scenario at price level 1.0. Although it is possible even in the FSC scenario to reach the BAS harvest level in the beginning of the simulation by increasing the price, the subsequent decrease in harvest level reaches about 10%. If we sum the harvests over 100 years in the different scenarios and compare them with the harvests of scenario BAS with price 1.0, we obtain the results shown in Table 5. Prices affect mainly the distribution of harvests over time. It is evident that the total harvest over the 100 years is more or less the Table 5 Average harvest over 100 years in relation to the BAS scenario for different price levels BAS
Fig. 2. Annual harvest over time for the four different scenarios at a relative price level of 1.0.
FSC
FSC+
Relative 1.0 1.1 1.2 1.0 1.1 1.2 1.0 1.1 1.2 price level Relative 1.00 0.99 0.99 0.91 0.91 0.90 0.87 0.86 0.86 harvest
460
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
Fig. 4. The percentage of final fellings of the total harvests for different scenarios (price level 1.0).
same irrespective of the price level, with a slight tendency for high initial harvests associated with higher price levels resulting in a somewhat lower total. Another result is that the FSC scenario bcostsQ 10–11% of the harvest level of BAS at all price levels. The corresponding figure for the FSC+ scenario is 13–14%. A marked difference between the different scenarios is the part of the total harvest that stems from final felling as shown in Fig. 4. In the FSC+ scenario, the area allocated to modified management, and consequently to thinning as the only available management option, is larger than in the BAS and FSC scenarios. Consequently the thinning part of the harvest increases. The initial peak in harvest in Fig. 2 is also found in this figure with respect to final harvest. This underlines the interpretation that in the first period the state of the forest is adjusted to the economic assumptions of the model; those forest areas that are over-mature in relation to the economic conditions are harvested.
5. Discussion Comparing the scenarios in terms of harvests, the first 10 years show that the management of today, expressed as the BAS scenario, features a supply level that is 7% lower than that of a situation where all forest land is used entirely for timber production. The FSC and FSC+ scenarios compared with BAS feature 10% and 15% lower levels, respectively. These differences correspond well with the fraction of the
forest area that is assigned to uses other than pure timber production. This observation corresponds well with the fact that the areas transferred from production forest to modified management and reserves has about the same site quality composition as the areas retained for production (see Table 2). The relatively low elasticities, even at a price level of 1.0, imply that to compensate in the short run for the decrease of supply that would follow from the extended conservation measures in the FSC scenario, prices would have to rise substantially. Other compensating mechanisms include increased imports and reduced wood consumption. Whatever effects prevail, the implication is that extended conservation measures would increase the strain on the wood consuming industry. A number of other assumptions, which range from basic economic assumptions to details of the submodels, need to be scrutinized in order to appreciate correctly the results. The scenarios of the study are, of course, more-or-less theoretical constructs. The BAS scenario is intended to reflect management that only meets the requirements of the current Forestry Act. There are only limited data to confirm that the ensuing result is what would have been forestry without certification. The FSC scenario has a better-founded basis in the stipulations of the FSC standard, although it is questionable how far the standard will go beyond the Forestry Act. It should be emphasized that the BAS scenario should not be confused with current forestry practice as more than half the Swedish forest area is already subject to FSC or other certification standards. The economic model used in this study uses two important assumptions. Forest managers are supposed to act according to beconomic manQ rationality; i.e., they are in every situation supposed to take the action that maximizes the present net value of their forestry. This is a simplification, which however can be motivated by the fact that the study should not primarily be interpreted as a prognoses; it is rather a model for clarifying the relationships between economic entities, such as harvest volumes and timber income, on the one hand, and environmental policies on the other. The second assumption is that the forest manager at every decision making step expects prices and costs not to change in the future. This assumption, that markets exhibit a rational expectations equilibrium, is
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
supported by a large number of investigations of financial markets (Brealey and Myers, 2000). It has also found support in studies of timber markets (Washburn and Binkley, 1990), although contradictory results have been reported (Lohmander, 1988). The model gains credibility when, as expected, its elasticities decrease with rising prices and consequently higher supply. Also, the elasticities, around 0.5, are of the same magnitude as those estimated from empirical data; Bra¨ nnlund (1988) presents estimates between 0.4 and 0.8 and Bra¨nnlund and Kristro¨m (1993) a figure in the range 0.1 to 0.2. At a discount rate of 2.5%, the modeled harvest is in line with current harvests in Sweden. It seems reasonable to seek a harvest level by choosing a realistic discount rate. The model’s real after-tax discount rate of 2.5% corresponds closely to empirical data on returns on investments for comparable assets (Brealey and Myers, 2000). The reasonableness of the combination of discount rate and harvest level also gives credibility to the model as such. However, the model’s distribution of harvests over balance regions does not fully reflect the factual situation in Sweden today. Using the ranking of the plots as the basis for distribution among treatment groups implies that forest areas with the highest conservation value will get the most extended protection. The distribution probably becomes more efficient from a nature conservation point of view than is possible in practice. In the latter case, effects of estate borders, lack of accurate data and other factors will also influence choice. A related issue is the assumption that the distribution of forest between prescribed areas will be permanent throughout time. This means that the setting aside of forest areas for conservational purposes or the assignment of specific management programs is unchanged through the simulations. Consequently no areas grow binQ or boutQ of a treatment group, although some of the criteria used for distributing the area refer to variables that are connected to the tree layer and thereby could change over time. This is probably of limited importance with respect to the results of the study. Retention trees have, in the model, the effect that a certain area is classified as reserve and the area is not accessible for harvest. However, in practice, this area could be subject to thinning during the current
461
rotation. The amount extracted will subsequently be underestimated. This effect is very small owing to the combination of, on the one hand, limited areas subject to thinning in this treatment group and, on the other hand, the thinning intensity in Swedish forests. The same transition probabilities are used irrespective of treatment group. This means that growth under continuous forest cover–i.e., modified management–is projected with the same growth model as forest management under a final felling regime. There are limited data to support this assumption. The effects on the results should be very small, as the harvest volumes from this treatment group are limited. Diameter and species composition are not projected by the forest growth module. Instead, these parameters are estimated by regression functions from the present state of the forest, and are applied for the entire simulation period. This means that the diameter and species composition of a certain forest type do not change during simulations, irrespective of management. For instance, the model does not allow for management that improves the conditions for broadleaved species. The forest growth module inherent in the model probably yields too low an increment level, especially in the southern part of Sweden, as has been shown in other studies where the growth model has been employed (Sallna¨s, 1990). Thus, future harvest levels may be underestimated, notably in balance region 4. However, this should not greatly affect the relations between the scenarios.
