Timothy Searchinger is a Senior Research Scholar at Princeton University and Technical Director for Agriculture, Forestry and Ecosystems at WRI. This article was written in collaboration with Steve Berry, David Swenson Professor of Economics at Yale University and Faculty Director at the Tobin Center for Economic Policy.
To reduce global carbon emissions, should people harvest and use more wood or less? This question underlies the merits of policies that encourage power plants and heating facilities to burn more wood pellets and builders to construct more tall wood buildings. As one illustration of the question’s importance, the U.S. government has recently requested input on whether a lucrative tax credit for carbon-neutral electricity should apply to burning wood.
In the Carbon Costs of Global Wood Harvests, published in Nature in 2023, WRI researchers using a biophysical model estimated that annual wood harvests over the next few decades will emit 3.5-4.2 billion tons of carbon dioxide (CO2) per year. That is more than 3 times the world’s current annual average aviation emissions. These wood-harvest emissions occur because the great majority of carbon stored in trees is released to the atmosphere after harvest when roots and slash decompose; as most wood is burned directly for heat or electricity or for energy at sawmills or paper mills; and when discarded paper products, furniture and other wood products decompose or burn. Another recent paper in Nature found that the word’s remaining forests have lost even more carbon, primarily due to harvesting wood, than was lost historically by converting forests to agriculture (other studies have found similar results1). Based on these analyses, a natural climate solution would involve harvesting less wood and letting more forests regrow. This would store more carbon as well as enhance forest biodiversity.
Carbon Costs focused on the pure physical emissions from wood harvest and timber management relative to leaving forests alone. This is consistent with the approach used for decades by the IPCC and numerous other papers to estimate the emissions from new wood harvests.2 However, it differs from some papers that claim the carbon emitted to the atmosphere by harvesting and using wood should generally be ignored. These papers assume that wood is carbon neutral, just like solar or wind energy, so long as other forest tracts in a large area (often a whole country) are growing enough to keep the total amount of carbon stored in forests stable — which is true of forests in most countries. By itself, this argument makes little sense: If some parts of a country’s forests are not harvested, forests in that country overall will grow more and absorb more carbon, which reduces global warming. This rationale for carbon neutrality is roughly equivalent to claiming that a money-losing company does not lose money if a country’s companies are profitable overall.
Yet, some researchers, such as the developers of the Global Timber Model (GTM), also have a more refined argument for why harvesting wood causes low, no, or even negative emissions. In a blog and a critique submitted to Nature, their core claim is that the effect of forestry on carbon is an economic question that requires analysis using an economic model rather than a biophysical one. According to the GTM, increased wood demand for any one product leads to a range of results that can lower carbon costs; these include causing people to plant more forests, to reduce their consumption of other wood products, and to intensify forest management. The first idea, that increased wood demand leads to more forests, is related to a broader idea: that forests exist because of the demand for wood. This underlies the views of many others who see wood as carbon neutral.
The GTM is by far the most cited economic model for analyzing the carbon consequences of global wood use, so its findings could have serious policy implications. Importantly, the model has been used to claim the climate advantages of harvesting more wood for bioenergy, particularly to burn in power plants. One GTM paper estimates that substantially increasing demand for wood for bioenergy could lead to roughly 1.1 billion hectares of agricultural land being converted to forests around the world. That is an area almost four times the size of India and equal to more than 70% of current global croplands — which raises the question of where the world’s food would come from.
This dialogue, to which WRI has responded in an exchange under review at Nature, provides a useful basis for exploring the effects of wood consumption on climate change and what they mean for policy. The U.S. government has specifically asked for comments about the role of economic models in treating wood as carbon neutral or negative. Here, we take a closer look at both economic and biophysical models and what each does or doesn’t tell us about the climate consequences of using wood.
Does Increased Wood Demand Lead to More Forests?
Although economic models rely on a different logic, the GTM creators and others sometimes argue that the carbon released by harvesting trees is inherently carbon neutral because it is cancelled out by the carbon that was absorbed when the trees grew.3 This theory could be valid only if all harvested forests existed solely because of the economic incentives created by wood use. If that were true, wood use would be not just carbon neutral, but carbon negative, because the very existence of these forests and the carbon they store could be attributed to the demand created by using wood.
