Directory UMM :Journals:Journal Of Policy Modeling:Vol22.Issue4.2000:

Forest Conservation: A General Equilibrium

Analysis of Malaysia

Frank Harrigan,Asian Development Bank

This article examines the loss in “metered” aggregate income that could occur if Malaysia surrendered the lumber value of its tropical forest resources to nonlumber uses. We estimate these losses under a variety of assumptions about what “business as usual” and “conservation” might entail. We also consider the sensitivity of income losses to variations in our model’s parameter values and to some of its maintained hypothesis. In a context where lumber’s initial contribution to aggregate income is around 2 percent, we estimate that a switch from lumber to nonlumber uses of tropical forests could cost up to 4 percent of baseline income. Of broader significance is the implication that the associated dynamic general equilibrium multipliers are consistently greater than unity, and often close to two in value. These large income multipliers are observed despite an assumed recovery of income through the reallocation of mobile factors initially employed in lumber activities, and an increase in returns to capital. In this study, a terms of trade deterioration, prompted by the loss of lumber foreign exchange revenue, accounts for about one-half of the total income losses we observe. For economies that rely to a greater extent than Malaysia on lumber foreign exchange revenue, these terms of trade-induced losses could be more important still. From a policy perspective, our results provide a benchmark against which the contingent valuations and other imputations of the monetary value of nonlumber use values of tropical forests may be gauged. 2000 Society for Policy Modeling. Published by Elsevier Science Inc.

Address correspondence to F. Harrigan, Strategy and Policy Office, Asian Development Bank, P.O. Box 789, Manila, Philippines.

With the usual disclaimer, the author would like to thank David Demery, J. Malcolm Dowling, Jr., Chris Edwards, Peter McGregor, Kim Swales, and Jeffrey Vincent for helpful comments. The paper also benefited from the comments of participants of the Economics Department seminar program at the University of Melbourne, and from participants comments at the Harvard Institute for International Development inaugural workshop on Malaysian economic development.

Received February 1997; final draft accepted September 1997. Journal of Policy Modeling22(4):491–531 (2000)

2000 Society for Policy Modeling 0161-8938/00/$–see front matter Published by Elsevier Science Inc. PII S0161-8938(97)00158-0

1. INTRODUCTION

Policies that seek to correct environmental externalities raise distributional tensions not only within countries but also some-times between them. Where a nation owns resources that others regard as part of the global commons, it could be compensated for exporting that part of any conservation benefit stream not captured locally. This principle has appeal, and may be a political precondition for conservation, where the resource economy is “poor” and other countries are “rich.”

While disagreement about environmental cost–benefit calculus is likely to be commonplace, partly because of uncertainty about the associated science (Cline, 1991), the economic principles that might govern compensation are clearer. The costs borne by a nation that conserves a global resource will be bounded from above by the sacrifice of consumption that conservation entails. The actual costs will almost certainly be less than this ceiling, because some residents of the resource-owning economy will ben-efit from conservation. The minimum other nations should be prepared to pay for conservation is what it would cost them to secure equivalent net benefits by alternative means.

In this paper, we attempt to measure the loss of “metered”1

income that could occur when an economy surrenders the lumber value of its forests to nonlumber uses. We estimate these costs for Malaysia, one of the world’s largest producers of tropical lumber products. However, our analysis, which is largely coun-terfactual in nature, has more general applicability. We find that the indirect income loss entailed by a switch of forest resources to nonlumber uses dominates the direct loss, and that the associated general equilibrium income multiplier is greater than one. In this study, a terms of trade deterioration triggered by a loss of lumber foreign exchange revenue has a large adverse effect on household real income.

In principle, there is no difficulty in weighing these income losses against the environmental benefits that conservation would secure. But, in practice, there are formidable measurement prob-lems. Many of the markets in nonlumber uses of forests (e.g.,

1_{Here we use the term “metered” income in the sense of Nordhaus (1991) to refer to}

incomes that have a counterpart market transaction, and which would therefore normally enter into the measurement of national income. The adjective is intended to signify that there are other incomes that go unmeasured.

biodiversity) are largely missing, and obtaining reliable estimates of existence and option values is problematic. While contingent valuation and other methods of imputation promise some progress in the assignment of monetary values to conservation benefits, these exercises are costly, and not without their limitations.2_For

these reasons, we do not attempt to measure those offsetting conservation benefits that would accrue to Malaysian residents, nor do we measure those benefits whose incidence might be more widely felt.

In the next section of the paper, we describe the contribution of lumber resources to Malaysian economic development. In Sec-tion 3, we motivate our approach to the analysis of forest conserva-tion, and describe the main features of the general equilibrium framework that we use in our calculations. We explain the design of our stimulations in Section 4. Section 5 discusses the main results of our paper, and tests their sensitivity to key assumptions. The concluding section briefly considers some of the policy issues raised by our analysis.

2. MALAYSIA’S LUMBER RESOURCES AND ECONOMIC DEVELOPMENT

The FAO (1993) estimate that 60.2 percent of the land area of Malaysia was forest in 1990. Of this, 53.5 percent was natural forest and the remainder plantations. Between 1981 and 1990, the FAO estimate that the natural forest area of Malaysia was being depleted at an annual average rate of 2 percent, which is high by both regional and world standards. If deforestation were to continue at this rate, the half-life of Malaysia’s remaining tropical forests would be only 35 years.

Alarming as these figures may seem, estimates of the rate of deforestation may not fully account for changes in the stock or quality of forest resources. Often, the degradation of forest re-sources may be more severe than the loss of forest area suggests (Burns, 1986). Between 1980 and 1990, 2.6 percent of Malaysia’s forest area was logged annually, and the vast majority of this log-ging was in previously unlogged areas (FAO, 1993). The annual lumber harvest in Malaysia doubled between 1971 and 1989, and lumber stocks could easily have halved over the same period.

Logging, however, is only one of many sources of tropical defor-estation. In Malaysia, land settlement, shifting cultivation, fuel-wood consumption, natural fires, the conversion of native forest for commercial agricultural, and (in earlier years) tin mining have all contributed to deforestation. Unfortunately, it is impossible to accurately identify the share of deforestation attributable to these and to other causes. Nevertheless, it has been estimated that the reduction in the stock of Malaysian lumber due to nonlogging sources may be between two to three times that caused by logging itself (Gillis, 1988). But while land settlement and the conversion of forest area for commercial agriculture were responsible for much of Malaysia’s deforestation in earlier decades, their rate of encroachment on forest areas has since fallen. Today, lumber extraction makes a larger contribution to deforestation than in the past.

The environmental impact of deforestation depends on how it occurs. Some sources of deforestation have more benign environ-mental effects than others.3_{There can, however, be little doubt}

that, in general, deforestation has degraded animal (and human) habitats, interfered with nutrient cycles, resulted in a loss of biolog-ical diversity, threatened water resources, caused the loss of non-lumber forest outputs (such as rattan, foodstuffs and potential medicines), eroded educational, recreational, and tourist amenity, and contributed to the accumulation of carbon in the upper atmo-sphere.

To an extent, deforestation in Malaysia has been the result of a cocktail of market and institutional failures that have led to stumpage values4 _{that do not adequately reflect the nonlumber}

use and other nonuse (including option and existence) values of tropical forest resources. For example, ill-defined property rights and insecure tenure have created incentives for timber to be “mined” in a context of slow growing stumpage values (see, e.g., Vincent, 1992). However, deforestation has also been a conse-quence of policies that have consciously traded off the value of

3_{For example, the conversion of forest area for commercial agriculture, if it involves}

land clearance through burning, is unequivocally bad for CO2concentrations in the upper

atmosphere. If, however, lumber is felled and is used in durable wood products, carbon sequestration can continue for some time. Natural tropical forest fires are a particularly damaging form of deforestation. They are often aggravated by earlier deforestation that has eroded forest moisture. In 1994, tropical forest fires covered much of the of Borneo, West Malaysia, and Singapore in a haze for several months of 1994.

forest land area against other uses. The transfer of forest land to permanent tree crop agriculture (palm oil and rubber) has undoubtedly generated substantial income for Malaysia. Equally, concession royalties and other taxes extracted from lumber and tree crop sectors have long provided an important source of fi-nance for the Malaysian government. Nor should it be overlooked that land settlement has made a significant contribution to en-abling Malaysia’s rural poor (IBRD, 1993). But the benefits of other policies are in doubt. For example, Vincent (1992) concludes that attempts to foster the development of local wood-based indus-tries through log export bans and other measures have been costly. For every sawmill job created in Peninsular (West) Malaysia be-tween 1973 to 1989, the economy sacrificed both value-added and export revenue (Vincent, 1992). Also, by reducing the local price of lumber, and subsidizing inefficient producers, log bans may have accelerated deforestation (Braga, 1992).

Our objective in this paper is not, however, to determine whether Malaysia has managed its tropical forest resources pru-dently, or to suggest how they ought to be managed in the future. Dasgupta and Ma¨ler (1991) explain the complex nature of “sus-tainable development” and identify its relationship to renewable resource exploitation. And when preferences vary widely within or between societies, the identification of a socially “optima” pro-gram for resource use may not even be possible. In isolating the income repercussions of tropical forest conservation, we provide information that policy makers may weigh against the nonlumber value of tropical forests. But, in the absence of complementary information about conservation benefits, no welfare inferences can be drawn from our analysis.

