Transcript of "A mill specific-rounwood_demand_equation_for_southern_and_central_finland"
ARTICLE IN PRESS Journal of Forest Economics 11 (2005) 95–106 www.elsevier.de/jfeA mill-speciﬁc roundwood demand equation for southern andcentral FinlandOlaf SchwabÃ, Gary Bull, Thomas ManessDepartment of Forest Resources Management, University of British Columbia, 2045-2424 MainMall, Vancouver, BC, Canada V6T 1Z4Received 10 December 2003; accepted 4 May 2005Abstract The majority of the roundwood processed by the highly concentrated forest productsindustry in Finland is supplied by non-industrial private forest owners (NIPF). The industry’sheavy reliance on NIPF roundwood supplies and the NIPF owners’ high dependency on theindustry for revenue motivated this study of the spatial ﬁbre ﬂows in regional markets. Todescribe the direction and magnitude of these regional ﬁbre ﬂows we estimate a mill-speciﬁctimber demand equation. This empirical model of roundwood demand can be used as abenchmark for identifying inefﬁciencies in wood procurement procedures. This study expandson the theoretical and empirical literature by increasing the spatial resolution of timberdemand estimates.r 2005 Elsevier GmbH. All rights reserved.JEL Classiﬁcation: Q230Keywords: Finland; Non-industrial private forestry; Roundwood demand estimation; Spatialresolution ÃCorresponding author. Tel.: +001 604 822 0921; fax: +001 604 822 9106. E-mail address: email@example.com (O. Schwab).1104-6899/$ - see front matter r 2005 Elsevier GmbH. All rights reserved.doi:10.1016/j.jfe.2005.05.001
ARTICLE IN PRESS96 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106Introduction Approximately three quarters of the forest land in Finland is owned by non-industrial private forest owners (NIPF).1 More than 350,000 NIPF owners supplyover 90% of the wood processed by the three major forest products companies inFinland. The relationship between the NIPF owners and the forest products industrycan be characterized as mutually dependent. The strong reliance of the forestindustry on NIPF timber harvests has created substantial concerns about thesecurity of current and future timber supplies. These concerns are centred on theshift from traditional objectives (timber harvesting) to non-timber objectives such asmanaging forests for their aesthetic values and recreational use. These new objectivesare gaining widespread acceptance among forest owners and could result insubstantial timber supply shortages in the future (Kuuluvainen et al., 1996). NIPF owners face two major challenges. First, they usually do not have theresources to market their timber outside the local area. Second, Finnish laws requireforest owners to provide free public access to resources such as recreationopportunities, aesthetic values and other non-timber forest products. Consequently,forest owners can generate revenue only from timber sales.2 Although the relevanceof forest-based income has been declining steadily over the last few decades, ruralhouseholds still rely heavily on timber sales to supplement their income (Karppinen,1998; Saastamoinen and Pukkala, 2001). For these NIPF owners, identifying allpotential customers is essential for negotiating proﬁtable timber sales becausecompetition between the three major forest products companies and the smaller,independent sawmills will ensure competitive prices. A comprehensive review ofstudies related to NIPF can be found in Amacher (2003). In contrast to existing roundwood supply models, the spatial resolution ofroundwood demand models has been relatively low. Demand data that wasaggregated at the regional or national level was commonly used (Brannlund, 1989; ¨Toppinen and Kuuluvainen, 1997; Bergman and Nilson, 1999; Ronnila andToppinen, 2000; Kallio, 2001). Models using more disaggregated data on round-wood demand exist; however, these models still do not provide any information onthe source of the roundwood (Baardsen, 2000; Roos et al., 2001; Nyrud andBergseng, 2002; Nyrud and Baardsen, 2003). Therefore, the objective of this study isto identify the structural relationships that affect the direction and magnitude ofﬁbre ﬂows in regional roundwood markets by estimating a mill-speciﬁc timberdemand equation. Modeling the structure and dynamics of ﬁbre ﬂows in theseregional markets is becoming increasingly important with both the changingindustry structure and recent shifts in the demographics and objectives of NIPFowners. 1 NIPF are deﬁned as forest owners who do not own or control primary or secondary processingfacilities. 2 Hunting rights being the notable exception.
