Document(s) 66 de 95

ABSTRACT

This paper addresses the effects of economic, demographic andinstitutional factors on land allocation between forestry and other uses.A panel data set from Hainan Island in China and a generalized least squaresestimation method, allowing individual effects for counties, are applied.The results indicate that higher timber prices have led to accelerationin rainforest exploitation, but encouraged forest investment in plantationforests. Population growth is the driving force behind the loss of naturalforests, but is positively related to plantation forests. De-collectivizationseems to have promoted plantation forests, but have not saved therainforest. A higher share of forestry land owned by state-owned enterprisesalso fosters afforestation on wasteland, but seems to lead to faster exploitationof natural forest, at least initially. The uncertainty that existed inthe early period of economic reform quickened the pace of resource extractionand deterred investment.

Key words: Forestland tenure, economic transition,deforestation, reforestation, panel data
 
 

1.0 INTRODUCTION

Recently, the causal factors behind the processes that lead to conversionof forests into farmland, tropical cropland and other non-forest landshave attracted considerable attention in the literature on forest landchanges in less developed countries (LDCs). Forest loss has many seriousnegative environmental and socio-economic impacts in LDCs, especially inthe long run. Research has been able to shed some light on the factorsbehind the process. However, land conversion that leads to forest landexpansion has received only little attention in the literature on landuse economics. Yet, forestland expansion, mainly through an increase inplantation forests and active natural forest management, can reverse many,even if not all, of the adverse consequences of loss of natural forests.

Many earlier studies that have addressed the problems related to forestland changes in tropical countries have analyzed the effects of demographicand economic conditions either using cross-country data (e.g., Capistrano1994, Palo 1994, Rudel 1998) or sub-country level data (e.g., Kummer andSham 1994, Panayoutou and Sungsuwan 1994). However, in the new institutionaleconomics it has been argued that in addition to economic factors, suchas the relative prices, institutional factors also affect economic development(e.g., North 1990). In the context of land use and forestry, the impactsof institutions have been examined by Bromley (1989 and 1991), Wallaceand Newman (1986), Mendelsohn (1994), Deacon (1994 and 1999), Besley (1995),Zhang (1996), Zhang and Pearse (1996), Laarman (1997), Yin and Newman (1997),Place and Otsuka (1998).

In this study, we examine the role of economic and institutional factorsin the development of both plantations and natural forests in the Provinceof Hainan. One of the reasons why development of plantation forests hasbeen rarely studied is that plantation forests are still relatively smallin the tropics, and successful plantation programs are even more rare.With its established programs aimed at increasing forest plantations, thecase of Hainan Island provides a good opportunity to shed some light onthe underlying causes and long-term development of forestry in LDCs. Inparticular, we are able to address some of the factors that may reversethe decline in total forestland area.

A major innovation of this paper is to distinguish between natural forestand forest plantation, and contrast how various factors or same factorsaffect the two differently. Forest transition is the consequence of theshift from natural forest exploitation to harvesting managed forests. Ofparticular interest in Hainan, as in other LDCs, is the effectiveness ofstructural adjustment policies aimed at adjusting government control, clarifyingforestland property rights and improving the possibilities for their enforcement.We also consider the roles of relative prices and transition policiesin determining forestland changes in Hainan. Our relatively long paneldata allow us to use regional dummy variables that are able to captureenvironmental and other directly unobservable differences across counties.

We start by describing the forest sector and economic reforms in Hainanprovince. The third section introduces the theoretical model for land allocationamong different uses leading to an estimable land allocation function.The following section introduces the panel data. Estimation results, forboth natural forests and plantations forests, are presented after this,followed by conclusions.
 
 

2.0 FORESTRY AND ECONOMIC REFORM IN HAINAN

During the past 20 years the most important ways to re-form land tenurein Hainan have been the following: i) the control of public forestry landnominally owned by the state is given to the state-owned enterprises; ii)the management authority for state-owned forestry land is transferred throughde-centralization from national to provincial level, and then from provincialto county level, or from state to community level; iii) forestry land isdistributed through de-collectivization to households, e.g., on the basisof the number of persons in the household; alternatively, parts (or all)of the management authority are transferred to contractors; and iv) propertyrights are redefined by organizing various kinds of joint forest managementventures that combine different inputs. Below, we briefly describe developmentof forests in Hainan before discussing the above land tenure reforms inmore detail.

Hainan Island, which covers a total area of 3.4 million ha, is locatedin the South China Sea and separated from the Chinese mainland by the QiongzhouStraits. It is almost as large as Taiwan Island. Hainan has a typical tropicalmonsoon climate. In 1988 Hainan was upgraded to the provincial level andgiven the status of Special Economic Zone. Since then, the economic growthrate, averaging 20% annually from 1988 to 1995, has been the fastest atprovincial level in China. The population is also growing dramatically,from 5 million in the mid-1980s to 8 million at present (the actual populationcould be over 10 million).