6. Conclusions Given the assumptions defined in the study, some conclusions may be drawn. ! Under forest management regimes that follow the certification standards, the future harvest level will be significantly lower than the potential assessed by earlier, more biologically based, studies. ! An extended adaptation of Swedish forestry practices to the certification standards would, at constant prices, result in decreased supply from the Swedish forest, or in strong price increases if the domestic timber supply is to be kept at the present level.
462
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
! The positive short-term effect of a price increase on supplied quantity would not prevail in the long term. The ongoing adaptation of Swedish forestry to the standards of the certification programs will have substantial effects on the supply of timber. Ensuring the Swedish forest industry’s future access to domestic raw material may demand price changes that could affect its international competitiveness. Effects and tendencies like these should form integral parts of future analyses of wood balances and wood supply from a regional perspective.
Appendix A. Prices and associated data Timber prices Region Species Top diameter class (cm) 12 14 16 18 20 22 24 26 28 30+ BR1 BR2 BR3 BR4
Pine Spruce Pine Spruce Pine Spruce Pine Spruce
382 451 468 259 384 414 435 484 414 454 435 484 414 454 403 412 480 407 411 478
531 444 518 465 518 465 495 513
563 460 542 478 542 478 526 521
567 464 559 489 559 489 537 539
580 467 574 500 574 500 548 540
580 471 584 507 584 507 559 545
590 472 589 516 589 516 574 561
592 476 595 526 595 526 585 580
Pulpwood prices Region
Pine
Spruce
Hardwood
BR1 BR2 BR3 BR4
219 229 229 254
236 250 250 271
228 254 254 275
Harvesting and silvicultural costs (SEK per hour) Region BR1 BR2 BR3 BR4 Forwarder including operator Harvester in final felling including operator Harvester in thinning including operator Cleaning and precommercial thinning Scarification including operator Planting excluding plants Plants (SEK per 1000 plants)
357 818 612 203 934 170 2200
357 818 641 203 934 170 2200
357 818 678 162 1116 170 2200
357 818 720 162 1116 170 2200
References Berck, P., Bible, T., 1984. Solving and interpreting large-scale harvest scheduling problems by duality and decomposition. Forest Science 30, 173 – 182. Bra¨nnlund, R., 1988. The Swedish Roundwood Market. Report 82. Department of Forest Economics, SLU, Umea˚. Bra¨nnlund, R., Kristro¨m, B., 1993. Assessing the impact of environmental charges: a partial general equilibrium model for the forest sector. Environmental and Resource Economics 3, 297 – 312. Brealey, R.A., Myers, S.C., 2000. Principles of Corporate Finance, 6th ed. McGraw-Hill, Boston. Denardo, E., 1982. Dynamic Programming. Prentice-Hall, Englewood Cliffs, NJ. Eriksson, L.O., Sallna¨s, O., Sta˚hl, G., Imamovic, D., Pettersson, H., 2002. Wood supply from Swedish forests managed according to the FSC-standard. Working Paper, vol. 94. Department of Forest Resource Management and Geomatics, SLU, Umea˚. FSC council, 1998. Swedish FSC Standard for Forest Certification. Report from the Swedish FSC council. (www.fsc-sverige.org). Jacobsson, J., 2002. Assessing the Impact on Production and Costs from the Application of Forest Certification Standards. Pi Environmental Consulting, Pully, Switzerland, http:// www.piec.org/pathfinder/Pathfinder_portal/Instruments_Engl/ C4-economic/web/C4-screen.pdf. Johansson, P.-O., Lo¨fgren, K.G., 1985. The Economics of Forestry and Natural Resources. Basil Blackwell, New York. Lohmander, P., 1988. Pulse extraction under risk and a numerical forestry example. Systems Analysis, Modelling, Simulation 4, 339 – 354. Lundstro¨m, A., Nilsson, P., Sta˚hl, G., 1997. Certifieringens konsekvenser fo¨r mo¨jliga framtida uttag av industri- och energived. En pilotstudie. SLU, Department of Forest Resource Management and Geomatics. Working paper 23. Nabuurs, G.J., Pa¨ ivinen, R., Schanz, H., 2001. Sustainable management regimes for Europe’s forests—a projection with EFISCEN until 2050. Forest Policy and Economics 3 (3–4), 155 – 173. Na¨slund, M., 1947. Funktioner och tabeller fo¨r kubering av sta˚ende tra¨d: tall, gran och bjo¨rk i so¨dra Sverige samt i hela landet. Meddelanden fra˚n Statens skogsforskningsinstitut, vol. 36 (3) (Stockholm). Ollas, R., 1980. Nya utbytesfunktioner fo¨r tra¨d och besta˚nd. Ekonomi, vol. 5. Forskningsstiftelsen Skogsarbeten, Stockholm. Sallna¨s, O., 1990. A matrix growth model of the Swedish forest. Studia Forestalia Suecica, No. 183. SLU, Almqvist & Wiksell International, Stockholm. Sallna¨s, O., Eriksson, L.O., 1989. Management variation and price expectations in an intertemporal forest sector model. Natural Resource Modeling 3, 385 – 398. Skogsstyrelsen, 1999a. Skogsstatistisk a˚rsbok 1999 (Statistical Yearbook of Forestry 1999). Skogsstyrelsen (National Board of Forestry), Jo¨nko¨ping. Skogsstyrelsen, 1999b. Storskogsbrukets kostnader fo¨r a˚tga¨rder pa˚ egen skogsmark a˚r 1998. Jfr. nr. 1998 59. Skogsstyrelsen (National Board of Forestry), Jo¨nko¨ping.
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463 Skogsstyrelsen, 2000. Skogliga konsekvensanalyser 1999. Report No. 2. Swedish Board of Forestry, Jo¨nko¨ping. SLU, 1989. Omra˚deskalkyler-skog, mellersta Sverige. Rapport 151. Konsulentavdelningen, SLU, Uppsala. Svenska Skogsplantor, 1999. Svenska Skogsplantor 2000. Svenska Skogsplantor AB, Jo¨nko¨ping. Verkaik, E., Nabuurs, G.J., 2000. Wood production potentials of Fenno-Scandinavian forests under nature-orientated management. Scandinavian Journal of Forest Research 15, 445 – 454.