3. THE MODELING APPROACH

Conceptually, we portray tropical forest conservation in terms of the permanent withdrawal of an immobile resource from the traded goods sector of a small open economy. Malaysia is, of course, small in the sense that its activity has little impact on the rest of the world, but it is not necessarily small in the sense that it is a price taker in world goods markets (see below). We know from Corden and Neary (1982) that, in general, the withdrawal of an immobile resource from the traded goods sector of an econ-omy will lead to a reduction in (“metered”) aggregate income,

an accompanying depreciation of the real exchange rate,5_{and a}

movement of resources from the non-traded to the traded goods sectors. Because the magnitude of the resource shock that we consider is large, these interlocking general equilibrium repercus-sions cannot safely be ignored. It is for this reason that we use an empirical general equilibrium model of the Malaysian economy to illustrate the effects of conservation. As we shall see, a partial equilibrium analysis would give seriously misleading results, and an empirical macro model would not cover well the mechanisms through which income changes are propagated.

Our applied general equilibrium model (Demery et al. (1992)) is not unlike the family of models described by Robinson (1991), and more fully articulated by Bourguignon et al. (1992). It identi-fies traded and non-traded goods sectors within a broader classifi-cation of 13 commodities. “Lumber” and “Tree Crop” activities (and commodities) are separately identified. The commodity out-put of both these activities is traded. A third agricultural activity (“Other Agricultural”) produces commodities (such as fish and fruit) that are primarily destined for domestic consumption. There are three manufacturing sectors, one of which is “Resource Based.” This “Resource-Based” manufacturing sector includes some downstream lumber activities, including the manufacture of wooden furniture, and paper and pulp products. The other manufacturing sectors are classified as “Export Oriented” and “Domestic.” Non-traded goods and activities include “Construc-tion,” “Dwellings,” and “Public Services.” The outputs of “Private Services” and “Utilities” are largely but not completely non-traded. Finally, “Oil and Gas” and “Other Mining” activities com-plete our sectoral division of economic activities.

There are both mobile and immobile factors of production. Capital and land, which are aggregated in a Hicks composite factor, are quasi-fixed, with their sectoral allocation responding to relative rewards through the allocation of net investment. The aggregation of capital and land is dictated by a lack of data. Such aggregation is valid if the relative rewards to capital and land do not change, or if technology is separable in capital and land. We consider the significance of the aggregation of capital and land, and its quasi-fixed nature, further in Section 5 below. Labor is

5_{Throughout, the “real exchange rate” refers to the price of nontraded to traded goods,}

intersectorally mobile. We distinguish between three different skill types (“Unskilled,” “Semiskilled,” and “Professional and Skilled”) and between workers in “Land-Based” and “Nonland-Based” ac-tivities.

Producers in each sector, except lumber, solve static profit max-imization problems subject to multilevel (constant returns CES) technology constraints. These determine input demands and out-put supplies. In the lumber activity the representative producer faces an output constraint and, given output, minimizes cost. Pa-rameteric variation of this output constraint facilitates the mea-surement of changes in income associated with lumber conserva-tion. The market for Malaysian lumber is cleared through rationing exports. Unsatisfied domestic demand is met through imports. This treatment seems reasonable given the priority that Malaysia has given to securing lumber for local supply (see above). In other sectors, equilibrium is achieved through relative price adjustments. Where lumber stocks are managed as a renewable resource, producers maximize the expected net present value of their lumber stock by harvesting lumber until the real interest rate equals the lumber growth rate plus the rate of expected appreciation of its stumpage value. This assumes that stumpage values are non-negative, and that second-order conditions are satisfied. However, a conjunction of insecure property rights, old lumber, and slowly growing stumpage values has encouraged concession holders to “mine” lumber in Malaysia (Gillis, 1988; Vincent, 1992). In circum-stances like these, where planning horizons are short, intertempo-ral first-order conditions are irrelevant, and producer behavior can reasonably be represented in static terms.

In our model, aggregate consumption is a function of household disposable income and private sector nonhuman wealth.6_A

repre-sentative household then allocates its aggregate expenditure be-tween goods using a linear expenditure system. We model import and export demands using Armington (1969) functions. Activity investment demand is exogenous in lumber and tree crop sectors, and is an increasing function of Tobin’s “q”7_elsewhere.

Transactors’ stocks of net wealth (measured at the beginning of the period) are supplemented by their (end of period) financial

6_{We assume that all transactors form their expectations adaptively. If we were instead to}

assume that consumers optimized intertemporally and that they formed their expectations rationally, then income losses would, by definition, be reduced over those we estimate.

surpluses. Net wealth is then allocated over assets using portfolio selection models in which relative returns and perceived risks are important. End-of-period asset market equilibrium occurs through the endogenous adjustment of asset yields, and the nominal ex-change rate.

Lumber workers, who are predominantly “unskilled,” sell their labor services in the “Land-Based” labor market that covers work-ers in agricultural, fishing, and lumber sectors. In the remaining sectors of the economy, there are separate labor markets for each of the three skill categories. In each period, wages adjust to clear the Land-Based labor market and the markets for Semiskilled and Professional workers. In the Unskilled labor market, we allow for the possibility that some workers who loose their jobs may experience a spell of unemployment. However, our wage adjust-ment function is parameterized so that unemployadjust-ment reverts to its equilibrium rate quickly. This reflects the flexibility that Malaysia’s labor markets have exhibited for many years.8_Workers

migrate from the land-based to other labor markets in response to (lagged) real consumption wage and unemployment differentials. Other than through migration, labor is inelastically supplied, and labor force growth is exogenous.

There are four main departures in our treatment of the Malay-sian economy from the analytical framework employed by Corden and Neary (1982). First, there is no requirement for trade balance in the sense of zero net exports. Second, and related to this, the terms of trade is endogenous. Third, the “behavior” of government is explicitly modeled. Finally, our model is dynamic rather than comparative static in nature. Each of these modifications lends greater realism to our calculations.

External balance in a real trade model, such as Corden and Neary’s, is normally identified with a balanced trade restriction. Usually this “closure” requires that domestic absorption adjusts to ensure that net exports are zero (or some other exogenously determined number). In our model, the current account balance is endogenous. The nominal exchange rate adjusts (together with other asset yields) to satisfy the conditions for asset market equilib-rium requirement, which include the requirement that the total

8_{In fact, Malaysia for a long time imported significant quantities of labor and faced}

demand for foreign exchange be equated with its supply.9

Never-theless, domestic absorption still responds to the current account balance through the influence the latter exerts on domestic wealth. Deficits, for example, reduce private wealth and consumption, thus allowing resources to be released from the non-traded to the traded goods sectors of the economy. This imparts stabilizing feedback to the balance of payments. In most simulations, we assume that the supply of foreign capital to the Malaysian economy is elastic. This assumption is not unrealistic, given Malaysia’s his-torical reliance on foreign saving.10_{We do, however, also provide}

estimates of the income losses that could occur if foreign saving were inelastically supplied. In our model, the repercussions of the current account are felt by domestic transactors through accompa-nying changes in their net private wealth.

The endogeneity of the terms of trade has important implica-tions for our analysis. To the extent that conservation leads to a reduction in foreign exchange, the Malaysian economy may have to endure a fall in the price of its nonlumber exports and an increase in the local price of its imports to replace lost foreign exchange. The magnitude of such changes will be larger; the larger is the price elasticity of demand for Malaysian lumber, and the smaller the price elasticity of demand for nonlumber exports.

Vincent (1992) notes that there are ready temperate substitutes for industrial roundwood and plywood produced from tropical lumber,11 _{and notes that competition from temperate woods is}

one reason why tropical lumber prices have not risen more quickly as world stocks have become depleted. Available econometric evidence supports this view and indicates that imports of tropical

9_{The balance of payments condition is, in effect, the external closure, and the exchange}

rate is determined together with endogenous asset yields to satisfy the condition for asset stock equilibrium (demand5supply) in each period in each asset market. Ours is a Branson-like model of exchange rate determination in which long-run equilibrium is characterized by the adjustment of the real exchange rate to ensure a current account balance consistent with a stable ratio of external debt to income. The steady-state debt to income ratio depends on initial conditions, the path of autonomous variables, and the model’s structure.

10_{Malaysia has frequently run current account deficits in the range of 5 to 10 percent}

of its GDP, although they have also posted more modest deficits in some years. Debt servicing has not proved problematic due to the fast nature of its income and export growth. Malaysia, of course, experienced a large withdrawal of foreign capital in the wake of the Asian crisis.

11_{There are of course specialty woods such as teak, mahogany, and nara for which}

lumber are highly sensitive to price (see the studies cited by Vin-cent, 1992, and Parthama and VinVin-cent, 1992). Therefore, our base-line assumption is that Malaysia is a price taker in world lumber markets.