Originally, Hainan was covered mostly by rainforests (Situ 1992), butthe effect of long-term human activities has caused serious deforestationand degradation of forestland. As in the rest of the tropical world, deforestationin Hainan was also directly caused by logging, shifting cultivation, agriculturalland encroachment and residential and industrial expansion. In addition,an even greater share of the land was abandoned without management, leavingopen forest, shrubs and long-term fallow.

It is widely agreed that the forest cover was about 30% at the beginningof the 1950s and the remaining forests were mostly located in the centerof mountain ranges (HAZC 1980). Since 1950s, the rainforest had sufferedfurther from logging due to growing demand from within and outside theisland. As harvesting was carried out like mining and there was insufficientsilvicultural investment, Hainan changed from being a net timber exporterin the 1950s-1970s to an importer in the 1980s. Tropical crops were introducedin Hainan at the beginning of this century, but it was not until the 1950sthat they started to expand dramatically as a result of increasing domesticdemand for natural rubber materials (trade was blocked by the western worldat that time). Over the past decades, agricultural land has not expandedsignificantly, while shifting cultivation, which has been widely practicedby local minorities, has destroyed large tracts of rainforest. It is estimatedthat 2,700 ha of forests were destroyed annually in the 1970s-1980s (Bao1991). After three decades of destruction, the forests shrank to theirminimum level, 15% of total land area, in the late 1970s (HAZC 1980).

During the last two decades the decreasing trend in forestland areahas been reversed and this area in Hainan is presently expanding (see Figures1-3). This provides empirical support for the 'forest transition' hypothesisclaiming that the declining forestland area will eventually begin to increaseas the country develops as suggested in Mather (1990), Hyde et al. (1996),Rudel (1998) and Zhang (2000) among others. Since 1990, income from wood-chipexports has become an important source of foreign currency earnings andan important part of rural economy. Physically, forest expansion has beendirectly caused by massive plantation of forests both for producing rawmaterial for wood-chip production and for environmental functions suchas wind-break and water conservation. Also, the rehabilitation of degradedland by closing access to it and a significant reduction in logging ofnatural forest through stricter regulation have had a positive effect onmaintaining rainforest.

Land tenure and administration in Hainan does not differ noticeablyfrom that in the rest of China. The only special feature is that state-ownedforest is mixed with collective forest and that forests play a specialrole in the island's environment and eco-tourism. Under China's LandReform implemented in 1950-1956, all land owned by landlords was confiscatedand most of it distributed to local farmers. However, remote areas, particularlyin the central island, i.e. the rainforests, were reallocated to the state-ownedlogging firms. During the 1950s and 1960s, about 350,000 ha of forestryland was allocated to state logging firms, of which 80% went to the 11largest ones (HAZC 1980).

Figure 1. Forest coverage/inventory (1950s-1990s)

Figure 2. Timber production (1950s-1990s)

In socialistic transformation that started in 1957 the land wastransferred from individuals back to the collectives in the first phaseof collectivization, and then to the People's Communes inthe second phase. The People's Communes continued until the Mao'sdeath in 1976. There were also a couple of episodes of de-collectivization,e.g., after the collapse of the Great-Leap-Forward in the late 1950s(Walker 1965). Compared to agriculture, the forestry sector was much morecollectivised and centralized.

The economic reform started in the late 1970s. The most significantreform in agriculture has been the introduction of the Household ResponsibilitySystem (HRS), which actually represents a kind of privatization of landuse rights. In the forest sector, the reform in general began a littlelater and is far less intensive and extensive than in agriculture (Zhanget al. 1999). The institutional reforms in Hainan forestry sector duringthe last 20 years are summarized in the following paragraphs.

Joint plantation of fast-growing species of trees

When the eucalyptus wood-chip market was expanding and a market-orientedeconomy was gradually emerging in China in the late 1970s, the Ministryof Forestry started planning to establish 130,000 ha of fast-growing treesin Hainan. A project, referred to as "2 million mu (15 mu= 1 ha) joint plantation of fast-growing and high-yield species of trees",was formally launched in 1982. The joint partners were the Ministry ofForestry, which provided the capital, the local collectives, which providedthe land, and the local people, who supplied labor. Encouraged by thisinitiative, several other joint plantation projects followed. A few banks,including the World Bank, provided substantial loans for these projects.

In general, these projects were successful. A total of 130,000 ha ofeucalyptus and some other fast growing species of trees was planted between1982 and 1995, accounting for more than one third of plantation forestsduring this period, and contributed to a 4% increase in total forest cover.The General Fast-growing Plantation Forest Company, established in 1984,has been in charge of administration, technology assistance, and the wood-chipexport monopoly.

Closed access management of degraded forestry land

From the 1950s to the 1980s, the category of degraded forestry land,including, e.g., open forest, shrubs, and long-term fallow, was largerin area than the other categories of forested land. Degraded forestryland accounted for one quarter of the total land area in the mid-1970s(Figure 3), resulting to a great extent from repeated damage caused byfuel-wood gathering and shifting cultivation (HAZC 1980).