463
˚ rsredovisning fo¨r verksamhetsa˚ret 1997/98. VMF VMF, 1998. A Nord ek. fo¨r., Umea˚; Wermla¨ndska inma¨tningsfo¨reningen ek. fo¨r., Karlstad; Virkesma¨tningsfo¨reningen Syd ek. fo¨r., Jo¨nko¨ping. Washburn, C.L., Binkley, C.S., 1990. Informational efficiency of markets for stumpage. American Journal of Agricultural Economics 72, 394 – 405.
Forest certification and Swedish wood supply Ljusk Ola Eriksson a,*, Ola Sallna¨s b,1, Go¨ran Sta˚hl a,2 a
Department of Forest Resource Management and Geomatics, SLU, SE-901 83 Umea˚, Sweden b Southern Swedish Forest Research Centre, P.O. Box 49, SLU, SE-230 53 Alnarp, Sweden
Received 19 April 2004; received in revised form 20 October 2005; accepted 8 November 2005
Abstract A number of measures have been introduced into Swedish forestry in order to satisfy demands on biodiversity and sustainability. Protection measures include set asides, areas with continuous cover forestry and retention trees on harvesting sites, to name but a few. Most of these practices will have implications for the total wood supply of the Swedish forests. In 1998, the Swedish standard for forest certification according to the Forest Stewardship Council (FSC) was approved. The standard adopts many detailed regulations regarding how forests should be managed. The aim of this study is to assess the likely effect of FSC certification on short- and long-term supply of roundwood in Sweden. A scenario, expressed as the distribution of the forest land base on production forests, areas with restrictions for management and reserves, is input into a simulation model in which forest owners are assumed to be guided by economic criteria when deciding on management. Four scenarios were created based on the data of the Swedish National Forest Inventory. For each scenario the short-term wood supply was assessed. Also, a number of long run simulations with different relative price levels are presented. The results indicate that full adoption of the FSC standard on the entire land base, compared with adherence only to the Forestry Act, could result in a substantial reduction in supply, or, conversely, could induce a price increase in case supplied quantities should be maintained at current levels. This is under the assumption that no compensating mechanisms, exogenous to the model, come into effect. Furthermore, a sustained price increase that would compensate for lost volumes today does not seem to prevail in the long run. In conclusion, the ongoing adaptation of Swedish forestry to the standards of the certification programs could have substantial effects on the timber supply and will probably influence the international competitiveness of the Swedish forest sector. Effects and tendencies like these should form integral parts of future analyses of wood balances and wood supply in a regional perspective. D 2006 Elsevier B.V. All rights reserved. Keywords: Long range projection; Forest sector; Price expectations; Markov model
* Corresponding author. Tel.: +46 90 7865840; fax: +46 90 778116. E-mail addresses: [email protected] (L.O. Eriksson), [email protected] (O. Sallna¨s), [email protected] (G. Sta˚hl). 1 Tel.: +46 4 415181; fax: +46 40 462325. 2 Tel.: +46 90 7865837; fax: +46 90 778116. 1389-9341/$ - see front matter D 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.forpol.2005.11.001
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
1. Introduction The Swedish standard for forest certification according to Forest Stewardship Council (FSC) principles and criteria was approved in 1998 (FSC council, 1998). Within a short period, all major forest companies in Sweden had adopted the standard and modified their management practices. The overall aim of the standard is to outline management principles that maintain the ecosystem’s productivity and biodiversity, secure local people’s livelihood, and promote long-term valuable wood production. In particular, the standard comprises many detailed regulations that determine how forests should be managed. For example, 5% of forest areas should be set aside for free development or be managed to promote biological values. Certain areas must be managed to promote deciduous forests, and boundary zones to streams, lakes, and non-productive land should be maintained, as should also a certain minimum number of trees (10 ha 1) during harvesting operations (FSC council, 1998). Compared with management practices of most companies before adoption of the FSC standard, the new management regimes imply lower levels of potential harvest. Some studies have quantified the effects of the new management on harvesting levels (e.g., Lundstro¨m et al., 1997; Skogsstyrelsen, 2000; Jacobsson, 2002). These studies show the range of harvesting reduction due to the FSC standard to be between 10% and 20%. Similar results are presented by Verkaik and Nabuurs (2000) in a study of the consequences of adopting bnature-orientatedQ management in Scandinavian forests, and by Nabuurs et al. (2001) analyzing the effects of a multi-purpose management scenario for European forests. However, in the latter studies, no attempts were made to apply the FSC standard as such. Long-term sustainable harvesting levels in Sweden before the adoption of the FSC criteria have been assessed to be slightly more than 100 million m3 per year (Skogsstyrelsen, 2000). The annual cutting levels during the last decade have been about 70–75 million m3 per year. There are several reasons for the relatively large difference between the two figures. First, many forests in Sweden are still relatively young, and thus not mature for final felling. Second, we are dealing with different concepts here. The level
453
around 100 million m3 is what would be a sustainable harvest from the forests from a biological point of view, reflecting the potential yield. In this calculation, no restrictions of a technical, social or economic kind are taken into account. The recorded figure of 70– 75 million m3 reflects the real supply of wood, given the economic setting and restrictions on forestry of different kinds. In most of the above studies of the effect of the FSC standard on harvesting levels, the modeling approach has been to control management activities by a set of rules formulated to mimic certain forest management practices. These rules are applied even if a deeper analysis (which is typically not made) for some forestry activities would imply negative results from the economic point of view. Together this implies that these studies are dealing with the bpotentialQ discussed above. A different approach is to put forestry into an economic framework, allowing harvesting activities to be controlled largely by economic considerations. The result of such an analysis most likely would be a different supply of roundwood, leading to lower harvesting levels compared with the levels that correspond to the biological potential. This kind of analysis would provide a better assessment of the consequences on harvesting levels of new demands on forestry; e.g., those caused by FSC certification. The aim of this study is to analyze the implications of the FSC standard for the supply of roundwood in Sweden with respect to both the effects on current supply relationships and the long-term level of harvesting. Two different assumptions about the rigor of the FSC standard are analyzed. The study contrasts these two scenarios with a modified version of the environmental restrictions of the recent SKA99 study (Skogsstyrelsen, 2000) used as a baseline case. The constraints are less restrictive than in the SKA99 study and are intended to reflect forest management under the current Forestry Act, with no adherence to the particular stipulations of the FSC standard. Furthermore, an extreme case with no environmental considerations is included in order to assess the maximum economic potential. Whereas the Swedish Forestry Act establishes that environmental and production goals are equally important, it contains very few binding requirements of, e.g., leaving trees and stands of the kind stipulated by
454
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
the FSC certification standards. Thus, the environmental concerns in a scenario adopting the minimum requirements of the Forestry Act are not as high as in a scenario adopting the certification standards. The different levels of environmental objective form the basis for four scenarios. The scenarios differ with respect to the areas available for production or assigned for modified management and reserves. Except for the distribution of the forest area for permissible management, the economic and other conditions remain the same. The analysis uses a model of forest management decisions developed by Sallna¨s and Eriksson (1989). The model derives economically optimal harvesting regimes for all types of forest given certain assumptions regarding prices and costs.