Yet, no one seriously suggests that all or even most harvested forests exist only because of wood demand. That would include the rainforests of the Congo Basin, the Amazon Basin, South-East Asia and Alaska, each of which continues to be subject to significant harvests. It would also include the vast, heavily harvested forests of Siberia and Canada where it is too cold for agriculture. In fact, 75% of the world’s forests are owned by governments, which respond to multiple incentives. Even in the United States, a commercially oriented country, only 30% of non-corporate forest owners, who own most private forests, report timber revenue as one of the many reasons they own forests. No one seriously argues that all forests came into being because of wood demand or would disappear without it.
Even so, it is possible that increased wood demand — by boosting wood prices and therefore the profitability of forestry — could lead to planting or preserving some more forests (in addition to harvesting existing forests more). That is the claim in some GTM model outputs. But for wood demand to preserve carbon overall, it must cause forests to increase globally, not just in some areas. It does not help if forests replace agriculture in one place only to have agriculture expand into forests elsewhere to replace the lost food production. Since the GTM is the primary model making the claim that demand leads to more forests globally, what is its evidence?
This GTM result relies on a single parameter built into the model, which specifies that doubling the profitability of each type of forest leads to a 30% increase in that forest type, and this applies to all forests in the world.4 This parameter underlies the finding in one GTM paper that using enough wood for bioenergy will increase global forest area by almost 30%, equal to almost 1.5 times the size of the Continental U.S. Although applied in every world region, the GTM authors attribute this parameter to a single study of small land-use changes in the United States by economist Ruben Lubowski, published in various forms from 2002-2006.
However, what the Lubowski study found was that a doubling of forest profitability in the United States had only an “extremely small” effect on forest area. The study’s true estimate was that a doubling of profitability would lead to only about a 0.4% increase in forest area — 1/75th of the parameter used in GTM. (We confirmed the elasticity used by GTM in emails with its lead modeler, confirmed the actual elasticity in conversations with Lubowski, and explain the source of the misinterpretation in the box below.)
The Lubowski study also emphasized that this “extremely small” effect was local only; it did not necessarily predict an increase in forest area across the whole United States, let alone globally. This is because converting agricultural land to forest tends to outsource food production elsewhere. Opening a paper mill in one location might encourage some adjacent agricultural land to be used as a forest plantation, but that would lead to at least some agricultural conversion of forest elsewhere to replace the lost food production. Counting such rebound effects, the “extremely small” local effect could easily mean no global effect or even a net loss of forests — particularly if some food production shifts from the United States or Europe to developing countries where the same food production typically requires more land (due to lower yields) and loses more carbon.
For some GTM papers, the modelers have added a term that is designed to reflect this rebound effect and reduces the extent of global forest expansion. Having any curb on the misinterpreted expansion effect is a step in the right direction. But this pushback is based on a single parameter at the global level, intended to capture the full effects of every type and use of agriculture, which has no credible empirical justification, as far as we can tell.5 Regardless, forest area expansion in the model is still heavily influenced by the overestimate of Lubowski.
Overall, we are aware of no credible evidence that wood demand has led to more forest area globally. The basic reason is that the economic returns from forestry are nearly always much lower than those from agriculture. Whether agriculture occurs in an area depends mainly on its ability to compete with agriculture elsewhere. The compelling evidence is that forests exist where they naturally grow and where such factors as temperature, steep slopes, rainfall, or limited transportation make agriculture unable to compete with farming in more favorable locations. This also means that there is no credible evidence that increasing wood demand for bioenergy or for building construction will lead to more forest area.