Along with population growth, economic development and advances in silviculturaltechniques, some good location and productive bare land has been afforested.Closedaccess management, which has been mainly applied to degradedforestry land, has greatly accelerated its rehabilitation. Exclusivenessis a pre-requisite for silvicultural investment (Zhang and Pearse 1996).In Hainan, contracts between the local collectives, who own the land andtrees, and households, who receive an annual guardian fee or share of thereturn from the final harvests have been used to reduce the costs of exclusion.Thus, this arrangement individualizes some of the benefits from enforcementof the forestland property rights at low costs, serving as a kind of privatization.During the 1980s, it was estimated that more than 100,000 ha of degradedland had been rehabilitated in Hainan by this approach (Zeng 1994), andalmost the same area during the 1990s. Therefore, 6-8% of the total territoryhas been reforested by this method.

Restrictions on timber production from tropical rainforests

The topography of Hainan is characterized by mountainous ranges in thecenter that gradually fall away towards the coast. The tropical rainforestsare mostly found in the center and are presently managed by the state-ownedforestry bureaus. After decades of exploitation, access is becoming moredifficult and the quality of the forest is declining.

The rainforest has attracted greater attention from the public and governmentsince it is widely believed that overall environmental deterioration, suchas soil erosion, shortage of water sources and loss of biodiversity, isdue to the loss of the rainforest in Hainan. Therefore, more restrictionshave gradually been imposed on rainforests. First, only selective cuttingwas allowed in the late 1970s. Then, a quota of allowable fellings wasset in 1984 and has been gradually reduced since then. Finally, in principle,harvesting of all rainforest has been banned since 1993 and 1994.

Since the state-owned forestry bureaus depended heavily on rainforestlogging, the economy suffered from the restriction. Financially, both thenational and provincial levels compensated for some of the loss; banks,either encouraged or, forced by governments, provided low interest loansto finance the shift to other fields of business, such as silviculture,furniture production, eco-tourism, tropical crop plantations, constructionof hydropower stations, mining, etc. After a decade's efforts, it seemsthat the forests in most of the forestry bureaus are recovering, even thoughthe local economies are still facing some difficulties.
 
 

3.0 THEORETICAL MODEL OF LAND ALLOCATION

The literature on spatial land allocation theory postulates that landis divided between alternative uses according to the highest land rentobtainable, given land's special characteristics with regards to locationand physical conditions (see. e.g. von Thunen 1875, Hyde 1980, Chomitzand Gray 1996). In China, land is publicly owned. Therefore, land marketsdo not in effect exist and land prices reflecting land rent of a particularland quality class in alternative uses are not available. However, it canbe assumed that individuals or (state) enterprises that have been giventhe control over land will allocate the land to different uses accordingto profitability in the particular use category.

The profitability of land use depends on relative prices of productsin different land use categories. Because we do not have data on pricesof land with different characteristics, we abstract from land quality classes.Assume that the landowner (an individual, a household or other administrativeunit with land tenure) seeks to maximize the total land rent from differentland uses on homogenous land under his control by allocating the totalland area optimally between n different uses. Distance to markets and otherfactors affecting the return on land, enter the model in our specificationindirectly through the production function properties. In particular, productiony_iof the ith product is a strictly concave function of land area A_iallocated to ith land use category and of composite input E_iused. The landowner's problem is to maximize the profit, constrained bythe total land area under his control:

Where p_i's (i = 1, n) are scalars for the product prices, E_i are the aggregate inputs used to produce the outputs y_i, and A_i's are the sizes of land areas allocated to produce the respective products and c is the unit cost of composite input. The production functions are such that and , i.e., the marginal products of land and composite input are positive but decreasing in each land use category. Setting up the Lagrangean, we obtain:

(2) R (p_i, A) = [ p_i y_i(A_i, E_i) - cE_i]+l (A-A_i)

The Kuhn-Tucker conditions are:

The terms arethe marginal income from additional land to be used. Thus they representshadow land rents. According to (3) land is optimally allocated so thatthese rents for a homogenous land are equal for each land use category.For an interior solution we obtain the following the equilibrium conditionfrom equation (3):

According to (4) the ratio of the marginal products of land under land-usej and i equals the ratio of the net prices of products i and j. The conditiongives the marginal rate of substitution between two different land uses.For example, if productivity of a certain land use category increases,more land will be allocated to this particular use.

The optimal land areas in different uses implied by first-order conditionsare functions of the product prices, the composite input price and theavailable total land area, and they can be written as

where p is the vector of product prices (Chambers 1988, p. 281about restricted production optimization).

The optimal land allocation functions, the 's,are nondecreasing in p_j (i = j). It can be shown thatthe own-price effects are positive. For the purposes of the empirical modeling,these results give us the expected positive sign for the effect of a changein timber price on the amount of land allocated for forest, and an expectednegative sign for the cross-price effects of the various land use classes.The effect of c, the composite input price, on dependson the substitution effect between land and the aggregate input, plus theoutput effect. A priori the total effect of c is indeterminate insign.