ARTICLE IN PRESS O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 97Methodology This section summarizes the steps taken in the analysis of previous work on timberdemand estimation, data collection, and ﬁnally, procedures for the estimation of themill-speciﬁc timber demand equation presented in this study. To analyze woodprocurement patterns of mills in southern and central Finland we used data providedby UPM-Kymmene Metsa, a subsidiary of UPM-Kymmene Oy. The Metsa group is ¨ ¨responsible for the purchase, harvest, and transportation of roundwood to UPM-Kymmene’s processing facilities (UPM-Kymmene Metsa, 2000). The dataset used ¨for this study contains spatial ﬁbre ﬂow information for all roundwood procured byUPM-Kymmene Metsa during the 2000 calendar year. In brief, roundwood was ¨supplied to 105 mills of which 31 were owned by the parent company UPM-Kymmene Oy. Fig. 1 shows the ﬁve timber supply areas: Kainuu, Ostrobothnia,Central Finland, South-East Finland, and Western Finland. The remainder of the logs utilized in Finnish manufacturing for the year 2000came either from Russian and Baltic States log imports (12%) or from logs alreadyin the mill inventory from the previous year (20%). For each timber supply area, theroundwood information is further segregated into six roundwood assortments:Norway spruce [Picea abies (L.) Karst.], Scots pine [Pinus sylvestris L.] and birch[Betula sp.] sawlogs; and spruce, pine, and birch pulpwood. Fig. 1. Timber supply areas in southern and central Finland.
ARTICLE IN PRESS98 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 Initially, the database contained 579 observations with a cumulative ﬁbre ﬂowof 20.5 million m3. However, since the geographic origin of roundwood procuredfrom imports and from mill inventories could not be determined (approximately32%); these observations were excluded from the analysis. Furthermore, sinceinformation on log prices were only available for South-East Finland, CentralFinland, and Ostrobothnia, ﬁbre ﬂows from Kainuu and Western Finland were alsoexcluded (approximately 27%). Therefore, the regression analysis was based on theremaining 177 observations, accounting for 8.4 million m3 (approximately 41% ofthe total ﬁbre ﬂow). To determine the average transportation distance from theharvesting site to the mill locations we assumed the centroid of each timber supplyarea as the point of origin for all timber procured from within this area.Transportation distances were measured following major roadways on a1:1,000,000 scale map. Table 1 summarizes the parameters that were selected foranalyzing regional ﬁbre ﬂows, as well as references to the timber demand studies thatthis selection was based on. The econometric model for the mill-speciﬁc timber demand equation and theexpected signs of the parameters are shown in Eq. (1), with being the error term.Volume ¼ f fDistance; Mill capacity; Roundwood price; Price volatilityg þ . (1) ðÀÞ ðþÞ ðÀÞ ðÀÞ Table 2 summarizes the parameter types, as well as the range of values before andafter transformations for both the dependent and the independent variables used inestimating the demand equation. The dataset was analysed using the statistical software SAS, release 8.02 (SAS,2001). Initially, ordinary least squares (OLS) regression was applied. To test for theapplicability of the Gauss-Markov assumptions for OLS we used a Breusch-Pagantest (Greene, 2003; Wooldridge, 2003), which indicated that the error term washeteroskedastic with a test statistic of 12.72, 4 degrees of freedom and p ¼ 0:01.Analysing the error term obtained from OLS estimation, we found that the errorvariance was proportional to the squared natural logarithm of the volume.Therefore, to correct for the heteroskedasticity, we re-estimated the demandequation using weighted least squares (WLS) regression with a weight of: wi ¼ 1= ln ðVolumeÞ2 . (2)For this WLS regression we obtained a Breusch-Pagan statistic of 5.87, with 4degrees of freedom, and p ¼ 0:25. The functional form speciﬁcations of the independent variables were tested usingthe Davidson-MacKinnon test (Wooldridge, 2003). Non-logarithmic forms of theseparameters were rejected since the t-statistic for the residuals were signiﬁcant. Sincethis type of functional form speciﬁcation test cannot be applied to a logarithmicdependent variable, we calculated a goodness-of-ﬁt measure following Wooldridge(2003). While an R2 of 0.29 was obtained using the linear form of the dependentvariable, an equivalent of 0.38 was calculated for the logarithmic form of thedependent variable.