2. Methods 2.1. Basic model concepts The results of this report rest on solutions to a forest management problem that confronts the forest owner. The outcome of different projections is the aggregate of the owners’ actions. We begin with the basic assumptions behind the management problem, then present the different models and solution procedures in the next section. The forest owner seeks to maximize monetary profits; i.e., the net present value of the forest holding. Three assumptions will underpin the representation of the management problem in the model. First, when calculating the optimal treatment for a certain piece of land at a certain time, we assume that the owner expects the current price level to persist, an assumption that should be consistent with efficient timber markets. Second, we assume that all forest owners encounter the same economic conditions. This means that the net revenue of a treatment of a certain piece of land, whether in establishment or harvesting, should be the same irrespective of ownership category. From this assumption, it also follows that financial markets must be efficient, as all owners would face the same interest rate. Third, it is assumed that decisions by the forest owner are not affected by considerations that span the whole forest estate, such as non-declining harvest constraints or objectives related to the composition of species and ages of the entire estate.
We use a stage-structured Markov model. These assumptions will then have the following implications for modeling profit maximizing by Swedish forest owners. The third assumption implies that the problem is separable. Given a set of prices, an optimal solution to the forest management problem is derived by selecting an optimal treatment for each piece of land separately (Johansson and Lo¨fgren, 1985). With a Markov model, each state of the model may be inspected one-by-one while still preserving optimality. The second assumption, together with the third, allows all forest belonging to the same forest state to be treated in the same (optimal) way. The first assumption implies that, for a given price level, the optimal decision will be the same for forests in a given state regardless of when it is made, as historic prices are irrelevant. One may note that, while the second and third assumptions are more technical in nature and primarily serve the purpose of simplifying the administration and running of the model, the constant price assumption could have fundamental consequences for the kind of results that emerge (Sallna¨s and Eriksson, 1989). 2.2. Forest model and solution procedure The solution procedures are built on the dynamics of the forest model (Sallna¨s, 1990). The Markov forest model is given by X xtþ1 ¼ ytik Tijk 8 ja I ð1Þ j ik
xti ¼
X
ytik
8 ia I
ð2Þ
k
where xti is the area residing in state i in period t, y ikt is the area in state i subject to treatment k in period t, and Tijk is the probability for an area residing in state i in period t to be found in state j in period t + 1 when subject to treatment k. Thus, (1) gives the transition of areas between states between period t and t + 1 as a function of the treatment of the forest and (2) ensures that the whole forest area is allocated a treatment (no action is also regarded as a treatment). The total set of states is I and the stage duration is 5 years. The forest area is divided into three broad treatment groups depending on what kinds of treatment are allowed. The amount that belongs to each
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
group depends on what scenario is analyzed (see Section 3.2). Within each treatment group, the forest area is differentiated between established forest (forest with an average height above 6 m) and young forest (including bare land). The state space of established forest is defined by geographical region, owner category (non-industrial or other), site quality, species composition, age, volume, thinning state (thinned or not thinned in the previous period) and comprises a total of 51,840 states. The state space of young forest is delineated by geographical region, owner category, site quality and age, and is composed of 384 states. The forest management problem in period t consists of maximizing the net present value; i.e., ! X X max nik þ d Tijk ej ytik ð3Þ jaI
ik
455
state of the forest in terms of the number of hectares in each state of the Markov model, x1i . 2. Set the economic conditions in terms of, for instance, discount rate and timber prices, and compute net revenues, n ik , and expected values, e i . 3. Make a projection for a number of 5-year periods by (i) solving the management problem for period 1 with Eqs. (3)–(5), (ii) projecting the forest into period 2 with Eq. (1), and then repeating steps (i) and (ii) for the number of periods that the projection should encompass. Once the projection has been made for all treatment groups, the result is consolidated to cover the whole forest area. By varying the amount of area distributed on different treatment groups or by changing the economic parameters, different developments may be created.