How the GTM Misinterpreted the Effect of Increased Forest Profitability on Forest Area
A major assumption built into the Global Timber Model is that doubling the profitability of forestry (profit per acre) would lead to an expansion of forest area locally by 30% (technically a forest area elasticity of 0.30). In an appendix to a 2018 GTM paper, GTM modelers wrote:
“Lubowski, Plantinga, and Stavins (2006) suggests that the price elasticity of land conversions from crops to forestry (and vice-versa) is 0.30 in the United States [30%]. We have found no similar studies for other regions. Unlike with forest yield functions, where we have data on aggregate biomass volumes for separate age classes, or where ecologists have published data from study sites in a wide range of biomes, economists have not produced land supply elasticities outside the United States. We thus apply the estimate from Lubowski, Plantinga, and Stavins (2006) to other regions.”
But the GTM authors misunderstood what this 30% elasticity represented. It was not an estimate of the percentage change in the total area of forest. Rather, Lubowski estimated that, regardless of forest profitability, a very small percentage of cropland in the United States transitioned to forest every five years (and some forest in turn became cropland). The 30% elasticity meant that a doubling of the profitability of forestry increased this very small percentage by 30%. To estimate the effect of increased forestry profitability on the total area of forest, the two percentages must be multiplied together, which results in an even smaller percentage. (For example, a 30% change in a 1% rate of change would be only 0.3%.) This was best explained in the most comprehensive version of the Lubowski study, published in 2002. As Lubowski wrote there: With a “near doubling of forest profits . . . the overall increase in forest areas . . . attributable to the increase in forest profits [is] extremely small as a percentage of the total forest area in the nation (only . . . 0.4%).” (This small effect was further shown in Figure 2 of a subsequent paper).
In short, the actual local elasticity of forest area estimated by Lubowski was only around 1/75th of the elasticity used by GTM (0.4%/30%).
Even this was only a local effect: Because of the need to replace any food displaced by forest expansion, Lubowski’s result did not necessarily imply any global forest expansion. In an apparent effort to account for this rebound effect in some way, some versions of GTM introduce a constraint on the level of forest expansion (as discussed in endnote 5), but the misinterpretation of the Lubowski elasticity is still a force in the model that drives forest expansion.
What Explains Why Forests Are Storing More Carbon Globally?
If not the demand for wood, what explains why the carbon stock of global forests is growing — despite carbon losses due to extensive logging and ongoing deforestation for agriculture?
The primary answer is climate change itself. Higher carbon dioxide levels in the atmosphere cause forests to grow faster. This effect is aided by warmer temperatures in cold regions, which allow forests to grow for more of the year. While there remains uncertainty regarding the precise quantity, estimates today are that climate change causes forests to absorb roughly 10 billion additional metric tons of carbon dioxide (CO2) per year. That is more than one-quarter of global CO2 emissions. In addition, many northern hemisphere countries have reduced their reliance on forests for fuelwood and grazing. They have also reduced their agricultural land areas due to a variety of factors, including a shift in agriculture to the tropics, where deforestation is rising in part due to the global North’s outsourcing of food production and to the replacement of horses with cars and tractors, freeing up large areas once devoted to feeding horses (discussed here). These reductions in agriculture have allowed many forests in the temperate zone to regrow. (See this graphic for Europe as one example.)
In short, forests are growing despite wood harvests, not because of them.
Significantly for policy, this forest carbon uptake is already factored into global warming estimates and is effectively assigned as a kind of implicit carbon credit to all people. When people add CO2 to the atmosphere, scientists estimate that slightly less than half stays there (called the “airborne fraction”), so the warming effect is less than half a ton of CO2. Something like one quarter of a ton is taken up by the ocean, but probably a little more than one-quarter is taken up by additional forest growth spurred by higher CO2 and climate change itself. In other words, without this additional forest sink, the fraction of CO2 remaining in the atmosphere would be not half but three-quarters. This means that without the advantageous feedback effect on forest growth, the increase of CO2 in the atmosphere caused by people would be roughly 50% higher.
In fact, this helpful forest feedback effect of CO2 is built into estimates of CO2’s “global warming potential.” Put colloquially, this means that emitters are responsible for roughly 25% less warming than they would be responsible for without the feedback. This works as a kind of implicit carbon credit to each person who emits a ton of CO2. Without double counting, this added forest growth cannot be credited as an implicit offset to wood harvests — meaning it cannot be used treat wood as carbon neutral, except by taking it away from everyone else.