ARTICLE IN PRESS O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 99Table 1. Parameter description and relation to other studiesParameter Description Data or proxies used ReferencesDistance Distance from Distance from Hultkrantz (1987), harvesting site to centroid of timber Faminov and Benson mill gate supply area to mill (1990), Palander (1995), gate Roehner (1996), Cea and ´ Jofre (2000), Troncoso and Garribo (2005)Mill capacity Physical processing Volume of Hultkranz and Aronsson capacity roundwood supplied (1989), Bergman and by UPM Kymmene Lofgren (1991), Brannlund ¨ ¨ Metsa¨ (1991), Brannlund (1993), ¨ ´ Cea and Jofre (2000), Kallio (2001), Størdal and Baardsen (2002), Troncoso and Garribo (2005)Roundwood Market price of UPM data on the Brannlund et al. (1985), ¨price different roundwood annual average Martinello (1985), Lofgren ¨ assortments roundwood price by and Ranneby (1987), species (birch, Hultkranz and Aronsson spruce, pine) and (1989), Bergman and assortment Lofgren (1991), Brannlund ¨ ¨ (pulpwood, sawlogs) (1991), Hetemaki and ¨ for each timber Kuuluvainen (1992), supply area Kuuluvainen et al. (1996), Toppinen and Kuuluvainen (1997), Toppinen (1998a), Gomez et al. (1999), Cea ´ and Jofre (2000), Latta and Adams (2000), Bolkesjo and Solberg (2003), Lundmark and Soderholm ¨ (2003)Price Standard deviation UPM data on Roehner (1996), Gomez etvolatility of roundwood prices annual roundwood al. (1999), Yin and within each timber price by species and Newman (1999) supply area assortment for 4–5 sub-units per timber supply area. To ensure that potentially endogenous variables as well as omitted variable bias donot affect the parameter estimates, we implemented an instrumental variableapproach using weighted 2-stage least squares (2SLS) estimation. We introducedharvesting costs and harvesting cost volatility as additional instrumental variables.
ARTICLE IN PRESS100 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106Table 2. Parameter types, descriptive statistics, and transformationsParameter Volume Distance Mill Roundwood Price Sawmill capacity price volatility size indexType Continuous Continuous Continuous Continuous Continuous IndicatorUnit m3 km m3 FIM/m3 — —Range 835–804,584 16–434 1,121–2,186,693 164.58–371.25 1.54–9.26 0;1(before transformation)Mean 47,627 152 358,666 288.50 11.08 —Standard deviation 97,150 82 527,425 76.05 6.76 —Transformation Logarithmic Logarithmic Logarithmic — — —Range 2.92–5.91 1.20–2.64 3.05–6.34 — — —(after transformation)For a discussion of why price-related variables may be endogenous in demand andsupply estimation see Latta and Adams (2000). Both harvesting costs and harvestingcost volatility meet the identiﬁcation condition of correlation with price volatility,and are therefore suitable instrumental variables (Wooldridge, 2003). Following2SLS estimation, we applied a RESET test (Pesaran and Taylor, 1999), whichindicated that the functional form was correctly speciﬁed with a test statistic of 1.51and p ¼ 0:13. We then performed a Hausman speciﬁcation test comparing theparameter estimates from WLS and 2SLS estimation (Greene, 2003; Wooldridge,2003). This test indicated that the WLS estimates are preferred over the 2SLSestimates with a Hausman test statistic of 0.53, 5 degrees of freedom, andPr4w2 ¼ 0:99. Based on the Hausman and RESET tests we can conclude that noneof the explanatory variables are endogenous, and that no substantial omittedvariable bias is present (Wooldridge, 2003).Results and discussion The seemingly imbalanced structure of roundwood markets in Finland wouldsuggest the existence of market imperfections; however, the majority of studies onthe competitiveness of these markets have rejected this hypothesis (Koskela andOllikainen, 1998; Toppinen, 1998a, b; Ronnila and Toppinen, 2000; Tilli et al., 2001).Therefore, it is reasonable to assume that the spatial ﬁbre ﬂows observed in Finlandare based on a sufﬁciently competitive market model. Table 3 summarizes themodeling results, including parameter estimates, standard errors, and levels ofstatistical signiﬁcance for estimating the mill-speciﬁc timber demand equation. The estimated equation has an adjusted R2 of 0.55, and a mean squared error of0.13. All parameter estimates show the expected sign. However, it should be notedthat, contrary to expectations, price did not have a statistically signiﬁcant effect onthe volume a mill purchases from a particular timber supply area. Based on theseparameter estimates the mill-speciﬁc demand equation can be written aslnðVolumeÞ ¼ 7:235 À 0:692 Ã lnðDistanceÞ þ 0:496 Ã lnðMill capacityÞ À 0:034 Ã ðPrice volatilityÞ þ 1:266 Ã ðSawmill size indexÞ þ . ð3Þ