For Malaysia’s manufacturing exports we impose export price elasticities that are comparatively large (>5). Our values are close to those estimated by Riedel (1989) for Hong Kong’s manufac-tures. However, Muscatelli et al. (1992, 1994) suggest that manu-facturing export price elasticities may be as low as 0.5. It is possible, therefore, that the nonlumber export price elasticities that we use in our reference simulations are upwardly biased. If such bias is present, we will tend to understate the terms of trade losses that would accompany lumber conservation. Because there is some uncertainty about the magnitude of export price elasticities for both nonlumber and lumber commodities alike, and because these play a central role in determining terms of trade changes, we also test the sensitivity of our results for departures from our baseline parameter assumptions.

A loss of lumber output and export revenue would result in an erosion of Malaysia’s tax base. It is unlikely that a fiscal deteriora-tion would go uncorrected. In our simuladeteriora-tions, we assume that the Malaysian government is precomitted to real levels of recurrent expenditure, to a path for monetary (and bond stock) expansion, and to targets for public debt (expressed as percentage of GDP). To accommodate these, we assume that the government adjusts (equiproportionately) income taxes and its capital expenditure.

The last important departure from a real trade model that we make is to provide a dynamic setting for our simulations. We consider income losses over a 10-year period. A comparative static analysis of the effects of conservation could be misleading because the adjustments to relative price and sectoral balance sheets are likely to be slow working in the presence of stock adjustments. Equally, the modeling of the steady-state effects of conservation is problematic.12 _{A 10-year period has been chosen because it is}

not too long to render our assumptions about the growth and structural evolution of the Malaysian economy fanciful, but it is sufficiently long to facilitate analysis in a context where there may

12_{We would, for example, have to define precisely what is meant by “balanced resource}

technological change” (Nordhaus, 1991) and to make conjectures about the composition of Malaysian far into the future.

be further rapid tropical deforestation and significant structural change in the economy.

The dynamic properties of our model resemble closely those of a neo-classical growth model. Aggregate economic growth reflects both factor augmentation and productivity growth (which is as-sumed to be Harrod neutral). We calibrate the model using 1990 Malaysian national accounts and related data. These data are then reproduced as our base year solution. The model is parameterized, and its exogenous data inputs set to generate a trend growth rate of GDP of 8 percent through to 1999. This growth rate is very close to the actual outcome over the period 1987–94. We configure our model in such a way that about 1 percent of aggregate GDP growth is attributable to growth of the “raw” labor force, and 3.5 percent each to the growth of the capital stock and to total factor productivity. The accumulation of human capital is subsumed in total factor productivity gains. Our assumptions imply substantial capital deepening over the period. In broad terms, we assume that the structure of the Malaysian economy evolves in such a way that it arrives by 1999 to the point the Korean economy had reached by 1990.13

A technical appendix provides fuller details of the model’s structure.

4. SIMULATION DESIGN

Recent evidence points to a leveling off in Malaysian lumber harvests following a period of rapid growth (Government Press, 1994). But stationary or even declining harvests do not necessarily imply the sustainable management of lumber stocks (zero net harvests). In Malaysia, lumber harvests would have to fall substan-tially to be brought into balance with the growth of lumber. Con-cern over available stocks prompted a temporary ban on log ex-ports that has been in force in Sabah since 1992. Sarawak has also announced plans to severely curtail logging in its forests.14

The absence of good data about stock levels and the shifting nature of policies toward logging raise some uncertainty about what “business as usual” might entail. For this reason, we measure

13_{By 1999, we assume that both secondary and tertiary sectors of Malaysia contribute}

about 38 percent each to aggregate output. This is close to the 1990 Korean figures. To some degree these assumptions have been overtaken by the Asian crisis.

14_{In Sarawak, the state government plans that the current (1993) log harvest of 18.8}

income changes from two alternative baselines. The first baseline (A) assumes that lumber harvests and related output remains stationary at their initial (1990) levels through to the end of the decade. The second (B) envisages that harvests will decline at an average rate of 8 percent per year. Such a decline in harvests could occur because of the “hollowing out” of forest areas or because of the adoption of policies intended to lead to the better management of Malaysia’s forest resources. It seems likely that the actual outcome will occur somewhere within the band defined by our baseline assumptions.15_{By applying these baselines in the}

measurement of income losses, we hope to identify a range within which actual losses might lie.

Unfortunately, it is not possible to identify our baseline assump-tions either with projected rates of deforestation or with reducassump-tions in lumber stocks. The stock-flow identities that govern the evolu-tion of lumber stocks require informaevolu-tion on opening balances, lumber growth rates, stocks destroyed by nonlogging sources of deforestation, in addition to assumptions about lumber harvested and replanted. Projecting the area under forest cover is more difficult still (see FAO, 1993). The data required for such projec-tions are not available for Malaysia. However, we can infer from the FAO data that, absent other changes, stationary harvests would mean that deforestation would continue at a rate of at least 2 percent per annum. This would imply that by the turn of the century a maximum of only 44 percent of Malaysia’s land area would remain under natural forest cover. An 8 percent decline in harvests will probably secure a larger stock of lumber stands, but this is unlikely to increase Malaysia’s forest land area. Indeed, the Malaysian government currently plans to transfer another 3.5 million hectares of state forests to other land uses, which is a little over 10 percent of its remaining forest area.

The issue of whether an incentive-based approach to conserva-tion is preferable to a command-and-control approach is complex. In practice, elements of both approaches may have to be applied, and the dividing line between the two is not always clear.16_Markets

in forest and ecological resources are difficult to create but,

15_{Recent data indicate that largely because of quotas imposed by state governments}

sawlog production fell by 5.1 percent on its 1993 level. Exports of sawlogs fell by 14 percent in volume terms. See Government Press (1994).

16_{For example, even for tradable permits associated quantities (or units) still have to}

equally, the administrative and efficiency costs of regulatory ap-proaches to conservation may be large. Command and control policies make most sense where either there is considerable uncer-tainty about the effects of incentive based approaches, or the objective is to eliminate an unwanted activity altogether.17 _The

types of conservation policy we have in mind include permanently gazetting larger areas as protected or reserve forest and, more generally, tightening land use regulations in relation to existing forest areas. In our model, we represent “conservation” in terms of lower lumber harvests (flow outputs), and a commensurate reduction in stock inputs (composite land and capital) used in lumber production. To reflect possible constraints on agricultural extension, we restrict net investment flows in agricultural sectors. Limiting the encroachment of other activities, principally tree crop plantations, on forest area is likely to be an important aspect of conservation. In our model, tree crop output is endogenous, but will depend on the growth of (exogenous) net investment expenditures. The conversion of forest area for plantations entails substantial net investment in land conversion. Therefore, by reduc-ing net investment expenditures in tree crop and other agricultural sectors we can, to the extent that capital and land are complemen-tary, mimic land use restrictions.

We measure conservation costs using a “compensating varia-tion” measure. The compensating variation is the amount that households would have to receive to leave their consumption unchanged following conservation. In our model, we identify the compensating variation through the increase in unrequited trans-fers from the rest of the world (to domestic households) needed to satisfy the consumption constraint. Note that although these compensation payments are nondistortionary in the sense that they have no direct substitution effects, alternative lump-sum transfer could yield different results.18

Our measure of the compensating variation is: NCV5

o

1999 t51990

dYrow t

(11 r)t219904

o

T

t51990 Yt

(11 r)t219903100

u(Ct(dYrowt . . .)5Ct(. . .)∀t), TPh1990,1999j, (1)

17_{However, the announcement of a pending ban on logging might succeed only in}

accelerating deforestation if it further weakens concession holders tenure.

18_{This is because their general equilibrium (indirect) effects will usually be different.}

Note that small changes in the value of the compensating variation can occur when we vary the elasticity to consume from income.

where dYrow _{is the change in real endogenous income transfers}

that, in each period, maintains private real consumption at its preconservation (baseline) level, Ct, r is the discount rate, Y is real gross domestic income (GDI) in the preconservation (base-line) economy, andtindexes time. We report income losses both in base year (T 5 1990) GDI units and in units of the present value stream of GDI (T51999). We measure real income using a GDI rather than a constant price GDP measure because there are, as we shall see, large terms of trade changes associated with tropical forest conservation. GDI is equal to (constant price) GDP plus a terms of trade adjustment defined as the “capacity to im-port,”lessthe volume of exports. The “capacity to import” equals nominal exports divided by the import price deflator. This defini-tion of GDI is that used in the World Bank’s World Tables.

To draw “welfare” implications from our analysis, our measure of the compensating variation would have to be adjusted to reflect the increase in the nonlumber value of forests created by conserva-tion. Note, however, that because we assume that trees and land area is permanently transferred to nonlumber uses, we do not have to account for the increase in the value of the terminal lumber stock. Our assumptions imply that the lumber value of trees is lost as nonlumber values are created through conservation. Although it would be interesting to consider policies that preserved trees today with a view to supporting larger sustainable harvests in the future, we do not have the stock-flow information that would be needed to calibrate the required model of lumber growth. Also, were the policy issue one of sustainable harvesting, the income losses in Equation 1 would be offset by the present value of the (potentially infinite) sequence of future lumber harvests made possible by the deferral of harvests in earlier periods.