As we argue, besides markets and prices, institutional factors playa key role in rural land allocation in Hainan. Therefore, we include inthe forest land allocation model one variable describing the economic developmentand one for population, as well as two variables describing the land propertyrights policies in China. The first two variables are gross output value(GOP) and population density (PD) in each of the Hainan counties.²

The land reform variables are the percentage share of land under thehousehold-responsibility system (H), and the percentage of forestry landunder state ownership (SF). These variables thus measure the effects ofland-tenure shift and decentralization of forestland management in Hainan.Accounting for the market-driven and the institutional factors, the implicitmodel for forestland allocation within the counties in Hainan can be writtenas:

(6) Forestland Allocation = F (p, c, A; GOP, PD; SF, H)

where p is a vector of output prices of rural crops includingtimber prices, c is the cost index for rural inputs, A is the total landarea in the county, and the rest of the symbols stand for the above mentionedinstitutional variables. In Hainan, the institutional changes from socialisttransformation to current economic reforms have been less endogenous withthe socio-economic development, but greatly determined by political ideology.Therefore, they can be treated as exogenous variables (Besley 1995).

3.1 Data

Before proceeding, we divide land use in Hainan into 4 broad classes(see Figure 3 and Table 1). In Table 1 total forest area has been dividedinto three categories: actively managed forests, degraded but afforestableland, and natural forests. The first and the third categories are classifiedas being part of forestland in official Chinese statistics, whileall three combined form the forestry land category. Therefore, inTable 1, category 4 represents the forestry land class. In our study wewill examine categories 4.1 and 4.3.

Figure 3. Land use/cover classifications and their changes (1950sto 1990s)

Table 1. Land classification

Categories	Description and notes
A1: Built-up,roads & unproductive land	Comprises mainlyresidential and industrial areas, roads, mining areas. This category alsoincludes technologically unforestable land such as water bodies and excessivelydegraded land which makes up 8% of the total land.
A2: Cultivationland	Subsistenceagricultural land (rice, vegetable, non-tree fruits, etc.)
A3:Tropical cropland	Tropicalcrops (rubber trees and other crops, tropical tree-fruits, etc.)
A4.1: Activelymanaged forest	Comprises mainlyplantation forest but includes some actively managed secondary naturalforest.
A4.2: Degradedbut afforestable land	Low biomasswoody land such as open forest, shrubs and long term fallow. But all aretechnologically afforestable.
A4.3: Non-activelymanaged forests	Frontier forestsor natural forest with very limited human disturbance.

Counties in Hainan are economic and administrative units, but they alsoreflect the natural, geological, economic and even social integration ofeach unit. Therefore, the cross-section observations in the study are basedon counties.

Hainan has a total of 19 counties and one state reclamation bureau.Due to frequent changes that occurred in the administrative system, wecombinethe neighboring counties of Haikou City and Qiangshan County; Tongza, Qingzhongand Baotin; Chenmai and Linguo; and Baisa and Changjiang. This ensuresthat the 13 cross-section units thus obtained remain the same throughoutthe period covered. Since data on the state reclamation bureau are verysparse, the bureau is therefore excluded from this study. Our sample covers17 points in time in the period from 1957 to 1994 and 13 cross sectionunits, i.e. 221 observations.

We use gross output value per capita (GOP) instead of GDP because GDPwas not used in the earlier statistical reports, and also because GOP isa good proxy for GDP. Nominal GOP has been deflated by the retail priceindex (base year = 1957). Prices are obtained from local statistics, mainlyHainan Statistical Yearbooks (HSB 1957-1995). Timber price used for plantationforests is the price for Eucalyptus while the timber price for rainforestsis an average hardwood price. All prices are at provincial leveland deflated using the general retail price index. The data on forest resourceswere combined with the forestry statistics of forestry bureaus that arebased on the Institute of Forest Resources Investigation (HFB 1957-1995).Data on the natural variables were obtained from HAZC (1980).

The percentage share of forestry land managed by state logging firmsand farms (SF) was obtained from the Hainan Forestry Bureau, while H, thepercentage share of land under the household responsibility system,was collected together with information provided by forestry bureaus atcounty level.

3.2 Econometric specification and methods

The econometric specification for both managed forests and rainforestsis based on equation (6). To avoid omitted variables bias, due tounobserved county-specific institutional and environmental factors, a paneldata analysis with individual constant terms for counties, was used. Fixed-effectpanel data estimation is appropriate when cross-section units are not sampledfrom a large population, and the differences between units can be viewedas parametric shifts of the regression function (Greene 1997, p. 623).This clearly is the case in the present study, because our sample includesall the counties in Hainan. Therefore, the estimated parameters cannotbe used to predict behavior outside the present sample, although our methodshould be applicable in other studies of land allocation.