ARTICLE IN PRESS O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 101Table 3. Modeling results for weighted least squares and weighted 2-stage least squaresestimationParameter Weighted least squares Weighted 2-stage least squares Estimate Standard Pr4jtj Estimate Standard Pr4jtj error errorIntercept 7.235 0.762 o0.0001 7.534 0.867 o0.0001ln(Distance) À0.692 0.133 o0.0001 À0.690 0.133 o0.0001ln(Mill capacity) 0.496 0.044 o0.0001 0.481 0.049 o0.0001Price volatility À0.034 0.012 0.0040 À0.046 0.020 0.0232Sawmill size index 1.266 0.384 0.0012 1.325 0.393 0.0009Wood procurement patterns can be described by the relationship of volume anddistance. As shown in Eq. (3) the volume of roundwood procured declines steadilythe further away the timber supply area is from the location of the mill. As expected,the majority of the roundwood is procured relatively close to the mill. Sinceroundwood is a relatively bulky, low volume commodity transportation costs areone of the main factors in determining the allocation (Troncoso and Garribo, 2005).In their simplest form, transportation costs consist of a ﬁxed component for loadingand unloading and a variable charge per kilometre traveled. In Europe, the mostcommonly used transportation systems are trucking, train transport, and barges/freight ships (Hecker, 2003). Although a switch in the dominant transportationsystem from trucking to train and barge for larger distances was expected, a Chowtest following Greene (2003) failed to detect any structural breaks in the distanceparameter. Two parameters describing mill characteristics (mill capacity, and sawmill sizeindex) were found to be statistically signiﬁcant. As expected, mill capacity ispositively related to the volume of roundwood procured within a particular timbersupply area. Since the total volume of roundwood available within the immediatearea around the mill is limited by the biophysical growing capacity of the forest andthe harvesting rates, mills with a high processing capacity are forced to procureroundwood from within a larger wood procurement area to satisfy their volumerequirements. The sawmill size index is an indicator variable for sawmills with anannual capacity of greater or equal to 500,000 m3. The positive sign of this parameterin Eq. (3) indicates an upward shift of the intercept in the demand equation for thesemills. Since only relatively few observations were available for these high-capacitysawmills (n ¼ 12), it was not possible to determine the functional form of theparameters necessary for a more complex spline model. This upward shift induced bythe indicator variable can be interpreted as counteracting the negative effect of thehigher prices for sawlogs compared to pulpwood. The ratio of value to volumebecomes one of the key factors in determining the economically feasible maximumtransportation distance. In 2000, the weighted average stumpage price of sawlogswas 45.62 h/m3, while pulpwood stumpage was valued at 16.53 h/m3 on average(METLA, 2001). Transportation costs are the same for a cubic meter of sawlogs and
ARTICLE IN PRESS102 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106a cubic meter of pulpwood. Due to the higher value per cubic meter sawlogs can betransported a longer distance before transportation costs exceed the value of thewood. Therefore, compared to a pulpmill, a sawmill could economically justify asubstantially larger wood procurement area. Contrary to expectations, the roundwood price was not found to be statisticallysigniﬁcant in our data analysis. The parameter for log price did show the expectednegative sign in the initial OLS estimation, indicating that mills purchase lessroundwood in those areas with higher average roundwood prices. However, the levelof statistical signiﬁcance was very low (p-value of 0.89); therefore we dropped thisparameter from the empirical timber demand equation. We suspect that the effect ofthe highly variable transportation costs (average transportation