Precisely for this reason, the Paris Agreement allows countries only to take credit for “anthropogenic removals,” or those caused by their own land-management changes. And when the countries that were part of the Kyoto Protocol negotiated the rules for forests — the only time climate agreements established detailed rules of this kind — the rules explicitly provided that countries could not take credit for the carbon sequestration in their forests due either to CO2 fertilization or to the regrowth of forests harvested or reestablished previously.6
Could More Intensive Forest Management Reduce Overall Emissions?
Unlike overall forest area, which the evidence shows responds to other factors, more intensive timber management is a direct result of wood demand and can affect the carbon costs of wood production and use. The largest management change is to convert natural forests into fast-growing wood plantations, often using single species. Carbon Costs factors this plantation management into its emissions estimates, using two possible future scenarios of extreme intensification: In one, all natural forests are converted to plantations after harvest; in another, plantation yields grow by 50%. In these scenarios of extreme intensification of forest management, the annualized emissions decline from 4.2 to 3.5 gigatons of CO2 per year.
Crucially, the only reason plantations can save carbon is because wood is not carbon neutral. Because plantation forests are harvested young and repeatedly — in the tropics, often every ten years or less — they typically store less carbon than natural forests. However, plantations also produce more wood per hectare per year. This means they can save carbon overall by reducing the need to harvest wood from natural forests to meet the same demand for wood, allowing natural forests to store more carbon. If harvesting natural forests were carbon neutral, there would be no carbon losses to reduce.
Does Using Economic Models to Estimate “Avoided Emissions” from Wood Harvest Alter their Physical Emissions?
An even more fundamental question is whether emissions from wood harvest and use should be assessed using biophysical models or economic ones. The functions of the different types of models are commonly misunderstood. The basic explanation is that biophysical models calculate the physical emissions caused by human activities, whether using paper, heating a home or eating a hamburger. Economic models like the GTM, if credible, can help analyze the impact of different policies by factoring in “avoided emissions” through shifting from one type of activity to another. This is quite a different focus.
Driving large and small cars illustrates the distinction. To estimate the climate impacts of driving cars — whether large or small — a biophysical model would be used to estimate the average emissions of each compared to not driving at all (which emits no carbon). For example, by one set of definitions, driving a large car in the United States emits an average of 6.2 metric tons of carbon dioxide per year, and a small car 3.1 metric tons per year.
By contrast, an economic model used to evaluate the impact of a policy to reduce the supply of large cars might reasonably estimate people would shift to small cars. Since small cars generate half the emissions per vehicle, the economic model would estimate that such a policy would reduce driving emissions by half. The same model might estimate that a policy that reduces the supply of small cars would shift people to large cars and double driving emissions. This could be valuable policy information. But it does not mean that driving a large car only releases half its physical emissions (on the grounds that the alternative would be a small car); that driving small cars is carbon negative (on the grounds that the alternative would be a large car); or that driving half large and half small cars is carbon neutral.
Forestry is no different. To understand the true climate impact of wood harvesting, its emissions must be compared to no human activity at all — leaving the forest land alone with no harvests and no timber management — rather than to an alternative human activity. This is how Carbon Costs, IPCC reports, and many other studies evaluate forestry (see endnote 1): the alternative is that the forest remains standing. This is also how the IPCC and others evaluate the impacts of clearing forests for agriculture. And the no-human-activity alternative is how emissions are typically calculated for all other human activities.
By contrast, an economic model like GTM seeks to estimate the effects of wood demand compared to the most likely alternative human activities. It therefore deducts from forestry emissions the emissions avoided by not doing other human activities, similar to the avoided emissions in switching from large to small cars. For example, if the model estimates that some forests would most likely be converted to cropland absent wood demand, then wood use is credited with avoiding the emissions from cropland conversion. That could be useful information if true, but it would not mean that forestry has no emissions; it would just mean converting land to cropland is worse.