5. SIMULATION RESULTS

5A. The Anatomy of Income Changes

To facilitate identification of the various mechanisms through which income losses are generated, we begin by measuring the effects of a 50 percent reduction in lumber flow outputs and stock inputs over their 1990 levels. This entails a uniform reduction in the levels of flow output and stock inputs in the lumber sector in each year. In all simulations, the reductions in flow output and stock inputs take effect in 1991 and continue through to the termi-nal year (1999). In 1990, investment demand is reduced to reflect

the lower level of stock inputs applied in lumber activity in later years. In this simulation, we place no restrictions on agricultural extension and employ baseline A.

The compensating variation (NCV) under these assumptions is 2.2 percent (T51990) of the capitalized stream of baseline GDI, and 18.2 percent when measured in base year GDI units (T 5

1990). In Table 1 (row 1) we show the values of the compensating transfers from the rest of the world in each period, expressed as a proportion of contemporaneous GDI. In the remaining rows we show lumber’s baseline and conservation income shares (rows 2 and 3), and we also record the impact of our conservation assumptions on a variety of variables when compensating transfers are suppressed (rows 4 to 14). On suppression of compensating transfers, GDI falls by 2.3 percent of its (capitalized) baseline value. This is very close to the income loss registered in theNCV. This is because the incidence of income losses falls almost entirely on households. Note also that the percentage reductions in private real consumption (row 5) are about twice the magnitude of the percentage declines in income. This is because the incidence of income losses falls almost entirely on households, and the con-sumption base from which they are measured is less than 60 per-cent of GDI. Reductions in wealth to some degree acper-centuate consumption losses.

The most notable aspect of these results is that the magnitude of the overall decline in income is large compared to the initial downsizing of lumber activity. Since, over the period 1990–99, the share of lumber in GDI averages 2.1 percent in our baseline, a 50 percent reduction in output and stock inputs leads to a direct withdrawal of income of around 1 percent of GDI, some of which is recovered when mobile factors released from lumber find em-ployment elsewhere in the economy. Therefore, it follows that the indirect loss of income is at least as great as the initial direct withdrawal. In this particular simulation the dynamic general equi-librium income multiplier is of the order of 2.

To understand why income losses are magnified, it helps if they are decomposed. We start with a simple household income identity:

y;W

Pc·L1 u· R

Pc·K. (2)

In Equation 2,y is household income measured in consumption units,W is the nominal wage, Pc _{is the consumption deflator,L}

F.

Harrigan

Table 1: Summary of the Effects of Log Harvest Restrictions (% Differences from Baseline A)1

Row Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

1 Remittances (%GDI) 0.03 4.48 2.97 2.55 2.44 2.31 2.16 2.02 1.83 1.72

2 Baseline share lumber (%GDI) 3.06 2.88 2.67 2.51 2.32 2.18 2.04 1.91 1.78 1.66

3 Conservation share lumber (%GDI) 3.06 1.43 1.34 1.26 1.16 1.09 1.02 0.95 0.89 0.83

4 GDI 20.02 23.23 22.87 23.07 22.75 22.40 22.28 22.19 22.01 21.85

5 Consumption 20.12 28.46 26.05 22.62 23.71 24.84 24.39 23.88 23.79 23.78

6 Exports 0.46 4.19 1.94 21.03 0.56 1.98 1.55 1.19 1.40 1.62

7 Imports 20.46 22.23 21.45 21.07 20.75 20.57 20.49 20.39 20.22 20.10

8 Terms of Trade 20.13 24.35 23.25 21.90 22.31 22.63 22.29 22.00 21.93 21.88

9 External debt to GDI (%)2 _{20.36 22.33 21.55 1.39 1.94 1.48 1.67 2.02 2.03 1.83}

10 Real consumption wage 20.23 28.41 26.38 24.45 24.29 24.58 24.24 23.78 23.49 23.36

11 Unemployment rate unskilled2 _{0.01 21.07 0.36 0.84 0.15 20.04 0.17 0.18 0.08 0.03}

12 Lending rate2 _{20.01 20.22 1.86 1.00 0.02 0.18 0.36 0.18 0.02 20.03}

13 Aggregate physical capital stock 0.00 22.55 22.53 22.36 22.24 22.12 21.99 21.87 21.75 21.65

14 Agricultural labor force 0.00 20.02 20.63 21.23 21.73 22.05 22.25 22.41 22.51 22.57

1_{Lines 4 to 14 refer to results from a simulation in which we compare pre- and postconservation outcomes, having relaxed the consumption}

constraint. A discount rate of 10 percent is applied in the calculation of the compensating variation. Because the discount rate enters both the numerator and denominator of Equation 1, the compensating variation is comparatively insensitive to the choice of discount rate (see also Table 4).

is aggregate labor units (assume full employment of labor), Ris the nominal capital rental rate,Kis the stock of capital and 0, u ,1 is the share of “profits” accruing to households. To prevent the algebra from getting cluttered, taxes are suppressed, labor is aggregated over markets and capital over sectors, and household claims on profits are assumed to be parameteric in Equation 2.

Applying the difference operator to Equation 2 and manipula-tion gives19

Dy;t1· (Dw· (L2 c·LF)1 u·Dr· (K2 c·KF))1 Dt·y0 2 t1· ((w0·c·LF1 u·r0·c·KF)

2(w1·c·LF_{1(12 l) ·u·r1·c·K}F_{)), (3)}

where D is the difference operator, w is the real product (not consumption) wage,r is the real product rental, t is the ratio of (current period) local producer prices to consumption prices, c

is the percent reduction in lumber activity over the period (ex-pressed as a decimal fraction),lis the share of the lumber capital stock that is immobile, the subscripts 0 and 1 index the periods before and after the reallocation of labor and the loss of lumber capital, and the superscriptFdenotes factor quantities initially in the lumber sector.

The first term in parenthesis on the RHS of Equation 3 is a measure of the change in household income arising from changes in real product factor prices for given endowments of labor and capital. The second term measures the loss in the command of initial household income over consumption goods when consumer prices rise relative to the cost of local output. We attribute this effect to terms of trade changes. The third expression has two components. The first part is the direct loss in household income that would occur if a fractioncof lumber capital and labor income were eliminated, and all factors were immobile. The second term is the recovery of household income that occurs when mobile factors initially employed in lumber are reallocated to other sectors.

Using Equation 3 we can now condense the income changes that are reported in Table 1. We do this for the change in GDI

19_{Changes in factor incomes are weighted by the current ratio of producer to consumer}

prices (the “terms of trade”), and the “terms-of-trade” change in the second term of Equation 3 is measured from base income. Income changes can also be decomposed where the weight on factor income changes is the base ratio of producer to consumer prices, and “terms-of-trade” changes could be measured for current rather than base household income.

in the impact period, 1991, reported in row 1 of Table 1. It is important to understand that these are the income changes that would occur in the absence of compensation. Also, to facilitate comparison, the left and right hand sides of Equation 3 are divided through by base GDI in 1991. All values in Table 2 are calculated directly from the results of model simulations or are defined by the nature of the experiment itself.

Table 2 identifies “terms-of-trade” changes, as reflected in the change in the relative price of producer to consumption goods, as the effect that contributes most to the net income loss, account-ing for just under half of the total decline.20 _{Next in importance}

is the decline in the real product wage that occurs when lumber capital is withdrawn from production. This accounts for around one-quarter of the net decline in income. The incidence of real wage reductions would, of course, be felt most directly by those workers and households in the agricultural sector, and by unskilled workers elsewhere. Therefore, it seems that conservation would be likely to have an adverse effect on the distribution of worker and household income.21 _{However, the reduction in household}

income caused by increased taxes is small, and accounts for only 5 percent of the net loss. Our computations also make it clear that a substantial proportion of the direct loss of lumber income is recovered both when mobile factors are reallocated, and when the return to capital increases as its stock shrinks. Taken together, just under 65 percent of the direct loss of lumber income is recov-ered through the reallocation of mobile factors and through in-creased profits.

The terms-of-trade–induced income losses that are highlighted in Table 2 (and row 8 of Table 1) occur as the price of imports rise relative to the price of exports. The real exchange rate also depreciates as the price of traded to non-traded goods increases.

20_{This ranking and the respective contributions may, however, change over time.}

Spe-cifically, once wages fall in the agricultural sector this will lead to out-migration and there will be income losses as migrants experience temporary unemployment in nonagricultural labor markets. These effects are, however, small, and wage adjustments are fast. Also, when investment is reduced in agriculture to mimic land restrictions, this will represent an additional source of income loss. Space considerations preclude a decomposition for each and every year.

21_{There is only one household identified in the version of the model used in these}

simulations so that it is not possible to measure distributional effects directly. It is unlikely that households with unskilled workers are likely to benefit much from the increase in capital income.

FOREST

CONSERVATION

509

Table 2: The Decomposition of Income Changes in 1991

Household income changes to GDI Value (% GDI) Annotation

Total 24.48 Value of NCV for 1991, Table, row 1.

Terms of trade 22.20 Calculated as in Equation 3, expressed in real GDI units

Labor share of factor product 21.40 Calculated as in Equation 3, expressed in real GDI units

Capital share of factor product 10.33 Calculated as in Equation 3, expressed in real GDI units

Note:households total claim on operating surplus is approx.