distance 182 km,standard deviation 82 km) outweighs the relatively low level of variation between theaverage roundwood prices at the roadside for these three timber supply areas(Ostrobothnia 48.20 h/m3; Central Finland 49.92 h/m3; South-East Finland 47.32 h/m3). We expect that roundwood price would become statistically signiﬁcant if it waspossible to link these roundwood prices to ﬁbre ﬂows with a higher spatial resolutionwithin each timber supply area. Price volatility was calculated for each roundwood assortment and timber supplyarea as the standard deviation of roundwood prices (FIM/m3) from 4 to 5 smaller sub-units within each timber supply area. These more differentiated roundwood pricescould not be used directly since information on the regional roundwood ﬂows was notavailable at the level of these sub-units. The parameter estimate for price volatilityshowed the expected negative sign, indicating that less roundwood was procured fromthose timber supply areas with a high degree of variation in timber prices. There aretwo possible explanations for this response; one based on the wood procurementresponse of the mill, and the other based on the roundwood harvesting behaviour ofthe forest owner. From the perspective of the mill purchasing roundwood thisparameter can be interpreted as risk aversion. Supply security, that is, the ability toprocure a certain volume of roundwood at a stable price, is an important factor inroundwood procurement planning. The relevance of this concept of supply securitycan be seen from the arrangements that exist between the forest products industry andforest owners. For example, UPM-Kymmene Metsa offers comprehensive forest ¨management and harvesting services to forest owners as a means of securing a steadysupply of roundwood (UPM-Kymmene, 2001). Similar reasons are given for theprevalence of centrally negotiated timber sales agreements between the Swedish pulpand paper industry and forest owners (Bergman and Lofgren, 1991; Bergman and ¨Nilson, 1999). In regions with high price volatility this type of supply security isreduced substantially, thereby shifting timber demand to timber supply regions withless price ﬂuctuations. From the perspective of the forest owner, a high degree of priceﬂuctuation can discourage harvesting activities (Gomez et al., 1999). Therefore, theobservation that less roundwood is procured from timber supply regions with a highdegree of price variability can be interpreted as an effect of the lower volumes ofroundwood being harvested (or sold on the stock) by forest owners. When applying this mill-speciﬁc timber demand equation to sawmills andpulpmills respectively it should be noted that under certain circumstances sawlogs
ARTICLE IN PRESS O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 103 Fig. 2. Pulpwood share of total roundwood procurement.and pulpwood can be substitutes. This substitutability can be shown by calculatingthe procurement of pulpwood as a percentage of total roundwood purchases forindividual mills (Fig. 2). Although the majority of roundwood is being purchased by mills at the extremes ofthis ratio, there are some mills that procure a mix of both sawlogs and pulpwood.While approximately 58,000 m3 of roundwood are procured by mills with a minor shareof pulpwood purchases (5–15%), almost 2.8 million m3 of roundwood are procured bymills that meet between 5 and 15% of their volume requirements with sawlogs. This is reasonable, since pulp mills can substitute sawlogs for pulpwood andsometimes do so because of the additional costs of log sorting and transportation thatwould be incurred when separating out sawlogs during harvesting operations focusedon pulpwood. The substitution of sawlogs in the production of pulp and paper canalso be expected in situations where no sawmill is located within a reasonable distanceform the harvesting site, or when the market price of sawlogs falls below the price ofpulpwood. On the other side of the spectrum, the majority of the mills procure onlysawlogs, suggesting that pulpwood is generally not an adequate substitute for sawlogs(Brannlund, 1989; Nyrud, 2002; Størdal and Nyrud, 2003). ¨Conclusion The mill-speciﬁc timber demand equation presented in this paper makes it possibleto predict the volume of ﬁbre ﬂows between timber supply areas and a sawmill or