subject to xti ¼
X
ytik
8ia I
ð4Þ
k
ytik a Ki
3. Data 3.1. Regions
8 ia I
ð5Þ
where n ik is the net revenue of treatment k in forest belonging to state i, d is the one-period discount factor and e i is the expected value of forest and land of state i. Eq. (4) is equivalent to (2), and (5) ensures that the treatment of each state i belongs to the set of permissible treatments K i . The expected value of each class i, e i , is derived by solving the following set of equations. ( ei ¼ maxkaKi nik þ
X
) dTijk ei
8 ia I
ð6Þ
j
Thus, the expected values are derived under an infinite time horizon, assuming optimal treatments and constant prices. System (6) is solved with a successive approximation algorithm (Denardo, 1982). The first step in using the model is to distribute sample plot data of the forest on the three different treatment groups depending on scenario. Then, for each treatment group, the following steps are completed. 1. From the sample plots, compute the area residing in each state; i.e., obtain a description of the initial
Calculations are done for the four balance regions as defined in SKA99 (Skogsstyrelsen, 2000), here denoted by BR1–BR4. The regions correspond well to the regions of the growth model (Sallna¨s, 1990). Certain functions, such as those for reduction for bark, are differentiated on Northern and Southern Sweden. BR1 and BR2 are here classified as Northern Sweden, and BR3 and BR4 as Southern Sweden. Other data, such as minimum final felling age, need to be referred to by even smaller regions. Here, BR1 will be represented by the county of Va¨sterbotten, BR2 by Dalarna, BR3 by Va¨rmland, and BR4 by Jo¨nko¨ping. 3.2. Scenarios The forest owners are constrained in the choice of treatments by the endowment of forest areas in different treatment groups. There are three such groups: production forest, forest with nature-oriented modified management, and reserves (this corresponds to the groups bTraditionell sko¨tselQ, bNaturanpassad sko¨tselQ and bIngen avverkningQ in SKA99 (Skogsstyrelsen, 2000), respectively). A given distribution of the Swedish forest area among the treatment groups is
456
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
here termed a scenario. The distribution of the forest among treatment groups for a given scenario is static; i.e., the initial distribution of treatment groups for a given scenario remains constant throughout the entire projection period. Production forests correspond to forests where both thinning and final felling are allowed. Minimum final felling age is set according to the previous Forestry Act, which for normal site conditions is about 80 years. This gives permissible ages about 10 years higher than in current regulations, which is motivated by our impression that many forest owners, including large forest companies, owing to environmental concerns, economic considerations or a need to ration forest for final felling apply rules that are stricter than the current implementation of the Forestry Act. The model allows thinning only for age classes that are not permissible for final felling. Final felling is in the form of clear felling. The treatment group with modified management differs from the first by allowing only thinning. Thinning is permissible for all established forest irrespective of age. Forests set aside as reserves have neither thinning nor final felling. All treatment groups share the same growth model given by Eq. (1). A data set consisting of 21,301 sample plots of the National Forest Inventory (NFI) from 1996, 1997 and 1998 was prepared. The area is 22.4 million ha, which corresponds to the productive forest area in Sweden, reduced by 1 million ha of existing and planned reserves with legal protection. The planned reserves are based on an estimate made in the SKA99 study of the likely area to be protected up to about 2010 (Skogsstyrelsen, 2000); in the projections, they remain protected over the entire planning horizon. The areas associated with the NFI plots on treatment groups for a particular scenario were distributed as follows. The first step established the area for each region and owner category that should belong to a particular treatment group. Categories should, although they have identical economic conditions, be kept separate since they are differentiated in the growth model. The next step derived the distribution for each region and category of the forest area on values for two categorical sample plot variables describing the importance of the plot for nature conservation. One variable is more closely related to the legal status and usage of the plot, and
will below be denoted S, and the other to the state of the tree layer, and is below denoted F. Both variables have values ranging from 1 to 6 where 1 signifies the highest priority for retention and the 6 signifies no specific values. A particular combination of the categorical variables determined what treatment group, or groups in boundary cases, the area represented by a sample plot should belong to such that the target area established in the first step was met. The procedure was somewhat complicated due to the definitions associated with the variables. The delineation was done as follows for each region and category for reserves: all plots with F = {1,2} and all plots with S = 1 were allocated to reserves. Then, as many plots as needed with S = 2, in increasing order of variable F, was set aside. The procedure was then followed for S = 3, etc. With a few, and in area insignificant, exceptions, the following procedure was followed on the remaining plots for each region and category for modified management: as many plots as needed with S = 2, in increasing order of variable F, was allocated to modified management. The procedure was then followed for S = 3. Then, plots were allocated in order of minimum sum S + F. Ties–S,F = (4,4) has the same sum as S,F = (5,3)–were broken by giving priority, first, to the plot with minimum value on min(S,F), and second, on taking plots with a higher value on the F variable before taking plots with this value on the S variable. After this, the remaining plots were placed in the production forest group (the detailed results for each region and category are given in Eriksson et al., 2002, Appendix 2). Four scenarios are studied: the NUL, BAS, FSC and FSC+ scenarios. The principles behind the distribution of areas into treatment groups for different scenarios are the following. The NUL scenario is included here as a reference in order to assess the maximum economic potential, and lacks practical significance in today’s forestry. Observe that this scenario, as the other scenarios, excludes areas of existing and until year 2010 planned reserves as was stated above. Whereas the BAS scenario is intended to include only environmental considerations induced by the Forestry Act, the SKA99 study reflects the environmental restrictions applied by Swedish forestry at the end of the 1990s. Subsequently, the SKA99 study
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
encompasses forest owners that already, to a smaller or larger extent, have implemented the FSC or some other certification scheme. The distribution of areas for the BAS scenario follows the SKA99 study (Skogsstyrelsen, 2000) as regards areas with natureoriented modified management. The areas designated as reserves in the BAS scenario follow what in the SKA99 study are set aside as patches and biotopes demanding care, amounting to about 2% for each. However, the BAS scenario omits the effect of retention trees in the SKA99 study, rendering this scenario less restrictive than the SKA99 study. The BAS scenario therefore represents forestry with no adherence to the particular stipulations of the FSC standard. The FSC scenario is intended to reflect a situation where the FSC certification standard is applied to all forests in Sweden. The FSC standard prescribes that at least 5% of the productive forest area should be set aside (2% in the SKA99 study). A further 3.5% constitutes smaller patches demanding care that should not be harvested (2% in the SKA99 study). The effect of retention trees left at the harvesting sites reduces the available area by an estimated 4.5%. This produces a total restricted area of 13%. The area with modified management remains essentially the same as in the BAS scenario, but is somewhat rearranged owing to the increased area requirements of reserves. The FSC+ scenario means that another 4% and 5%, compared with the FSC