38 percent, and total profits increase by 1 percent of GDI in our simulation. The factort1equals 0.868.

Direct loss of lumber income 21.30 Calculated as in Equation 3. This is 50 percent of baseline

share of GDI recorded in Table 1

Recovery of mobile labor income 10.51 Calculated as in Equation 3. Expressed in real GDI units

Note:initial share of labor in Lumber value-added is 0.43

percent, and that the product wage of mobile labor falls in the simulation

Recovery of capital income 0.00 Capital is assumed ex-post immobile (l 51) (see p. 6)

Tax losses 20.27 Calculated from model simulation

Total losses 24.33 Summation of above

These relative price changes facilitate a movement of resources into the traded goods sector and leads to the recovery of some of the foreign exchange that is lost as lumber exports contract. In raising the cost of consumption relative to the price of domestic output, the deterioration in the terms of trade reduces real income through the Laursen/Metzler effect.22

The dynamics as well as the composition of income losses are of some interest. Although the negative impact of conservation on income and consumption recedes as lumber activity contracts, losses nevertheless persist (see Table 1, rows 1 and 4). Because we have assumed that the withdrawal of lumber capital is permanent, lower product wages in the agricultural labor market and capital incomes in the lumber sector endure. Also, an emerging deficit on the current account reduces private sector wealth and, with a lag, leads to a (stabilizing) fall in private consumption (rows 5–9 of Table 1). Finally, the resource gap that the current account deficit mirrors, causes a small increase in domestic real interest rates. This raises the user cost of capital, lowers Tobin’s “q,” causing the aggregate capital stock to fall (row 13 of Table 1). This further reduces the real product wage and household wealth. The qualitative nature of the resource switching effects that occur with lumber conservation are exactly as would be predicted by a real trade model. Output in those sectors that are compara-tively sheltered from world trade (Construction, Utilities, Public Services, Private Services, and Domestic Manufacturing) contract following conservation. The largest gains in output are in those sectors that we assume are most open to world trade (Export Manufacturing). By the terminal year of our simulation (1999), constant price value added in Construction falls over baseline by 2.8 percent and value added rises by 4.2 percent in Export Manufacturing.

5B. The Results with Alternative Baselines and Alternative Conservation Policies

Having now explained the basic mechanisms through which conservation reduces “metered” income, we now turn to the issue

22_{The income effect of exchange rate changes is often referred to as the Laursen/Metzler}

Table 3: Variations in theNCVwith Baseline and Conservation Assumptions

Conservation assumptions

Baseline Harvests Land use NCV T51999 NCV T51990

Stationary (A) 250% — 2.23% 18.23%

28% pa (B) 250% — 1.10% 8.91%

28% pa (B) 250% Restricted 2.16% 17.62%

Stationary (A) 250% Restricted 3.34% 31.20%

28% (B) 297.5% — 3.30% 26.77%

28% (B) 297.5% Restricted 4.32% 35.37%

Stationary (A) 297.5% — 4.39% 35.41%

Stationary (A) 297.5% Restricted 5.46% 44.85%

of how losses change with variations in our baseline and conserva-tion assumpconserva-tions (see Table 3). Consider first what happens when baseline lumber flows and stocks fall at an annual rate of 8 percent annum (i.e., baseline B), rather than remain constant (as in base-line A).23 _{Not surprisingly, this reduces income losses and the}

NCVindex falls to 1.1 percent of the capitalized stream of GDI, and to 8 percent measured in base year (1990) GDI units.

We noted earlier that the conversion of forest land area, espe-cially for commercial agricultural uses, has been a major source of deforestation in Malaysia. It seems likely, therefore, that conser-vation would also restrict agricultural encroachment on forest land area and inhibit the growth of agricultural output. If gross investment in the agricultural sector is set to replacement level, growth of agricultural output can then only occur either through technological progress or the application of more labor. This is precisely the way in which production possibilities in agriculture would be constrained if capital and land were complementary and further agricultural extension on forest area was prohibited. Of course, it is possible that agricultural extension could still take place on nonforest lands so that our treatment may exaggerate the effect of land restrictions on income. Moreover, to the extent that capital and land are substitutable, output losses could be ameliorated through the adoption of more capital intensive meth-ods of farming. Although substitution possibilities are likely to increase with the elapse of time, it would appear that in the

23_{In this simulation baseline lumber output is in fact a little lower than conservation}

production of Malaysian tree crops capital cannot be easily substi-tuted for other factors.24

In our baseline simulation, the growth of output of the tree crop sector averages 3.7 percent per annum. This is very close to the 3.8 percent increase in the volume output that was observed between 1990–93. On the imposition of land restrictions, output growth falls to an average of 1 percent per annum over the period 1990–99. As a consequence, income is further eroded and the NCV(T51999) increases by about 1 percent. This result is largely insensitive to the baseline assumptions applied.25

By comparing the effects of a 50 percent reduction in harvests with those of a 97.5 percent reduction, some indication of the marginal costs of conservation can be obtained.26_{For those cases}

where baseline lumber harvests are assumed to remain stationary (at 1990 levels), marginal costs are almost constant. Over the first 50 percent reduction in lumber activity, the compensating variation averages 0.0446 percent of GDI for each one percent reduction in activity. Over the next 47.5 percent reduction in the level of lumber activity, the average compensating variation increases to 0.0455 percent of GDI. Where we assume that baseline lumber activity declines by 8 percent per year, a different picture seems to emerge. Now, over the first 50 percent reduction in lumber activity, the average loss is 0.022 percent of GDI for each percent reduction in activity. But this figure rises to 0.046 percent of GDI over the next 47.5 percent reduction. This increase in cost occurs because with declining baseline output, the proportional contraction in lumber output approximately doubles as the level

24_{In recent years, output of the tree crop sector in Malaysia has been constrained by}

shortages of unskilled labor. This in itself has helped slow the pace of forest land conversion. Opportunities for the adoption of more capital intensive techniques seem limited (Govern-ment Press, 1994).

25_{It should be understood that the apparent invariance of the cost of land restrictions}

to the underlying level of harvests reflects the assumed independence of capital and land inputs in lumber and agricultural sectors. We cannot draw any inferences from these results about an “optimal” policy mix of restrictions on output and land use. Such a judgment would ultimately rest upon whether the objective is to preserve lumber stocks, forest area, or some combination of both. It would also reflect the desired attributes of the preserved forest area and of the implementation and monitoring costs of the different restrictions.

26_{For technical reasons, it was not possible to eliminate lumber harvests completely.}

In any case, a complete ban on lumber logging would be almost impossible to police. It is likely that illegal logging would persist even under the strictest conservation regime.

Table 4: Calculations ofNCV(T51999)

n50.04 g5

n50

r 0.08 0.07 0.06 0.05 0.04 0.03 g50.03

0.10 0.803 1.131 1.421 1.677 1.905 2.108 1.597 0.09 0.615 1.000 1.338 1.638 1.905 2.143 1.564

0.08 — 0.802 1.214 1.580 1.905 2.195 1.523

0.07 — — 1.009 1.483 1.905 2.281 1.469

0.06 — — — 1.289 1.905 2.454 1.398

0.05 — — — — 1.905 2.971 1.298

n50,r 50.1 0.604 0.853 1.072 1.267 1.441 1.597

reduction increases from 50 to 97.5 percent of initial (1990) output. Therefore, it appears that income losses increase almost linearly with the decline in lumber output.

A limitation of our analysis is that we have calculated costs over a time horizon that is short relative to the life cycle of tropical lumber species. It is also short relative to the period over which the benefits from conservation could be expected to accrue. As we have already observed, lengthening the simulation period would render our results sensitive to speculation about the rate and form of Malaysian economic development into the distant future. Nonetheless, by extrapolating from our results we can provide some indication of what costs might be over a longer period. An infinite horizon, approximation to Equation 1 is:27

NCV*5NCV1990·l 1(12 l) ·NCV1999·(r 2g) · (11n) (r 2n) · (11g)

ur .g,n T51999, (4)

where g is the (post-1999) assumed steady-state growth rate of GDI,nis the assumed steady-state growth rate of income compen-sation,ris the discount rate, NCV1990 is the observedNCV (T5

1999) value for the period 1990–99, NCV1999 (T 5 1999) is the

observed ratio of compensating income remittances to GDI in the terminal year, andl(g,r) measures discounted GDI over the period 1990–99 relative to the limit of the series. In table 4, we

27_{This is the limit of Equation 1 as time approaches infinity assuming all variables follow}

summarize the results that we obtain by applying this formula. Results are based on the same assumptions used for Table 1.

It can be seen that lowering the assumed steady-state growth rate of output increases measured costs. This is because the promi-nence of lumber activity in the economy now increases. Although, in general, lowering the discount rate reduces costs, it does so only where the rate of growth of output (g) is greater than the rate of growth of compensating remittances (n). If remittances grow more quickly than aggregate output, suggesting a post-1999 increase in the contribution of lumber to aggregate Malaysian income (something that is highly unlikely), a reduction in the discount rate will, in fact, magnify costs. Note also that varying the rate of growth of compensating remittances has a larger effect on costs the smaller the discount rate, and rate of growth of output. It is reasonable to infer from these results that long-run costs are likely to be less than our finite horizon estimates, but it is difficult to be precise by how much.