As macroeconomic, as well as geographic and socio-economic factors,are likely to affect land allocation in different counties in Hainan, weallow contemporaneous correlation of cross-section units (Greene 1997,p. 658, Hsiao 1986). Lagged values of exogenous variables did not improvethe statistical performance of the two estimated equations. Therefore,to take into account the relatively strong autocorrelation in the data,the Cochrane-Orcutt transformation is used and the model estimated usingthe iterative Generalized Least Squares method (Kmenta 1986 pp. 622-625,Greene 1997, pp. 651-669). The econometric specification is

(7) FC_it=a ₁+a_iD_i+b₁PD_it+b₂GOP_it+b_(2+n)Pⁿ_it+b₆SF_it+b₇H_it+b₈DU_t+b₉H_itDU_t+e_it,

where subscripts i and t indicate the counties and years, respectively;D_i are the regional dummy variables; FC denotes the forest cover(%); PD is population density (in persons per ha), and GOP is deflatedper capita output value (Yuan per capita); Pⁿ (n=1,2,3) arethe real price indexes of agricultural products, tropical crops and timber,respectively, i.e., they are divided by the index for rural industrialmaterials, following Yin and Newman (1997)³; SF denotes theratio of land owned by state forestry firms to total forestry land andH is the ratio of the household responsibility forestry land to total forestryland. All variables except SF and H are in natural logarithms.

The period we examined covers the years from 1957 to 1994. The years1980 to 1987 were characterized by uncertainty concerning the directionof economic reforms, which were in an experimental phase. Farmers and local,even provincial, government officials did not know whether and how thereforms would be carried through. Forests suffered greatly from this uncertaintyacross the country. Therefore, a dummy variable, DU, is used to allow boththe intercept, and the slope of the de-collectivization (H) to differ fromthe rest of the period studied.

To examine the difference between managed (mainly plantations) and naturalforests (mainly rainforests), a behavioral model is estimated for bothcategories.⁴ As stated above, hardwood price is used as thetimber price for the rainforest; the timber price for plantation forestequation is the price of timber produced from plantation, mainly Eucalyptus.Although the same economic principles apply to managed and rainforests,the development of managed and rainforest areas may respond differentlyto relative prices and the institutional variables because investment inplantation forestry resembles investment in other sectors, such as agriculture,while harvesting of natural forests is a mining-like activity.

In the case of rainforest, only 8 of the 13 cross-section units wereof significant size during the period studied. The harvesting of rainforesthas been regulated: it is either that only a selective cut has been allowedsince the late 1970s or harvesting has been prohibited altogether sincethe mid-1980s (in other words they have become "managed forest"). Therefore,we exclude the 5 cross-section units where rainforests make up less than2% of the total area of the county (in the late 1970s) and the period after1985 in the regression model explaining the area of rainforests.⁵The total number of observations is 72 (8 units for 9 points in time).The estimated function includes the same explanatory variables asequation (7).
 
 

4.0 ESTIMATED RESULTS

The estimated results for managed forest and rainforest obtained withequation (7) are given in Table 2. Standard errors clearly decreased whencross-sectional heteroscedasticy and contemporaneous correlation were allowed,indicating an increase in efficiency when using GLS. However, some of thevariation between cross-sectional units is not explained by the variablesincluded, as indicated by the statistically significant regional dummies.Inorder to take care of the directly unobservable regional differences, suchas climate, topography and history, etc., a common intercept and 12 regionaldummies were retained.

Table 2. GLS estimation results for managed and rainforest cover in Hainan, 1957-1994* (The t-values are in parentheses below these estimates)

	Variable description	Managed forest cover (%)	Rainforest cover (%)
a1	Constant**	-3.74 (7.03)	2.98 (3.40)
Pa	Agriculturalproducts price	0.11 (1.16)	0.33 (6.55)
Pc	Tropical crop/Product price	-0.40 (4.22)	1.29 (7.68)
Pt	Timber price	0.61 (5.81)	-1.52 (4.80)
PD	Populationdensity	2.21 (42.57)	-0.61 (21.51)
GOP	Per capitagross output value	0.59 (22.48)	-0.31 (18.33)
SF	Percentageof SF land to total (%)	0.42 (7.17)	-0.77 (7.80)
H	Ratio of HRSto total	0.79 (8.56)	-0.54 (16.06)
DU	Dummy for uncertainty	-0.09 (2.73)	-0.76 (0.88)
H·DU	Joint effectof HRS and uncertainty	-1.55 (15.49)	0.66 (17.07)
r***	Autocorrelationcoeff.	0.64	0.53
Log-L	Log-likelihood	134.7	144.6

Notes:
* Data for managed forests is from 1957 to 1994 across13 units; data for rainforest is from 1957 to 1985 across 8 units.
** The regional dummy variables are excluded in the Table.
*** A single autocorrelation coefficient is adopted becausewe found that different autocorrelation coefficients did not change theestimated results or standard errors markedly.

As expected, higher timber prices promote forestry investment in managedforest as measured by plantation forest cover. However, the rainforestland area decreases as timber price increases, implying that the increasinghardwood price has, at least in the past, increased mining-typeharvesting of rainforests. The agricultural product price has a positiveimpact on both managed and natural forest cover. This runs counter to thetheoretical hypothesis, and suggests that rising agricultural product pricesdo not lead to encroachment on forestland in Hainan. In a developing subsistenceeconomy, this result may be due to the fact that farmers can meet theirincome needs with less harvest income when agricultural prices increase.Also there does not seem to be any serious conflict between agricultureand forestry. A similar result was obtained for mainland China by Yin andNewman (1997). Plantation forests respond negatively to the tropical cropproduct price, implying that tropical crops compete for land with plantation.However, the rainforest and tropical crops may even be complementary asindicated by the positive coefficient of tropical crop price in the rainforestequation.