The most important reason to count physical emissions rather than “avoided emissions” is that calculations will otherwise underestimate emissions. Nearly every human activity has an alternative activity that also causes emissions, such as driving large versus small cars. If each human activity were credited with avoiding emissions from another, adding them together would total far fewer emissions than the physical reality. In fact, just as the avoided emissions from driving different sizes of cars can cancel out all driving emissions, counting up avoided emissions of multiple activities would often sum to none at all.
There are also many possible policies, and the level of actual emissions helps inform what different policies might achieve. For example, instead of just shifting from large cars to small cars, policies could seek to reduce all driving. The physical emissions from driving tell us how much carbon that would save. Similarly, even if wood demand did help limit the expansion of cropland, biophysical models show how much carbon could be saved using a combination of recycling policies and agricultural policies to both reduce wood consumption and curb cropland expansion.
Careful economic modeling can play a valuable role in comparing the effects of different policies. Using an economic model to estimate “avoided emissions” is also a method commonly applied to evaluate carbon offsets, which have a special but limited role to play in climate policy (although the difficulty of the estimates contributes to controversies about offsets). But the climate outputs of such models only compare emissions from one set of human activities with another; they do not estimate actual emissions. Confusing “avoided emissions” with actual emissions would cause vast human emissions to seemingly disappear.
Would Policies to Increase Wood-based Bioenergy Reduce Wood Use Elsewhere?
Another claim sometimes used to justify bioenergy policies is that using more wood for energy, by increasing wood prices, will substantially reduce the amount of wood people consume for paper and construction. These reductions would mean that wood could replace fossil fuels without requiring much additional wood harvest, and therefore with less effect on forests and their carbon.
In the GTM, this effect is large. It assumes, for example, that if enough new bioenergy demand doubled the price of wood, consumption of all paper and timber wood products would decline by 50% — from toilet paper to cheap desks to construction timber. The GTM also assumes that these reductions would occur everywhere in the world, from the United States to Russia to Tanzania. In one paper, GTM estimates that enough wood bioenergy would drive down consumption of all these other wood products by a remarkable 80%.
The GTM authors credit this parameter (technically a demand elasticity of -1) to a book that describes what was, in essence, the first version of GTM. But that book actually claims an appropriate demand response (technically an “own-price demand elasticity”) of only around one-quarter the size of the assumption later built into the GTM.7 Furthermore, the book did not cite studies for this estimate, but stated it was chosen to be “reasonable.” Economic studies that are available (generally in the United States) tend to find elasticities of around one-tenth the size of that chosen by GTM.8 The available economic evidence, therefore, is that harvesting additional wood to burn in powerplants has only a very small effect on consumption of wood by others, so nearly the full additional quantity of wood must be harvested.
Diverting wood from other uses also has its own climate costs. While using wood for any purpose causes emissions, it is broadly acknowledged that using timber for construction is better for the climate than burning wood in a power plant. If a model assumes that increasing one kind of wood use will reduce another wood use, then it should factor in the emissions from replacing that wood (such as the emissions from alternative building materials). The GTM does account for some storage of carbon in wood products. But it generally does not account for the need to use concrete, steel, plastics or other materials to replace wood products.
Why Global Economic Forestry Models Are Likely to Generate Biased Results with Low Carbon Cost Estimates
The GTM aims to use economics to project where in the world and through what form of forestry new wood demand will be met, and how that will alter global land use and carbon storage. Such a model, if credible, could be useful for policy. However, given data limitations today, we do not believe this type of model can be credible at this time. Perhaps even more importantly, the assumptions built into GTM and other models of its kind lead to a structural bias that systematically understates the climate costs of wood harvests.
The limitations become clear from the fact that the GTM applies parameters derived from studies only in the United States to the whole world. Wood use differs dramatically by country and by product type. It is therefore not credible to assume that a demand elasticity derived for construction in the U.S. would be the same for toilet paper in Germany or housing construction in Kenya. Similarly, the Lubowski study found that even within the U.S., the change in forest area due to changes in profitability of forestry differed greatly from place to place (although it was always small). These differences make sense because land uses reflect differences in rainfall, slope and soils and access to markets. It is therefore equally unlikely that an average U.S. response will meaningfully reflect the response in every country in the world.