5C. Sensitivity to Model Assumptions

Finally, we consider how sensitive our results are to our model specification. There is, of course, in any CGE model a large num-ber of “free parameters” and a large numnum-ber of behavioral assump-tions that can be varied. Even where theory informs us about the qualitative effect of such changes, their quantitative implications may not be clear. Therefore, in Table 5, we report and annotate the effects that variations in some of our more important assump-tions have on costs. Again, all changes are measured from the results reported in Table 1.

As we cautioned earlier, terms of trade effects are likely to be sensitive to our assumptions about export price elasticities for lumber and nonlumber products. In Table 5, we report the result of three experiments that bear on this matter. First, we halve all nonlumber export and import price elasticities, and recalculate income losses. Recall, that some evidence (Muscatelli et al., 1993) suggests that our reference elasticities may be biased upwards. When elasticities are halved, income losses (as measured by the NCV) increase by 0.242 percent of GDI over their reference value (2.2 percent of GDI). Next, in a symmetrical fashion, we double all reference nonlumber export and import price elasticities. This reduces income losses by 0.176 percent of GDI. Taken together, these results suggest that income losses respond nonlinearly to

FOREST

CONSERVATION

515

Table 5: Summary of Sensitivity Tests

DinNCVpercent

Assumption GDI (base52.2%) Explanation

Halving trade price elasticities in all but those sectors 10.24 Greater independence of domestic from world prices, where Malaysia is assumed to be a price taker in world terms of trade losses are increased

markets.

Doubling trade price elasticities in all but those sectors 20.18 Model more closely reflects a “law of one price” economy,

where Malaysia is assumed to be a price taker in terms of trade losses are now more muted

world markets.

Unit price elastic lumber exports 21.1 Lumber foreign exchange revenues are invariant with

respect to the flow of lumber output, income losses are reduced

Imposition of a constraint on the ratio of terminal external 10.8 The current account now exerts stronger effect on terms

debt to GDI of trade, income losses are increased.

No lumber capital income accrues to domestic households 20.29 Partial elimination of the income losses that arise because of capital immobility, income losses are reduced

trade price elasticities, with marginal changes in theNCVtapering off as price elasticities increase. If nonlumber trade price elasticit-ies were, in fact, as low as 0.5 (see, Muscatelli et al., 1993), then losses would rise from 2.2 percent to 3.1 percent of GDI.

We now examine the effect of the lumber export price elasticity on income losses. Recall that we have assumed that the foreign exchange price of Malaysian lumber is set independently of Malay-sia’s supply (see, Section 3). Now we model income losses assuming that there is a unitary export price elasticity for Malaysian lum-ber.28_{Accordingly, as Malaysia’s lumber exports decline the price}

of Malaysia’s lumber rises to maintain lumber foreign exchange revenues. Not surprisingly, the “sterilization” of lumber foreign exchange losses ameliorates the impact of the deterioration in Malaysia’s terms of trade. The compensating variation falls by about 1 percent of GDI. Reassuringly, this change in income compensation is consistent with the share of total losses that we attributed to terms of trade changes in Table 2. A lower export price elasticity for lumber implies that a larger incidence of conser-vation costs fall on nonresidents.29_{Note, however, that if}

conserva-tion were to entail the complete cessaconserva-tion of lumber trade, then, irrespective of the price elasticity of demand for lumber, there could be no mitigation of foreign exchange losses.

The composition as well as the scale of output losses change when foreign exchange losses are sterilized. A more moderate depreciation of the real exchange rate now means that the local currency price of traded to non-traded goods does not rise by as much as previously. Real profits in non lumber export sectors are, as a consequence, comparatively depressed: Tobin’s “q” falls, and the incidence of output losses falls more heavily on real investment expenditure. Net exports expand by much less than previously,

28_{Our objective is to “sterilize” foreign exchange losses as lumber exports are withdrawn.}

We cannot do this exactly because different users of lumber have different price elasticities of demand, and face different markups over basic price to accommodate taxes, transporta-tion, and distribution margins. In the simulation reported in the text, direct foreign exchange losses are sterilized almost exactly in the base period, but the effectiveness of sterilization is gradually eroded and by the end of the simulation period only 80 percent sterilization of the impact foreign exchange losses is achieved.

29_{If themarketelasticity for lumber were below unity, collusion between producing}

nations might both increase their income and help conserve forest area, perhaps even eliminating the need for compensation. If the elasticity facing Malaysia as a single producer were below unity, it could improve its terms of trade by restricting output without coordinat-ing its actions with other producers.

and private consumption falls by much less. However, as reduced real investment cuts into the capital stock, the growth of real wages in inhibited. This changes the temporal profile of consump-tion losses. Real consumpconsump-tion losses, though smaller, now show much more persistence.

Next, we consider the effect that our “external closure” has on measured income losses. External debt rises in our reference simulation, but by only 1.83 percent of GDI by the terminal year of our simulation (see Table 1). Although this could probably be easily financed, it is worth considering what might occur if there were a constraint on foreign saving. In such circumstances, the income losses associated with conservation would increase. Do-mestic absorption, including household consumption, would then have to fall to accommodate the reduction in the supply of saving from nonresidents. To a degree, we can mimic the required adjust-ments by precluding substitution between domestic and foreign liabilities. The current account deficit now contracts and the in-come compensation needed to compensate domestic consumers rises by 0.22 percent of GDI. The smaller current account deficits that now occur reduce the ratio of terminal debt to GDI by 0.5 percent. We infer from these responses that if a constraint on foreign borrowing were to be expressed in terms of a cap on the ratio of external debt to GDP, say set equal to its baseline terminal value, this would increase overall income losses from 2.2 percent to about 3.0 percent of GDI.30

Finally, we consider what would happen to income losses if households received no rents or operating surplus from lumber activities. This has the effect of eliminating some income losses because we have assumed that lumber capital income is entirely lost when lumber values are surrendered to other forest uses. Imposition of this assumption, reduces income losses fall by nearly 0.3 to 1.9 percent of GDI. However, because labor productivity is still adversely affected by the immobility of the lumber capital stock, not all indirect losses are eliminated in this simulation. If we were to assume that lumber capital were fully mobile, the results of Table 2 (row 5 less row 6) suggest that income losses

30_{It is not possible to impose this constraint directly in the model. We arrive at this}

inference by dividing the 1.8 percent increase in terminal debt (Table 1) by the reduction of 0.5 percent reported in the text. If income losses respond linearly to changes in debt, multiplying (1.840.5) by the 0.22 percent increase in income losses, gives an overall increase in income losses of about 0.8 percent of GDI.

of up to 0.8 percent of GDI might be recovered in the impact period (1991).31

6. CONCLUSIONS

In this paper we have estimated the loss in “metered income” that Malaysia might experience if it conserved its tropical forest resources. Because no account has been taken of the possible benefits of conservation that arise through the increase in the nonlumber value of trees preserved, our estimates should not be interpreted as net costs, nor should they be identified with prospective changes in “welfare.”

Our principal findings are summarized in Tables 1, 2, and 3, and the results of our sensitivity analyses in Tables 4 and 5. For a given conservation policy, “metered” income losses are smallest if mobile resources are efficiently and quickly reallocated, foreign exchange revenues are resilient to the loss of lumber output, and “business as usual” would have entailed fast deforestation. How-ever, even under these conditions, a 10-year moratorium on lum-ber harvesting and associated land use restrictions could easily generate income losses of over 2 percent of the capitalized stream of GDI. A more probable estimate of the income losses would be between 3 to 4 percent of GDI. Of more general significance are the large values of the dynamic general equilibrium income multipliers that these numbers reflect. Where lumber makes a significant contribution to foreign exchange earnings and these are adversely affected by conservation, direct income losses are amplified by terms of trade losses that more than outweigh the income, which is recovered as mobile factors are reallocated, and profits rise.

To facilitate the calculation of the compensating variation we have assumed that Malaysian residents are directly compensated by nonresidents by an amount sufficient to maintain their precon-servation consumption stream. In reality, arrangements like this would be impractical. Therefore, in concluding, it is worth briefly considering mechanisms through which the delivery of environ-mental benefits could be tied to compensating resource transfers.

31_{This estimate, however, takes no account of the general equilibrium repercussions}

that would follow. Also, to the extent that labor and capital are efficiently allocated to begin with, income recovery would necessarily be less than 0.8 percent of GDI.