457
Table 2 The distribution of the area of production forest on different site quality classes for different scenarios (%) Scenario
NUL BAS FSC FSC+
Site quality classa 1
2
3
4
11 10 10 10
26 25 25 24
29 30 30 31
33 35 35 35
Class is defined in m3 ha 1 year 1 , potential mean increment, for all classes except classes 1 and 2 (the least productive sites) in region 1 (Northern Sweden) where the definition is based on altitude (see Sallna¨s, 1990; Appendix 1). a
scenario, are added to the areas of modified management and reserves respectively. Following these principles and procedures, the ensuing distribution of areas is as shown in Table 1. We can note a minute tendency for the site quality composition of production forest to improve as areas are transferred from production forest to forest with modified management and reserves (Table 2). 3.3. Economic data Prices of timber and pulpwood were calculated based on averages for the years 1996–1998 for pine and spruce, whereas broadleaves only qualified as pulpwood. Pulpwood prices were collected from the Official Statistics of Sweden (Skogsstyrelsen, 1999a). Price lists for timber are for BR1 from Norra Skogsa¨garna,
Table 1 The distribution of the forest area into treatment groups (million hectares) Scenario NUL
BAS
FSC
FSC+
Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area Established forest Young forest Sum Share of total area
(%)
(%)
(%)
(%)
Production forest
Modified management
Reserve area
Sum
14.53 7.86 22.39 100 13.08 7.32 20.40 91 11.68 6.64 18.32 82 10.37 5.98 16.36 73
0 0 0 0 0.71 0.33 1.04 5 0.75 0.38 1.13 5 1.40 0.66 2.06 9
0 0 0 0 0.73 0.21 0.94 4 2.10 0.85 2.95 13 2.76 1.21 3.97 18
14.53 7.86 22.39 100 14.52 7.86 22.38 100 14.53 7.87 22.40 100 14.53 7.85 22.39 100
458
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
Table 3 Annual harvest during the first 10 years of simulation (million m3) at different discount rates under the BAS scenario Discount rate (%) 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 61.6 66.0 67.3 71.1 72.6 76.3 81.1 83.5 87.3 91.5 96.6
Umea˚ district, for BR2 and BR3 from Mellanskog and the price lists applicable for Dalarna and Va¨rmland, respectively, and for BR4 from So¨dra Skogsa¨garna, Jo¨nko¨ping district. Timber prices of standard assortment are given over diameter for five quality classes for pine and four classes for spruce in the price lists. They are weighted together into one price list for each species and region by using statistics on the distribution on quality from three timber measurements associations (VMF, 1998), assuming that the distribution of volumes over diameter classes does not differ between qualities. Log conversion is calculated using the method of Ollas (1980) with the relation of volume over and under bark given by Na¨slund (1947). Species composition and dimension are of importance for the economic return from harvests. The growth model includes three broad species classes: more than 50% conifers with pine dominating, more than 50% conifers with spruce dominating, and more than 50% broadleaves. For a better estimation of the species distribution in individual states of the model, species distribution functions were developed from NFI data (see Eriksson et al., 2002, Appendix 4). Average diameter functions from Eriksson et al. (2002, Appendix 5) are used as a basis for the application of the conversion functions in individual states of the growth model. Costs for cleaning, scarification, planting (excluding plants) and precommercial thinning on a perhectare basis are from Skogsstyrelsen (1999b) for Northern and Southern Sweden, respectively. Plant cost is from Svenska Skogsplantor (1999), with the number of plants differentiated by region and site index. Harvesting costs for thinning and final felling were estimated with functions for harvester and forwarder from SLU (1989). In order to account for increased productivity since 1987, the input data on cost per hour was reduced such that the cost per cubic meter
was the same for a tree with 20 cm diameter at breast height as the cost given by Skogsstyrelsen (1999a). (For details on prices, see Appendix A). The cost of capital–i.e., the discount rate–was estimated using the method presented by Berck and Bible (1984). Assuming that the model is correct, a real after-tax discount rate that gives model results coinciding with observed behavior is determined. Running the model for the first two 5-year periods under the BAS scenario shows that a discount rate of 2.5% generates a harvest level that quite closely coincides with the current harvest level in Sweden (Table 3). This discount rate will be used in subsequent analyses.
4. Results Results are here presented as an aggregation of growth and harvest of two stages instead of the separate 5-year periods for ease of interpretation; there is a variation in harvests, up and down, for the first two 5-year periods that make the single 5-year period results less relevant (see also explanation to Fig. 2). The total annual harvest during the first 10 years of simulation in the BAS scenario is 76.3 million m3 (Table 4)). This level is 9% higher than that of scenario FSC and 15% higher than FSC+, while the NUL alternative would yield some 7% higher harvest level than that of the BAS scenario. These differences are rather well reflected in the results for the individual balance regions, except for BR3, where we notice only a marginal difference between the FSC and FSC+ scenarios. In Fig. 1, the average supply for the first 10 years is given for a gradient of relative timber prices. One implication of the results is that in order to reach the same short-term harvesting level in the FSC scenario Table 4 Annual harvest for the first 10 years, for different scenarios Scenario
BR1
BR2
BR3
BR4
Total
NUL BAS FSC FSC+
110 27.1 89 84
106 18.4 91 85
104 7.7 92 91
105 23.1 91 84
107 76.3 91 85
The figures for scenario BAS are given in million m3 , and for the other scenarios in percent of scenario BAS.
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
459
Fig. 1. Annual supply during first 10 years for different scenarios and different relative prices.
as in the BAS scenario with relative price 1.0, the price level has to be increased by some 17%. The corresponding figure for the FSC+ scenario is just below 40%. Price elasticities provide another perspective on the same phenomenon. Calculating over a 20% interval on both sides of relative price 1.0, the elasticities range from 0.55 to 0.51, with the more benvironmentalQ scenarios in the lower end. As expected, the elasticities decrease with an increased price level. In Fig. 2, the development over time of annual harvests is given for the different scenarios at a price level of 1.0. The curves follow each other consistently. The dip in the second period is found for all scenarios. This is probably due to an adjustment of the forest estate to the economic assumptions of the model; existing over-mature forest, from an economic
Fig. 3. Harvests over time for some scenarios at different price levels as a relative difference from the level of the BAS scenario at price level 1.0.
point of view, is harvested in the first period, leading to a reduced supply in the next period. During the rest of the century we find a rather steady increase in harvests. After a decrease for some periods, the harvests reach the former high level at the end of the simulations. The price effects on harvest levels illustrated in Fig. 1 are in the short term, in this case the first 10 years. In Fig. 3, the long-term harvest development for a number of selected combinations of scenario and price level are compared to that of the BAS scenario at price level 1.0. Although it is possible even in the FSC scenario to reach the BAS harvest level in the beginning of the simulation by increasing the price, the subsequent decrease in harvest level reaches about 10%. If we sum the harvests over 100 years in the different scenarios and compare them with the harvests of scenario BAS with price 1.0, we obtain the results shown in Table 5. Prices affect mainly the distribution of harvests over time. It is evident that the total harvest over the 100 years is more or less the Table 5 Average harvest over 100 years in relation to the BAS scenario for different price levels BAS
Fig. 2. Annual harvest over time for the four different scenarios at a relative price level of 1.0.