Like many other studies, our estimated results indicate that populationhas a negative effect on rainforest coverage. This suggests that growingpopulation generally causes some rainforest to be converted intoagricultural, industrial and residential land, and probably into wastelandafter logging. However, the managed forest area in Hainan has increasedwith population. Thus, rising population has speeded up the exploitationof natural forest, but may have promoted plantation in Hainan.

The effect of economic development (or welfare), measured by per capitaoutput value, also has a positive effect on managed forest cover. One explanationfor this may be that economic development reduces transportation costsand costs of protecting property rights as infrastructure and law implementationare improved. The economic development may also have demand side effects.Strengthening demand for timber is reflected in timber prices. However,a wealthier society may also appreciate the in situ benefits of forestscausing a positive correlation between plantation forest and GOP. For examplein Hainan, about 20% of plantation forest is intended for environmentalpurposes. The positive effect of GOP on managed forests does not contradictthe negative sign of GOP in the rainforest equation. Natural forests areexamined over the period 1957-1985. Since then harvesting in rainforesthas been fully regulated, indicating increased environmental concerns anda change in preferences towards protection of natural forests. The expansionof managed forest accelerated in the 1990s in line with the economic growth.

The share of forestry land controlled by state-owned enterprises seemsto have a positive effect on plantation forest cover, but the effect issignificantly negative for rainforest before 1985. This can be explainedby the fact that state-owned forestry bureaus (or farms) were set up fortwo purposes: to afforest public wasteland and to exploit public naturalforest. Our results also show that de-collectivization (H) promoted theloss of natural forest in the early part of the explored period (the negativesign of H in the second column of Table 2), but may have worked the otherdirection in the 1980s (the positive sign of H*DU in the second columnof Table 2).

In summary, the disaggregation of total forestland area into managedforests and natural forests is justified when explaining the developmentforestland area. The results indicate different responses of plantationand rainforest to evolving timber prices, population growth, economic development,and institutional changes.

Cross-sectional differences

The regional constant terms represent differences between counties thatare omitted from the model because the data necessary for their inclusionare not available (Hsiao 1986). Even if it is not possible to explain allregional variation with the available data, the estimated regional constantterms can be used to study possible reasons for this variation. In thepresent case, it can be assumed that regional differences are at leastpartly due to natural and geological factors (e.g., Hoshino 1996). As wehave only 13 cross-section units for managed forest (8 units for rainforests),the results must be considered with caution. However, in order to tentativelytest whether geological factors explain the regional differences, we regressregion-specific constants, a *_i (a*₁=a₁;a*_i=a₁+a_i,for i ¹ 1), on the share of mountain land(%) above 500 m sea level, ML, the share of hilly land (with an altitudebetween 250-500 m above sea level) plus the land with a slope of 12 degreebut an altitude of less than 250 m above sea level, SL, and the annualrainfall, RF. The estimated model becomes:

(8) a *_i =a+b ₁ML_i+b₂SL_i+b₃RF_i+e_i

Table 3. The OLS estimates for intercepts equations

	Variable description	Intercepts (in Managed forest cover)	Intercepts (in Rainforests land cover)
a	Constant	-7.60 (3.45)	3.74 (0.70)
ML (%)	Mountain area	0.05 (0.62)	0.46 (1.53)
SL (%)	Hilly and slopeland	1.92 (4.93)	-0.51 (0.31)
RF (100 mm)	Rainfall	-0.02 (0.04)	-0.33 (0.30)
R2	AdjustedR square	0.80	0.27

Note: The values in parentheses.

According to the results reported in Table 3, the regional constantterms are statistically related to the natural variables. The dummies formanaged forest are very strongly related to the share of hilly and slopingland, while the dummies for rainforest seem to be related most to mountainland.⁶ This is intuitively interpretable. Flat and low landis best suited to agriculture and tropical crop plantations, the middleand high mountains are likely to be rainforest because of their difficultaccessibility, while the hilly and low mountains are best suited to treeplantations.

Rainfall does not explain regional variation in managed forest and rainforest.This is also understandable. Rainfall might be important for timber volumesper land unit, but not necessarily for allocation of forestland. The regionthat was first settled is on the western coast of the island where rainfallis much lower and cultivation is easier. Rainfall helps forest grow butit also helps agriculture, and the accessibility of rainforest mostly dependson distance and topography rather than the types of rainforest, which aredetermined by rainfall.
 
 

5.0 CONCLUSIONS

This study has demonstrated that natural forestland area and plantationforests react differently to economic and institutional factors explainingtheir development over time. Therefore, studies explaining the developmentof aggregated forestland area may give misleading results and also theirpolicy implications could be questioned. According to our results, highertimber prices have promoted clearing of rainforests in Hainan. In contrast,higher timber prices seem to have been an incentive to investment intoforestry and have had a positive effect on the expansion of plantationsin Hainan. The conflict between agricultural and forestry land seems tobe less serious than the conflict between tropical crops and forest plantations.