ARTICLE IN PRESS104 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106pulpmill for a deﬁned area in southern and central Finland. These predictions arebased on the distance between a particular mill and the timber supply area, theproduction capacity of the mill, the price volatility within each timer supply area,and an indicator variable for high-capacity sawmills. In addition to providing someinsight into the structure and dynamics of regional roundwood markets, this type ofeconometric analysis can be used as a benchmark to identify inefﬁciencies in woodprocurement procedures, and to estimate input parameters for optimum harvestallocation models such as the one presented by Troncoso and Garribo (2005). The empirical work presented in this study shows that it is possible to increase thespatial resolution of timber demand models substantially by estimating a timberdemand equation for individual mills. By expanding this mill-speciﬁc timber demandequation to include roundwood imports and time series information on the currentparameters, and by linking it with recent advances in predicting the roundwoodharvesting behaviour of NIPF owners, it will be possible to study the inter-dependencies that exist between the forest products industry and forest owners andto predict the effect that changes in the market structure will have on the ﬁnancialviability of small-scale forestry.Acknowledgments We would like to acknowledge the support of UPM-Kymmene Metsa, who ¨provided the data this paper is based on. The authors would also like to thank threereferees for their comments on an earlier version of the paper.ReferencesAmacher, G.S., 2003. Econometric analysis of nonindustrial forest landowners: Is there anything left to study? Journal of Forest Economics 9 (2), 137–164.Baardsen, S., 2000. An econometric analysis of Norwegian sawmilling 1974–1991 based on mill-level data. Forest Science 46, 537–547.Bergman, M., Lofgren, K.G., 1991. Supply risk management under imperfect competition - empirical ¨ applications to the Swedish pulp and paper industry. Empirical Economics 16, 447–466.Bergman, M.A., Nilson, M., 1999. Imports of pulpwood and price discrimination: a test of buying power in the Swedish pulpwood market. Journal of Forest Economics 5 (3), 365–387.Bolkesjo, T.F., Solberg, B., 2003. A panel data analysis of nonindustrial private roundwood supply with emphasis on the price elasticity. Forest Science 49 (4), 530–538.Brannlund, R., 1989. The social loss from imperfect competition - the case of the Swedish pulpwood ¨ market. Scandinavian Journal of Economics 91, 689–704.Brannlund, R., 1991. Disequilibrium and asymmetric price adjustment: the case of the Swedish timber ¨ market. Empirical Economics 16, 417–431.Brannlund, R., 1993. Welfare losses in disequilibrium markets - an empirical illustration. Scandinavian ¨ Journal of Economics 95, 209–225.Brannlund, R., Johansson, P.O., Lofgren, K.G., 1985. An econometric analysis of aggregate sawtimber ¨ ¨ and pulpwood supply in Sweden. Forest Science 31, 595–606. ´Cea, C., Jofre, A., 2000. Linking strategic and tactical forest planning decisions. Annals of Operations Research 95, 131–158.Faminov, M., Benson, B., 1990. Integration of spatial markets. American Journal of Agricultural Economics 72, 49–62.
ARTICLE IN PRESS O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106 105Gomez, I.A., Love, H.A., Burton, D.M., 1999. Alternative price expectations regimes in timber markets. Journal of Forest Economics 5 (2), 235–251.Greene, W.H., 2003. Econometric Analysis. Prentice Hall, Upper Saddle River, NJ.Hecker, M., 2003. Holztransport und Umweltschutz. AFZ – Der Wald 58 (4), 168–171.Hetemaki, L., Kuuluvainen, L., 1992. Incorporating data and theory in roundwood supply and demand ¨ estimation. American Journal of Agricultural Economics 74, 1010–1018.Hultkrantz, L., 1987. Access cost functions for approximating roundwood supply in northern Sweden. Scandinavian Journal of Forest Research 2, 517–524.Hultkranz, L., Aronsson, T., 1989. Factors affecting the supply and demand for timber from private nonindustrial forest lands in Sweden: an econometric study. Forest Science 35, 946–961.Kallio, A.M.I., 2001. Analysing the Finnish pulpwood market under alternative hypotheses of competition. Canadian Journal of Forest Research 31, 236–245.Karppinen, H., 