FSC
FSC+
Relative 1.0 1.1 1.2 1.0 1.1 1.2 1.0 1.1 1.2 price level Relative 1.00 0.99 0.99 0.91 0.91 0.90 0.87 0.86 0.86 harvest
460
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
Fig. 4. The percentage of final fellings of the total harvests for different scenarios (price level 1.0).
same irrespective of the price level, with a slight tendency for high initial harvests associated with higher price levels resulting in a somewhat lower total. Another result is that the FSC scenario bcostsQ 10–11% of the harvest level of BAS at all price levels. The corresponding figure for the FSC+ scenario is 13–14%. A marked difference between the different scenarios is the part of the total harvest that stems from final felling as shown in Fig. 4. In the FSC+ scenario, the area allocated to modified management, and consequently to thinning as the only available management option, is larger than in the BAS and FSC scenarios. Consequently the thinning part of the harvest increases. The initial peak in harvest in Fig. 2 is also found in this figure with respect to final harvest. This underlines the interpretation that in the first period the state of the forest is adjusted to the economic assumptions of the model; those forest areas that are over-mature in relation to the economic conditions are harvested.
5. Discussion Comparing the scenarios in terms of harvests, the first 10 years show that the management of today, expressed as the BAS scenario, features a supply level that is 7% lower than that of a situation where all forest land is used entirely for timber production. The FSC and FSC+ scenarios compared with BAS feature 10% and 15% lower levels, respectively. These differences correspond well with the fraction of the
forest area that is assigned to uses other than pure timber production. This observation corresponds well with the fact that the areas transferred from production forest to modified management and reserves has about the same site quality composition as the areas retained for production (see Table 2). The relatively low elasticities, even at a price level of 1.0, imply that to compensate in the short run for the decrease of supply that would follow from the extended conservation measures in the FSC scenario, prices would have to rise substantially. Other compensating mechanisms include increased imports and reduced wood consumption. Whatever effects prevail, the implication is that extended conservation measures would increase the strain on the wood consuming industry. A number of other assumptions, which range from basic economic assumptions to details of the submodels, need to be scrutinized in order to appreciate correctly the results. The scenarios of the study are, of course, more-or-less theoretical constructs. The BAS scenario is intended to reflect management that only meets the requirements of the current Forestry Act. There are only limited data to confirm that the ensuing result is what would have been forestry without certification. The FSC scenario has a better-founded basis in the stipulations of the FSC standard, although it is questionable how far the standard will go beyond the Forestry Act. It should be emphasized that the BAS scenario should not be confused with current forestry practice as more than half the Swedish forest area is already subject to FSC or other certification standards. The economic model used in this study uses two important assumptions. Forest managers are supposed to act according to beconomic manQ rationality; i.e., they are in every situation supposed to take the action that maximizes the present net value of their forestry. This is a simplification, which however can be motivated by the fact that the study should not primarily be interpreted as a prognoses; it is rather a model for clarifying the relationships between economic entities, such as harvest volumes and timber income, on the one hand, and environmental policies on the other. The second assumption is that the forest manager at every decision making step expects prices and costs not to change in the future. This assumption, that markets exhibit a rational expectations equilibrium, is
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
supported by a large number of investigations of financial markets (Brealey and Myers, 2000). It has also found support in studies of timber markets (Washburn and Binkley, 1990), although contradictory results have been reported (Lohmander, 1988). The model gains credibility when, as expected, its elasticities decrease with rising prices and consequently higher supply. Also, the elasticities, around 0.5, are of the same magnitude as those estimated from empirical data; Bra¨ nnlund (1988) presents estimates between 0.4 and 0.8 and Bra¨nnlund and Kristro¨m (1993) a figure in the range 0.1 to 0.2. At a discount rate of 2.5%, the modeled harvest is in line with current harvests in Sweden. It seems reasonable to seek a harvest level by choosing a realistic discount rate. The model’s real after-tax discount rate of 2.5% corresponds closely to empirical data on returns on investments for comparable assets (Brealey and Myers, 2000). The reasonableness of the combination of discount rate and harvest level also gives credibility to the model as such. However, the model’s distribution of harvests over balance regions does not fully reflect the factual situation in Sweden today. Using the ranking of the plots as the basis for distribution among treatment groups implies that forest areas with the highest conservation value will get the most extended protection. The distribution probably becomes more efficient from a nature conservation point of view than is possible in practice. In the latter case, effects of estate borders, lack of accurate data and other factors will also influence choice. A related issue is the assumption that the distribution of forest between prescribed areas will be permanent throughout time. This means that the setting aside of forest areas for conservational purposes or the assignment of specific management programs is unchanged through the simulations. Consequently no areas grow binQ or boutQ of a treatment group, although some of the criteria used for distributing the area refer to variables that are connected to the tree layer and thereby could change over time. This is probably of limited importance with respect to the results of the study. Retention trees have, in the model, the effect that a certain area is classified as reserve and the area is not accessible for harvest. However, in practice, this area could be subject to thinning during the current
461
rotation. The amount extracted will subsequently be underestimated. This effect is very small owing to the combination of, on the one hand, limited areas subject to thinning in this treatment group and, on the other hand, the thinning intensity in Swedish forests. The same transition probabilities are used irrespective of treatment group. This means that growth under continuous forest cover–i.e., modified management–is projected with the same growth model as forest management under a final felling regime. There are limited data to support this assumption. The effects on the results should be very small, as the harvest volumes from this treatment group are limited. Diameter and species composition are not projected by the forest growth module. Instead, these parameters are estimated by regression functions from the present state of the forest, and are applied for the entire simulation period. This means that the diameter and species composition of a certain forest type do not change during simulations, irrespective of management. For instance, the model does not allow for management that improves the conditions for broadleaved species. The forest growth module inherent in the model probably yields too low an increment level, especially in the southern part of Sweden, as has been shown in other studies where the growth model has been employed (Sallna¨s, 1990). Thus, future harvest levels may be underestimated, notably in balance region 4. However, this should not greatly affect the relations between the scenarios.
6. Conclusions Given the assumptions defined in the study, some conclusions may be drawn. ! Under forest management regimes that follow the certification standards, the future harvest level will be significantly lower than the potential assessed by earlier, more biologically based, studies. ! An extended adaptation of Swedish forestry practices to the certification standards would, at constant prices, result in decreased supply from the Swedish forest, or in strong price increases if the domestic timber supply is to be kept at the present level.