“Debt for nature swaps” are one way in which resource transfers have been linked to environmental rehabilitation and protection programs. However, “debt for nature swaps” have typically relied on the resources of nongovernment organizations and, in practice, have only involved small projects and resource transfers. It seems unlikely that such arrangements could work for large areas of forest, and large debt write offs. Coordination problems and the risk of moral hazard would be just two obstacles that would prevent “debt for nature swaps” working at a macro scale. Another route through which widely shared environmental concerns, including tropical deforestation, could be addressed is through the establish-ment of an appropriate multilateral structural adjustestablish-ment facility. However, the scale on which such a facility would have to be funded makes this unlikely.32

Perhaps, more realistically, trade policy could be used to encour-age sustainable forest manencour-agement and pricing practices. How-ever, a carrot rather than a stick approach is required (see Repetto, 1994). Impeding trade in tropical lumber (e.g., through boycotts, quotas or tariffs) would depress stumpage values, and thereby hasten deforestation. Besides, such actions would probably now contravene WTO rules. On the other hand, arrangements linking tropical forest conservation to improved market access for both the processed and manufactured goods of tropical lumber nations would be both environmentally beneficial and trade creating.33

APPENDIX

General Model Description

M4_{is, broadly speaking, of the family of financial CGE models}

described by Robinson (1991) or by Bourguignon et al. (1992). It has 13 goods markets (including a separate market for Forestry), four labor markets, and five asset markets. There are six institu-tional transactors in the model who participate in asset markets

32_{If, for example, Malaysia were to receive a transfer of 2 percent of its GDI to}

compen-sate it for abandoning the exploitation of its lumber resources this would cost around US$14 billion (1994 prices) over 10 years. By comparison, the recently created Global Environmental facility has an endowment fund of only US$4 billion.

33_{This assumes that any accompanying acceleration in industrialization will not itself}

degrade the environment by more than deforestation did. For this reason, concessional access might also be given to best practice and environmentally clean technologies. Repetto (1994) argues that developing countries should also adopt the “polluter pays” principle.

and for whom flow and stock accounts are separately collated. The model has no extrinsic dynamics, other than through its adaptive treatment of expectations, but it has extensive intrinsic dynamic relationships, governing the accumulation of all physical and fi-nancial stocks. Different kinds of closure restrictions may be easily imposed in M4_{, but here we restrict attention to those that we use}

in the text. In general, markets clear in each period (one calendar year), though there are some elements of quantity adjustment in goods markets, and wages in the market for unskilled labor re-spond sluggishly to excess labor supplies. Asset markets are cleared at the end of each period through the adjustment of interest rates and the nominal exchange rate.

In the following description we simplify many elements of the model to expedite presentation.

Producer Decisions

In each activity, except Public Services and Dwellings (which is an account for imputed rent), the representative producer maxi-mizes profit or minimaxi-mizes cost subject to multilevel technology constraints. The solution of these problems generates the demand for all variable factors of production, including intermediate com-modity inputs and different categories of labor. In the production of Public Services output a fixed coefficient technology is em-ployed. In Dwellings, net output is simply equated with imputed rent, calculated as the product of the user cost of housing and the beginning of the period stock of private dwellings.

The multilevel technology tree has three layers in each sector. Activity gross output is modeled as a CES function of an aggregate intermediate input and value added. The aggregate intermediate input is an aggregate of composite commodities, and composite commodities are themselves Armington aggregates of domestic and imported commodities. Value added in nonland-based activi-ties is produced by a Cobb-Douglas technology employing fixed capital (including land and other fixed factors) and composite labor. In long-based activities, value added is produced with a CES technology where substitution elasticities are low (0.25). Composite labor is a CES aggregation of professional and skilled, semiskilled, and unskilled occupational categories.

For each activity, these relationships generate factor input de-mands and associated unrestricted and restricted (at the value-added level) cost functions. Ignoring the complications that are

created by margins, by distinctions between traded and non-traded goods, and by joint production these relationship yield:

pjmji5pxwi·er· markup pjdji5pxi· markup

pxi5pces(pji,pvi)|iÓLOP pxi5pxwi·er|iPLOP

pji5pces(pjjji)

pvi5rpcd(wi,vai,ki[t21])|iÓLOP, iÓL pvi5rpces(wi,vai,ki[t21])|iÓLOP, iPL

pvi5ipces(pji,pxi)|iPLOP pjjji5parm(pjdji,pjmji)

wi5pces(wni) wni5 tni·wn,

where: pces is an unrestricted CES cost function; rpces is a re-stricted CES cost function;rpcdis a restricted Cobb Douglas cost function;ipcesis an inverted, unrestricted CES cost function;parm is an Armington price function, the index LOP denotes sectors where the law of one price operates and L indexes land-based activities (Export Agriculture, Domestic Agriculture, Forestry). pxwis the exogenous world commodity price in foreign currency units;eris the nominal effective exchange rate;pxis the domestic commodity price; pv is the price of value added; pj is the price of the intermediate aggregate;wis the nominal wage;vais quantity value added; k is the capital stock; pjd is the price of domestic intermediate commodities, andpjmis the price of imported inter-mediate commodities (both in purchasers’ prices). The last equa-tion lets wages paid to identical labor categories in different sectors vary (see text). The purchasers’ prices of domestic commodities sold to different uses are calculated as markups onpx in exactly the same way aspjd. The indexnis used here to identify different labor categories and markup is used where goods or factor prices are marked up over factor cost or theircif price. Note that time subscripts have been dropped except where lagged (beginning of the period) quantities or prices enter into the determination of a variable.

The counterpart factor demand relationships of our model are: ji5fces(pxi,pji,xi)

vai5fces(pxi,pvi,xi)|iÓLOP

vai5irpcd(pvi,wi,ki[t21])|iPLOP, iÓL jjji5fces(pjjji,pji,ji)

ni5fcd(wi,pvai,vai) nnni5fces(wni,wi,ni)

jdji5arm(pjdji,pjmji,jjji) jmji5arm(pjmji,pjdji,jjji),

where:fcesis a CES factor demand equation;fcdis a Cobb Douglas factor demand equation; irpcd is an inverted, restricted Cobb Douglas cost function;irpces is an inverted, restricted CES cost function, and arm is an Armington demand function; x is gross output; j is the aggregate intermediate input; jj are composite intermediate commodities; n is composite labor; nn is labor by skill; jd is domestic output sold to intermediate uses, and jm is imported output sold to intermediate uses.

Factor and Institutional Incomes

Here we deal explicitly only with households, government, pri-vate, and public corporations and the rest of world. The relation-ships for other transactors are suppressed to expedite presentation. Factor incomes are calculated as the product of factor prices (less taxes, plus subsidies) and the stocks of factor inputs em-ployed, summed over all activities. Operating surplus is deter-mined residually after subtracting gross emoluments from nominal value added. The aggregate operating surplus of the economy is then distributed between its different institutional transactors.

Household’s disposable income is calculated as: yh5

o

n

o

i

win·nnin2taxes1transfers1asset income 1unincorporated business income.

Included with “transfers” are the unrequited transfers from over-seas residents that adjust to accommodate our consumption con-straints (see text). Asset income here and for other transactors is calculated as the product of lagged (beginning of period) stocks and lagged interest rates, so that asset income (which may be negative) is received only after transactors have held stocks for one period. Although not shown, asset revaluations are accommo-dated in our calculation of all incomes.

“Corporate” net income is calculated as:

yc5operating surplus1transfers1asset income2taxes,

wherewkare working capital demands by government,dMis the change in the money stock (currency), dB is the change in the market value of the bond stock, ander.dFD*

gis the change in the stock of foreign currency denominated public sector debt.

Each transactor in the model demands working capital. The demand for working capital is generally assumed to be a function of their level of nominal transactions and the prevailing loan/ deposit rate. For example, households demand for working capital is a function of their nominal consumption expenditure and the operating surplus of unincorporated business. These demands for “transactions’ balances” then determine the sector’s demand for currency. In effect our LM curve is vertical. “Working capital” is financed together with other liabilities for debtor sectors and is drawn down from creditors’ stocks of net assets.

For creditor sectors (e.g., households), asset portfolio selection can be represented as:34

3

Dj

Bj

Ej

er · FA* j

4

5 Q·

3

id ib ie iw1 •

ere

4

1

3

ud

ub

ue 12 um2 ub2 ue

4

· (Aj[t]2wkj[t]) ,

whereD are interest bearing deposits,B is bonds,E is equities, FA* are net foreign assets in foreign currency units:idis the bank deposit rate;ibis the market return on bonds; ieis the return on equity;iw is the exogenous world nominal interest rate, anderis the nominal effective exchange rate. The matrix Q satisfies the usual sign and adding-up restrictions.

For debtor sectors, the corresponding portfolio selection equa-tions are:

3

Lj

Ej

er · FD* j

4

5 P·

3

il ie iw1er4• 1

3

vl

ve 12 vl2 ve

4

· (Aj[t]1wkj[t])

where L are loans and il is the loan rate. Again, the usual crossequation portfolio restrictions are satisfied by the parameters of this system of demand equations. The ratio of interest-bearing

34_{Portfolio selection in the model is, in fact, hierarchical and features many more assets}

and liabilities than are identified in this simple description of the model. Although our portfolio selection equations satisfy the restrictions identified in the text, they also permit model users to place upper bounds on the share of wealth/debt held in any single asset/ liability.

deposits to loans is (under flexible interest rate regimes) an exoge-nous parameter of the model, and the loan rate and deposit rate are linked through a fixed markup relationship.

End-of-period stock equilibrium requires money market, bond market, equity market, and loan-deposit market clearing, and that foreign net indebtedness equals the sum of the net claims on individual transactors. Hence,

M5Mh1Mc1Mpc

B5Bh1Bc1Bpc

Eh1Ef5Ec1Epc

D·h 5Lc1Lpc

Af5er· (FD*g1FDc*1FD*pc2FA*h)1Ef,

wherehis the fixed ratio of loans to deposits andEf, is exogenous (long-run) equity capital inflows.