In Hainan, there is much land that has a defined owner, but the owneris unable or unwilling to enforce property rights making these areas ineffect open access. Due to, e.g., poor land quality and low output pricesrevenues may not have been sufficient to cover the costs of protectingthe property (cf. Allen 1991 and Barzel 1997). Consequently, the propertyrights are not enforced and inactive management is a rational choice. Inspite of this, land tenure reforms play an important role in forest investmentsin China. Empirical results of this study suggest that de-collectivizationin the form of the household responsibility system is an incentive to forestinvestment into plantations, but may have initially increased rainforestexploitation.

Privatization is not the only way to reform land tenure. Changes inland tenure characteristics, such as renting, transferring part of propertyrights to individuals and cooperatives among different input owners, mayalso promote forestry. For example, the increased control of state ownedforestland by state-owned enterprises has increased plantations but ithas also promoted clearing of rainforest. Therefore, these results suggestthat privatization and increasing timber prices do not necessarily helprainforest conservation and that public intervention as practiced in Hainanto protect rainforests has been justified.

Population growth is not necessarily a serious problem for forestrydevelopment when managed forestry is concerned. The existing potentiallyforestable land area is as large as the currently forested land in Hainan(see Figure 3). Therefore, deforestation resulting from agriculture andindustrial expansion may actually not be as big a problem as the non-activemanagement of this forestable land.

Finally, some limitations of this study should be noted. Our data werecollected mainly from official sources. Changes in conceptual definitionsand statistical data collection methods occurred during the observationperiod. Also, land reforms affecting other land categories (such as realestate business and agricultural sectors) that clearly have direct or indirectinfluences on the relative value for forest management could not be studied.However, the results indicate that when studying land allocation betweenforestry and other land uses, it is important to disaggregate forest landinto managed and natural forest in order to derive relevant policy implications.
 

¹ This work was supported by the InternationalDevelopment Research Centre's (IDRC) Economy and Environment Programmefor Southeast Asia (EEPSEA). For helpful comments, the authors thank HeminiaFrancisco, William Hyde, Nancy Olewiler, Thomas P. Tomich and three anonymousreviewers. All errors are responsibilities of the authors. Back
² Generally population pressure affects landallocation, but the effect on managed and natural forests may differ. Populationdensity may also, among other things, be related to the distance of forestryland to markets. In addition, both economic prosperity (GOP) and populationdensity (PD) may affect forestland allocation through directly unobservablenontimber values of forests. Back
³ This specification provided the best statisticalperformance. Together with the logarithmic transformation the specificationimplies that the impacts of output prices and composite input price areassumed to be equal in absolute terms but to differ in sign. Back
⁴ The aggregation of natural and managed forestlandmay be appropriate for regions dominated by either natural forest or plantationforest. In the case of Hainan Island, natural forest and plantation forestseach account for half of total forests. Aggregation necessarily loses information,therefore, separate regressions are necessary. Back
⁵ Because the estimation periods differed,seemingly unrelated regression (SUR) -method could not be used to estimatethe managed and rainforest equations together and to improve the efficiencyof the estimates. No significant contemporaneous correlation between theerror terms from the equations for rainforest and managed forests was foundwhen the same sample length was used. Back
⁶ The lower statistical significance of thecoefficient for mountain area in the rainforest equation is clearly dueto the fact that we have omitted counties without mountain regions, wherethere is no rainforests left. Therefore, mountain area may be more criticalfor rainforests than our regression shows. Back
 
 

REFERENCES

Allen, D. 1991. What are transaction costs? Research in Law and Economics14: 1-18.

Bao, K. 1991. The Provincial Government's Making an Initiation: Experimentof Sustainable Development in Hainan. Proceedings of International AdvisoryMeeting on the Economic Development of Hainan in Harmony with the NaturalEnvironment. Haikou, China. pp.: 93-96.

Barzel, Y. 1997. Economic Analysis of Property Rights, 2^nded. Cambridge University Press, Cambridge, UK.

Besley, T. 1995. Property Rights and Investment Incentives: Theory andEvidence from Ghana. Journal of Political Economy 103 (5): 903-937.

Bromley, D. W. 1989. Property Relations and Economic Development: TheOther Land Reform. World Development 17 (6): 867-877.

Bromley, D. W. 1991. Environment and Economy: Property Rights and PublicPolicy. Blackwell Publishers, USA

Capistrano, A. D. 1994. Tropical Forest Depletion and the Changing Macroeconomy1967-85. In The Causes of Tropical Deforestation. Edited byBrown, K. and Pearce, D. W. UCL Press, London. pp. 68-85.

Chambers, R. T. 1988. Applied production analysis: A dual approach.Cambridge University Press. Cambridge.

Chomitz. K. M. and Gray, D. P., 1996. Roads, Land, Markets and Deforestation:A Spatial Model of Land Use in Belize. The World Bank Economic Review 10(3) 487-512.