462
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463
! The positive short-term effect of a price increase on supplied quantity would not prevail in the long term. The ongoing adaptation of Swedish forestry to the standards of the certification programs will have substantial effects on the supply of timber. Ensuring the Swedish forest industry’s future access to domestic raw material may demand price changes that could affect its international competitiveness. Effects and tendencies like these should form integral parts of future analyses of wood balances and wood supply from a regional perspective.
Appendix A. Prices and associated data Timber prices Region Species Top diameter class (cm) 12 14 16 18 20 22 24 26 28 30+ BR1 BR2 BR3 BR4
Pine Spruce Pine Spruce Pine Spruce Pine Spruce
382 451 468 259 384 414 435 484 414 454 435 484 414 454 403 412 480 407 411 478
531 444 518 465 518 465 495 513
563 460 542 478 542 478 526 521
567 464 559 489 559 489 537 539
580 467 574 500 574 500 548 540
580 471 584 507 584 507 559 545
590 472 589 516 589 516 574 561
592 476 595 526 595 526 585 580
Pulpwood prices Region
Pine
Spruce
Hardwood
BR1 BR2 BR3 BR4
219 229 229 254
236 250 250 271
228 254 254 275
Harvesting and silvicultural costs (SEK per hour) Region BR1 BR2 BR3 BR4 Forwarder including operator Harvester in final felling including operator Harvester in thinning including operator Cleaning and precommercial thinning Scarification including operator Planting excluding plants Plants (SEK per 1000 plants)
357 818 612 203 934 170 2200
357 818 641 203 934 170 2200
357 818 678 162 1116 170 2200
357 818 720 162 1116 170 2200
References Berck, P., Bible, T., 1984. Solving and interpreting large-scale harvest scheduling problems by duality and decomposition. Forest Science 30, 173 – 182. Bra¨nnlund, R., 1988. The Swedish Roundwood Market. Report 82. Department of Forest Economics, SLU, Umea˚. Bra¨nnlund, R., Kristro¨m, B., 1993. Assessing the impact of environmental charges: a partial general equilibrium model for the forest sector. Environmental and Resource Economics 3, 297 – 312. Brealey, R.A., Myers, S.C., 2000. Principles of Corporate Finance, 6th ed. McGraw-Hill, Boston. Denardo, E., 1982. Dynamic Programming. Prentice-Hall, Englewood Cliffs, NJ. Eriksson, L.O., Sallna¨s, O., Sta˚hl, G., Imamovic, D., Pettersson, H., 2002. Wood supply from Swedish forests managed according to the FSC-standard. Working Paper, vol. 94. Department of Forest Resource Management and Geomatics, SLU, Umea˚. FSC council, 1998. Swedish FSC Standard for Forest Certification. Report from the Swedish FSC council. (www.fsc-sverige.org). Jacobsson, J., 2002. Assessing the Impact on Production and Costs from the Application of Forest Certification Standards. Pi Environmental Consulting, Pully, Switzerland, http:// www.piec.org/pathfinder/Pathfinder_portal/Instruments_Engl/ C4-economic/web/C4-screen.pdf. Johansson, P.-O., Lo¨fgren, K.G., 1985. The Economics of Forestry and Natural Resources. Basil Blackwell, New York. Lohmander, P., 1988. Pulse extraction under risk and a numerical forestry example. Systems Analysis, Modelling, Simulation 4, 339 – 354. Lundstro¨m, A., Nilsson, P., Sta˚hl, G., 1997. Certifieringens konsekvenser fo¨r mo¨jliga framtida uttag av industri- och energived. En pilotstudie. SLU, Department of Forest Resource Management and Geomatics. Working paper 23. Nabuurs, G.J., Pa¨ ivinen, R., Schanz, H., 2001. Sustainable management regimes for Europe’s forests—a projection with EFISCEN until 2050. Forest Policy and Economics 3 (3–4), 155 – 173. Na¨slund, M., 1947. Funktioner och tabeller fo¨r kubering av sta˚ende tra¨d: tall, gran och bjo¨rk i so¨dra Sverige samt i hela landet. Meddelanden fra˚n Statens skogsforskningsinstitut, vol. 36 (3) (Stockholm). Ollas, R., 1980. Nya utbytesfunktioner fo¨r tra¨d och besta˚nd. Ekonomi, vol. 5. Forskningsstiftelsen Skogsarbeten, Stockholm. Sallna¨s, O., 1990. A matrix growth model of the Swedish forest. Studia Forestalia Suecica, No. 183. SLU, Almqvist & Wiksell International, Stockholm. Sallna¨s, O., Eriksson, L.O., 1989. Management variation and price expectations in an intertemporal forest sector model. Natural Resource Modeling 3, 385 – 398. Skogsstyrelsen, 1999a. Skogsstatistisk a˚rsbok 1999 (Statistical Yearbook of Forestry 1999). Skogsstyrelsen (National Board of Forestry), Jo¨nko¨ping. Skogsstyrelsen, 1999b. Storskogsbrukets kostnader fo¨r a˚tga¨rder pa˚ egen skogsmark a˚r 1998. Jfr. nr. 1998 59. Skogsstyrelsen (National Board of Forestry), Jo¨nko¨ping.
L.O. Eriksson et al. / Forest Policy and Economics 9 (2007) 452–463 Skogsstyrelsen, 2000. Skogliga konsekvensanalyser 1999. Report No. 2. Swedish Board of Forestry, Jo¨nko¨ping. SLU, 1989. Omra˚deskalkyler-skog, mellersta Sverige. Rapport 151. Konsulentavdelningen, SLU, Uppsala. Svenska Skogsplantor, 1999. Svenska Skogsplantor 2000. Svenska Skogsplantor AB, Jo¨nko¨ping. Verkaik, E., Nabuurs, G.J., 2000. Wood production potentials of Fenno-Scandinavian forests under nature-orientated management. Scandinavian Journal of Forest Research 15, 445 – 454.
463
˚ rsredovisning fo¨r verksamhetsa˚ret 1997/98. VMF VMF, 1998. A Nord ek. fo¨r., Umea˚; Wermla¨ndska inma¨tningsfo¨reningen ek. fo¨r., Karlstad; Virkesma¨tningsfo¨reningen Syd ek. fo¨r., Jo¨nko¨ping. Washburn, C.L., Binkley, C.S., 1990. Informational efficiency of markets for stumpage. American Journal of Agricultural Economics 72, 394 – 405.