1998. Objectives of non-industrial private forest owners: Differences and future trends in southern and northern Finland. Journal of Forest Economics 4 (2), 147–173.Koskela, E., Ollikainen, M., 1998. A game-theoretic model of timber prices with capital stock: An empirical application to the Finnish pulp and paper industry. Canadian Journal of Forest Research 28, 1481–1493.Kuuluvainen, J., Karppinen, H., Ovaskainen, V., 1996. Landowner objectives and nonindustrial private timber supply. Forest Science 42 (3), 300–309.Latta, G.S., Adams, D.M., 2000. An econometric analysis of output supply and input demand in the Canadian softwood lumber industry. Canadian Journal of Forest Research 30 (9), 1419–1428.Lofgren, K.G., Ranneby, B., 1987. Behavioral modes for a ﬁrm facing an uncertain supply or demand ¨ curve. Scandinavian Journal of Economics 89, 39–54.Lundmark, R., Soderholm, P., 2003. Structural changes in Swedish wastepaper demand: A variable cost ¨ function approach. Journal of Forest Economics 9 (1), 41–63.Martinello, F., 1985. Factor substitution, technical change, and returns to scale in Canadian forest industries. Canadian Journal of Forest Research 15, 1116–1124.METLA, 2001. Finnish Statistical Yearbook of Forestry 2001. Finnish Forest Research Institute, Helsinki.Nyrud, A.Q., 2002. Integration in the Norwegian pulpwood market: domestic prices versus external trade. Journal of Forest Economics 8 (3), 213–225.Nyrud, A.Q., Baardsen, S., 2003. Production efﬁciency and productivity growth in Norwegian sawmilling. Forest Science 49 (1), 89–97.Nyrud, A.Q., Bergseng, E.R., 2002. Production efﬁciency and size in Norwegian sawmilling. Scandinavian Journal of Forest Research 17, 566–575.Palander, T., 1995. Local factors and time-variable parameters in tactical planning models: a tool for adaptive timber procurement planning. Scandinavian Journal of Forest Research 10, 370–382.Pesaran, M.H., Taylor, L.W., 1999. Diagnostics for IV regressions. Oxford Bulletin of Economics and Statistics 61 (2), 255–281.Roehner, B.T., 1996. The role of transportation costs in the economics of commodity markets. American Journal of Agricultural Economics 78, 339–353.Ronnila, M., Toppinen, A., 2000. Testing for oligopsony power in the Finnish wood market. Journal of Forest Economics 6 (1), 7–22.Roos, A., Flinkman, M., Jappinen, A., Lonner, G., Warensjo, M., 2001. Production strategies in the ¨ ¨ ¨ Swedish sawmilling industry. Forest Policy and Economics 3, 189–197.Saastamoinen, O., Pukkala, T., 2001. The challenges of small-scale forestry in Finland: policy and planning perspectives. In: Niskanen, A., Vayrynen, J. (Eds.), EFI Proceedings. European Forestry ¨ Institute, Joensuu, pp. 107–117.SAS, 2001. Statistical Software. The SAS Institute Inc., Cary, NC.Størdal, S., Baardsen, S., 2002. Estimating price taking behavior with mill-level data: the Norwegian sawlog market 1974–1991. Canadian Journal of Forest Research 32, 401–411.Størdal, S., Nyrud, A.Q., 2003. Testing roundwood market efﬁciency using a multivariate cointegration estimator. Forest Policy and Economics 5, 57–68.Tilli, T., Toivonen, R., Toppinen, A., 2001. Modelling birch pulpwood imports to Finland. Scandinavian Journal of Forest Research 16, 173–179.Toppinen, A., 1998a. Econometric models on the Finnish roundwood market. Ph.D. Thesis, University of Helsinki.Toppinen, A., 1998b. Incorporating cointegration relations in a short-run model of the Finnish sawlog market. Canadian Journal of Forest Research 28, 291–298.
ARTICLE IN PRESS106 O. Schwab et al. / Journal of Forest Economics 11 (2005) 95–106Toppinen, A., Kuuluvainen, J., 1997. Structural changes in sawlog and pulpwood markets in Finland. Scandinavian Journal of Forest Research 12, 382–389.Troncoso, J.J., Garribo, R.A., 2005. Forestry production and logistics planning: an analysis using mixed- integer programming. Forest Policy and Economics 7 (4), 625–633.UPM-Kymmene, 2001. Annual Report 2000. UPM-Kymmene Group, Helsinki.UPM-Kymmene Metsa, 2000. UPM-Kymmene and Finland’s forests. UPM-Kymmene, Valkeakoski. ¨Wooldridge, J.M., 2003. Introductory Econometrics: A Modern Approach. South-Western College, Cincinnati, OH.Yin, R., Newman, D.H., 1999. A timber producer’s entry, exit, mothballing, and reactivation decision under market risk. Journal of Forest Economics 5 (2), 305–320.