Modern economics requires that these kinds of analyses rely on rich data variation and use econometrically credible statistical methods. Some rough global economic models based on global data might be achievable, but at this time, global models that project precise responses in different regions or countries cannot be credible. Even for policy, the better approach is to gain insights from rigorous local studies and to estimate the future using a biophysical model to analyze a range of different scenarios.
Even more importantly, the GTM and similar models rely on a core assumption that automatically drives low climate cost estimates. This is the assumption that all forests, at least if or once they become “accessible,” are managed solely to maximize timber profits. While that is undoubtedly true of some forests, it is not true of most. Because the model assumes that trees in most forests only exist to meet wood demand, the only reason trees remain standing at any time would be to meet already expected future demand. With these existing trees and future growth already, in effect, “taken,” meeting additional demand would requires growing more wood through planting more forest area or more intensive management. In effect, the model largely assumes the new demand will not be met just by harvesting more trees that would otherwise remain standing, which is the main action that reduces carbon stored in forests.9
These structural assumptions dictate the model’s finding that increasing wood demand causes limited reductions in forest carbon. In short, the assumed model structure assumes away most of the climate costs.
What Does This Mean for Wood Policy?
The world has valuable uses for wood, just as it has valuable uses for food and many other products that also cause emissions. And, just as for other goods, the goals should be both to reduce the greenhouse gas emissions involved in wood production and to hold down overall consumption. That’s the basic reason the world recycles paper and that policies have started to encourage wood reuse. It also explains why there is some role for intensive forest management.
Under business as usual — even without policies to increase the demand for wood — the world faces a doubling in demand for commercial wood harvests by 2050 relative to 2010. The exact mix of policies to meet this demand is a complex question. But it should start with the recognition that, just like other products, using wood is not carbon-free.
Notes:
1 Other global studies which find existing forests are missing very large quantities of their carbon include Erb et al. (2018) and Walker et al. (2022). Europe-focused papers include Keith et al. (2024).
2 IPCC Assessment reports estimate emissions from land use change that have incorporated emissions from wood harvest using biophysical “bookkeeping models,” similar to the model in Carbon Costs. Bookkeeping models start with estimates of wood harvest and use. The original model primarily relied upon by the IPCC was developed by Dr. Richard Houghton and published in many papers over the years (for example, see here and here). New models have since been added, including the “BLUE model”; these two models have also been reflected in the annual carbon budget estimates provided by the Global Carbon Project, which is the most commonly cited source for annual emissions. In each of these models, emissions are based on the alternative of leaving forests alone, not on an economically estimated counterfactual. As discussed in Carbon Costs, sometimes the emissions of new wood harvests were obscured by the reporting of net emissions that combined the effect of new emissions and the recovery of previously harvested forests in one number. Although this netting could obscure the impacts of new harvests, the estimated emissions of new harvests were still based on biophysical models, for which the alternative is no harvest. This problem of netting was itself pointed out by Houghton in several papers in which he separated the effects of new harvests of some forest areas from regrowth of other forest areas previously harvested (see here and here). Numerous other papers also use biophysical models to evaluate the carbon effects of forestry. Examples include: Naudts et al. 2016; Hudiburg et al. 2019; Kalliokoski et al. 2020; Skytt, Englund, and Jonsson 2021; and Chen et al. 2018.
The major differences in the Carbon Costs model were: (1) substantially added detail for global forestry estimates relative to other models (in part because they also focused on agricultural expansion); and (2) incorporating into the costs of wood harvests the carbon sequestered by faster forest growth post-harvest because young forests tend to grow faster than older forests. This addition reduced, rather than increased, the attributed carbon costs. Carbon Costs calculated the net effect of new harvests 40 and 100 years after harvest, and it also calculated the costs using different discount rates to value emissions in a way that recognizes earlier emissions are even more costly than later emissions.
3 The phrasing of the claim is that the effects of harvesting and using wood should be counted from the time trees start to grow. As a result, wood use first receives a credit for the carbon sequestered by forests, and only spends this credit when the trees are harvested, making wood inherently carbon neutral at least.