Material Balances

In each sector we require that total output demanded equals total output supplied in each period:

xj5

o

i

jdji1cdj1idj1exj1gj|jÓLOP.35

In the market for Public Services, government recurrent expendi-ture on goods adjusts to satisfy material balances for given supply and in the Dwellings “market,” the demand for housing is automat-ically equated with imputed rent.

Labor Market Closure

In all labor markets except the market for unskilled nonagricul-tural labor, wages (wn) adjust to equate the demand and supply of labor in each period. In the nonland-based, unskilled-labor market the rate of wage change is related to expected consumption goods price inflation, excess demand for labor and a trend captur-ing productivity growth. Hence, in those markets that clear we have:

lfn· (12nruen)5

o

i

nnni,

where lf is the labor force and nrue is the (exogenous) natural rate of unemployment. In the nonland-based market for unskilled labor:

dlnwn5 u01 u1dlncpie2 u2ln (1

2uen)

(12nruen),

whereueis the actual unemployment rate, andcpiis the consumer price index. We model net migration from land-based to nonland-based labor markets as:

dlnPOP L t

POPt

5φ0·

1

ln

wL t

wt

2 v*

2

1φ1·

1

ln 12ueL

t 12uet

2rue*

2

1φ2·

1

ln

POPL t21

POPt21

2 §02 §1· ln

wL t21

wt21

2 §2· ln 12ueL

t21 12uet21

2

,

from which net migration can be calculated as: nmgt5exp

3

ln

POPL t21

POPt21

1dlnPOPLt

POPt

4

·POPt21· (11ng)

2POPL

t21· (11ngL)5POPtL2POPLt21· (11ngL),

wherePOPis the end of the period population stock,nmgis net migration from nonland-based to land-based activities (L), and a bar above a variable represents its mean value in nonland-based markets. The terms v* and rue* denote, respectively, the wage gap and the log ratio of employment rates in equilibrium. The termngis the natural rate of growth of the labor force. Our net migration equation is an error correction formulation of a model in which we postulate the equilibrium distribution of the population stock is a function of relative market wages and unemployment rates. The labor force in each market is updated through:

lfn[t11]5lfn[t]· (11ngn[t])1εn·nmgn[t]|

o

n

nmgn[t]50,

o

nÓL

εn51,εL51. Other Intrinsic Dynamics

The capital stock in each activity is depreciated in each period and augmented by lagged investment:

ki[t]5ki[t21]· (12 d)1(iai[t21]1iagi[t21]1 j·iapci[t21]).

Private sector, nonhuman wealth is defined in nominal units as the value of the capital stock owned by the private sector, plus outside money plus the private sector housing stock less net foreign

indebtedness (net of government debt). wlt5

o

i

pki·ki[t]·ci1outside money2housing stock

2(Af[t]2er·FD*g 2er·FD*pc).

Note that in the current specification of the model neither public sector debt nor assets enter into the calculation of private wealth. In each period, technological progress occurs at the value-added level of each sector’s production function and is labor augmenting (Harrod neutral), but with variable rates across activities. Techno-logical progress is assumed to be “high” in manufacturing (0.08); moderate in services (0.04), and low in primary sectors (0.02), yielding an aggregate rate of about 0.05. Together with labor force growth of 3 percent pa., this generates trend growth of just over 8 percent pa.

Expectations of the change in consumer price inflation and of the nominal exchange rate are adaptive and are of the usual form:

pt112 pt5 lp·

1

d cpi cpit21

2 pt

2

•

ere t112

•

ere

t 5 ler·

1

d er ert21

2 •

ere t

2

,

wherepis expected inflation and •

ere_{is the expected depreciation} of the exchange rate.

Constraints

In addition to our market clearing conditions, we require that, in each period, the Forestry sector activity outputs and fixed input stocks satisfy our exogenous restrictions on their values:

xaF5xaF

kF5kF.

To satisfy the output constraint, Forestry exports adjust residu-ally (see above), and investment in Forestry accommodates the constraint on fixed inputs. Although negative investment does lead to the value of the capital stock (and wealth) being written down, it does not, given the irreversibility of capital investment decisions, release resources for use elsewhere in the economy.

Our consumption constraint is accommodated through fixing the value ofc/cpiat the appropriate baseline level for each year, inverting the aggregate consumption to find the necessarily level

of household disposable income, and then by calculating the level of unrequited transfers from the rest of the world (a component oftransfers) that will yield this income, given all other elements of income and adjusting for taxes.

Finally, target ratios of nominal public sector debt to nominal GDP are satisfied through endogenous adjustment factors that operate on “target” income tax rates and public corporations’ “target” capital expenditures.

Data and Baseline Calibration

The initial database for the model was a set of social accounts, flow of funds, and asset stock data collated for the period 1983–84. These data were subsequently updated to 1990 and the model has be recalibrated on these more recent data and reproduces these data as its 1990 solution. These updated data are consistent with available Malaysian national accounts information for 1990.

Model Solution

The model is solved using an implementation of the Levenberg-Marquard algorithm that is described in Appendix 4 of Harrigan et al. (1991). Essentially, through a recursive ordering of the equa-tions, variable elimination is achieved and the model is then solved as a constrained nonlinear, least-squares optimization problem. Overall, the model has about 2,500 endogenous variables, and although all of the system’s equations must be retained for the calculation of the numerical derivative information used by the algorithm, the rank of the associated Hessian matrix can be re-duced to just over 50 for an appropriate ordering of equations.

REFERENCES

Armington, P. (1969) A Theory of Demand for Products Distinguished by Place of Produc-tion.IMF Staff Papers16:159–178.

Braga, C.A.P. (1992) Tropical Forests and Trade Policy: The Case of Indonesia and Brazil. InInternational Trade and the Environment(P. Low, Ed.). World Bank Discussion Paper 159, Washington, DC, pp. 173–194.

Bourguignon, F., Branson, W.H., and De Melo, J. (1992) Adjustment and Income Distribu-tion: A Micro-Macro Model for Counterfactual Analysis.Journal of Development Economics31.

Burns, D. (1986)Runaway and Treadmill Deforestation, vol. 2.IUCN/IIED Tropical For-estry Policy Paper, London, UK.

Cline, W.R. (1991) Scientific Basis for the Greenhouse Effect.Economic Journal101:904– 919.

Corden, M., and Neary, J.P. (1982) Booming Sector and De-Industrialisation in a Small Open Economy.Economic Journal92:825–849.

Dasgupta, P., and Ma¨ler, K.G. (1991)The Environment and Emerging Development Issues. Proceedings of the World Bank Annual Conference on Development Economics, 1991, Supplement to the World Bank Economic Review and the World Bank Research Observer.

Demery D., Harrigan, F.J., and McGregor, P. (1992)M4_{: A Micro-Macro Simulation} Framework for Malaysia. Report to the Economic Planning Unit of the Prime Minister’s Department of Malaysia, Kuala Lumpur.

FAO (1993)Forest Resources Assessment 1990. Rome: FAO, United Nations.

Gillis M. (1988)Malaysia: Public Policies and the Tropical Forest, In Public Policies and the Misuse of Forest Resources(R. Repetto and M. Gillis, Eds.). Cambridge: Cambridge University Press.

Government Press. (1990)The Sixth Malaysian Plan.Malaysia: Kuala Lumpur. Government Press. (1994)Economic Report.Malaysia: (Treasury) Kuala Lumpur. Harrigan, F., McGregor, P., Swales, J.K., Perman, R., and Yin, Y.P. (1991) Amos: A

Micro-Macro Model of Scotland.Economic Modelling8:424–479. IBRD. (1993)East Asian Miracle. New York: Oxford University Press.

Muscatelli, A., Srinavasan, T.G., and Vines, D. (1994) The Empirical Modelling of NIE Exports: An Evaluation of Different Approaches. The Journal of Development Studies30:279–302.

Muscatelli, A., Srinavasan, T.G., and Vines, D. (1992) Supply and Demand Factors in the Determination of NIE Exports: A Simulataneous Error-correction Model for Hong Kong.Economic Journal, 102:1467–1477.

Nordhaus, W.D. (1991) To Slow or Not to Slow.Economic Journal101:920–937. Parthama, I.P., Vincent, J.R. (1992) United States Demand for Indonesian Plywood.

Bulletin of Indonesian Economic Studies28:101–112.

Reidel, J. (1988) The Demand for LDC Exports of Manufactures: Estimates for Hong Kong.Economic Journal98:138–148.

Repetto R., and M. Gillis, Eds. (1988)Public Policies and the Misuse of Forest Resources. Cambridge: Cambridge University Press.

Repetto R. (1994) Trade and Sustainable Development. InCritical Issues in Asian Develop-ment(M.G. Quibria, Ed.). Hong Kong: Oxford University Press.

Robinson, S. (1991) Macroeconomics, financial variables and CGEs.World Development

19:1509–1525.

Vincent, J.R. (1992) The Tropical Timber Trade and Sustainable Development.Science