Deacon, R. T. 1994. Deforestation and the rule of law, Land Economics70: 414-430.

Deacon, R. T. 1999. Deforestation and Ownership: Evidence from HistoricalAccounts and Contemporary Data. Land Economics 75 (3): 341-359.

Greene, W. H. 1997. Econometric Analysis, 3rd ed. Prentice-hall InternationalInc. New York.

HAZC (Hainan Agricultural Zoning Commission). 1980. Collection of HainanAgricultural Zoning Reports. Haikou, China.

HFB (Hainan Forestry Bureau). 1957-1995. Hainan Forestry StatisticalMaterials. Haikou, Hainan.

Hoshino, S. 1996. Statistical Analysis of Land-use Change and DrivingForces in the Kansai District, Japan, Working Paper 96-120. IIASA, Austria.

HSB (Hainan Statistical Bureau). 1957-1995. Hainan Statistical Materials.Haikou, China.

Hsiao, C. 1986. Analysis of Panel Data. Cambridge University Press,Cambridge.

Hyde, W. H. 1980. Timber Supply, Land Allocation, and Economic Efficiency.The Johns Hopkins University Press. Baltimore and London.

Hyde, W., Amacher, G. S. and Magrath, W. 1996. Deforestation, ScarceForest Resource, and Forest Land Use: Theory, Empirical Evidence, and PolicyImplication. World Bank Research Observer 11: 223-248.

Kmenta, J. 1986. Elements of Econometrics, 2nd ed., Macmillan.

Kummer, D. and Sham, C. H. 1994. The causes of tropical deforestation:a quantitative analysis and case study from the Philippines. InThe Causes of Tropical Deforestation. Edited by Brown, K. and Pearce,D. W. pp. 146-58.

Laarman, J. G. 1996. Government Policies Affecting Forests in LatinAmerica. Environment Division of Social Programs and Sustainable DevelopmentProgram, Inter-American Development Bank. Washington D. C.

Mather, A. 1990. Global Forest Resources. Timber Press. Portland, Oregon.

Mendelsohn, R. 1994. Property rights and tropical deforestation. OxfordEconomic Papers 46: 750-56.

North, D. 1990. Institutions, Institutional Change and Economic Performance.Cambridge University Press. Cambridge.

North, D. C. 1997. The contribution of the New Institutional Economicsto an Understanding of the transition Problem, UNU/WIDER Annual Lecture1. Helsinki.

Palo, M. 1994. Population and deforestation. In The Causes ofTropical Deforestation. Edited by Brown, K. and Pearce, D. W. UCLPress, London. pp. 42-56.

Panayoutou, T. and Sungsuwan, S. 1994. An econometric analysis of thecauses of tropical deforestation: the case of Northeast Thailand. InThe Causes of Tropical Deforestation. Edited by Brown, K. and Pearce,D. W. UCL Press, London. pp. 192-210.

Place, F. and Otsuka, K. 1997. Population density, land tenure, andtree resource management in Uganda (Draft). International Food Policy Centre.EPTD Discussion Paper 24.

Situ, S. 1992. Historical land use development in Hainan. Hainan Press.Haikou. (in Chinese).

Rudel, T. K. 1998. Is There a Forest Transition? Deforestation, Reforestation,and Development. Rural Sociology 63 (4): 533-552.

von Thunen, J. H. 1875. The Isolated State in Relation to LandUse and National Economy. Schmaucher Zarchlin, Berlin.

Walker, K. R. 1965. Planning in Chinese Agriculture: Socialization andthe Private Sector 1956-1962. Franl Cass & Company Ltd., London.

Wallace, T. D. and Newman, D. H. 1986. Measurement of ownership effectson forest productivity in North Carolina from 1874-1984. Can. J. For. Res.16: 733-738.

Yin, R. and Newman, D. 1997. Impacts of rural reforms: the case of theChinese forest sector. Environment and Development Economics 2: 289-303.

Zeng, Q. 1994. A brief introduction to closing the land for reforestationin Hainan Province. Tropical Forestry Newsletter 2. ITTO, Japan. No. 2.p24.

Zhang, D. 1996. Forest Tenure and Land Value in British Columbia. Journalof Forest Economics 2 (1): 7-29.

Zhang, D. and Pearse, P. 1996. Differences in sivicultural InvestmentUnder Various Types of Forest Tenure in British Columbia. Forest Science42(4): 442-449

Zhang, Y., Dai, G., Hung, H., Kong F., Tian, Z., Wang, X. and Zhang,L. 1999. The Forest Sector in China-Towards a Market Economy. In WorldForests, Society and Environment. Edited by Palo, M. and Uusivuori,J. Kluwer Academic Publishers, the Netherlands. pp.: 371-93.

Zhang, Y. 2000. Deforestation and Forest Transition: Theory and EvidenceIn China. World Forests from Deforestation to Transition? In Palo, M andVanhanen, H. (eds.), World Forests from Deforestation to Transition? KluwerAcademic Publishers, Dordrecht. pp: 19-43

Document(s) 66 de 95