4 Although most GTM modeling papers provide limited description, this description is provided in Appendix B of Tian et al. (2018) and quoted further in the Box.
5 This factor is described in equation 3 in Sohngen et al. (2019), although there is no reference to it in many other GTM papers. It essentially provides that as the misinterpreted Lubowski local elasticity drives forests expansion around the world, there is a pushback effect to slow this rate down. This pushback effect is based on the global change in forest area, rather than the local change. While this provision does moderate the extent of forest expansion, there is no citation, and to our knowledge no empirical support, for this function or the single parameter that controls it. A credible analysis of this rebound effect would have to account for different supply and demand elasticities for different major agricultural products, and since the model claims to be disaggregated by region, to different regions. Because the model does not include agricultural products, it does not have any of these components.
The added factor curbs expansion in some, but apparently not all, GTM papers, but does not alter the fact that forest expansion is still driven by the misinterpreted Lubowski local expansion effect of only around 0.4% rather than 30%. The interaction of this curbing factor with the local expansion effect can also cause unwarranted shifts in forest area in different regions even as Lubowski found even local effects to be “extremely small.”
6 At the first “conference of the parties” after agreement on the Kyoto Protocol, the signers adopted a specific principle for all “land use, land use change and forestry” as follows: “(h) That accounting excludes removals resulting from: (i) elevated carbon dioxide concentrations above their pre-industrial level; (ii) indirect nitrogen deposition; and (iii) the dynamic effects of age structure resulting from activities and practices before the reference year.” (Decision -/CMP.1). The last clause excludes the carbon gain from regrowing forests established prior to 1990, the base year for country reporting.
7 In Sohngen et al. (2001), GTM authors write, “Regional demand functions are calculated assuming a uniform price elasticity of 1.0 (Sedjo and Lyon 1990).” (Technically the elasticity was -1.) The reference is to a 1990 book (not available online) published by the organization Resources for the Future, entitled, The Long-Term Adequacy of World Timber Supply, which first described the model, the Timber Supply Model, that evolved into GTM. This book discusses demand elasticities only in the chapter entitled “Introducing Demand in the Timber Supply Model,” where it states: “The elasticity of world demand in the base scenario varies from 0.17 in the initial year to 0.18 in the stationary state, and for the high-demand growth scenario it varies from 0.19 to 0.30 for the two periods.” The book did not cite any studies to justify this assumption but states that they were chosen to be “reasonable” (adding the quotation markets itself).
8 In Appendix B to the 2018 GTM publication, the authors claim this -1 demand elasticity is only somewhat higher than wood “import elasticit[ies]” estimated in some papers. But wood import elasticities are not own price demand elasticities, which are those used by the model, i.e., how much wood consumption declines with a 1% rise in prices. Wood import elasticities estimate how much increases in international wood prices lead to reductions in U.S. wood imports only. Even high elasticities do not necessarily imply any reductions in the consumption of wood because U.S. imports can be entirely replaced by domestic supplies of wood. International suppliers are essentially competitors with domestic suppliers, and just like any other suppliers, will lose market share to competitors if they raise their prices. The same limitation applies to studies such as Newman (1987), also cited, which estimate the demand for wood on a regional basis, as they will measure not necessarily declines in consumption but just switches to supply to other regions or countries.
In the same publication, the GTM authors cite but do not discuss Hayes et al. (1981), yet that paper found elasticities generally around one-fifth of those in GTM. Its senior author also updated it in a subsequent paper to less than one-tenth those chosen by GTM. The GTM authors also cite Simangunson & Buongiorno (2001), which they claim had high elasticities. But that paper used multiple specifications of its model, all of which found estimates that were extremely low except one specification, which the paper criticized as having “excessive bias.” One of the two authors of that paper subsequently updated it, generating uniformly low elasticities by all methods.
9 GTM does assume that there are some “inaccessible” forests, and if wood supply from accessible forests becomes too expensive, then roads may be built, and inaccessible forests harvested. But accessible forests include most global forests, including nearly all in the U.S. and EU. (as described in this paper.) This means that the model will generally require that additional wood comes from planting more forest lands or more intensively managing existing forests.