Journal of Forest Economics 20 (2014) 126–140

Contents lists available at ScienceDirect

Journal of Forest Economics

journa l homepage: www.elsev ier .com/ locate / j fe

The effect of collective forestland tenure reformin China: Does land parcelization reduce forestmanagement intensity?

Yi Xiea,∗, Peichen Gongb, Xiao Hanc, Yali Wena

a School of Economics and Management, Beijing Forestry University, Beijing 100083, Chinab Department of Forest Economics, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Swedenc Faculty of Forestry, University of Toronto, 33 Willcocks Street, Toronto, Canada

a r t i c l e i n f o

Article history:Received 23 September 2013Accepted 14 March 2014

JEL classification:Q23N50

Keywords:Land tenure reformCollective forestForest managementForest policyDouble hurdle model

a b s t r a c t

China implemented a new round of collective forestland tenurereform during 2003–2013. In this reform, forestland owned byvillages or township collective organizations were divided into agreat number of small plots and allocated to member householdsof the collectives. A widespread concern about the reform is thatparcelization of forestland might limit farmers’ incentives to investin forest management. This paper examines the factors affectingfarmers’ investment in forest management using household datacollected in four provinces in 2010. The results show that the inten-sity of a household’s investment in forest management is negativelyaffected by its nonfarm income and the average size of forest plots,but positively affected by the easiness in obtaining loan and thetechnical assistance the household receives. We argue that thecounterintuitive effect of nonfarm income on investment inten-sity is due to the increasing marginal cost of own labor input. Theeffects of forest plot size and easiness in obtaining loan suggestthat households have limited amount of capital to invest in forestmanagement. Because of this constraint, parcelization of forestlandresulted from the recent reform has not yet caused any reductionof the intensity of investment in forest management.

© 2014 Department of Forest Economics, Swedish University ofAgricultural Sciences, Umeå. Published by Elsevier GmbH. All

rights reserved.

∗ Corresponding author. Tel.: +86 10 6233 8444; fax: +86 10 6233 7674.E-mail address: xybjfu@126.com (Y. Xie).

http://dx.doi.org/10.1016/j.jfe.2014.03.0011104-6899/© 2014 Department of Forest Economics, Swedish University of Agricultural Sciences, Umeå. Published by Elsevier GmbH. All rights reserved.

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 127

Introduction

About 60% of the forestland in China is collectively owned by villages or township collective eco-nomic organizations, and the rest is under State ownership (The State Forestry Administration, 2010).Collective ownership of forestlands was built up in the 1950s, and these forests were managed bythe collectives until the beginning of the 1980s. The defects and inefficiency of this managementregime were widely acknowledged by the end of the 1970s (Lu et al., 2002). Since 1981, the ChineseGovernment has implemented a series of reforms in order to improve the management of collectiveforests.

There is a huge body of literature on forestland tenure reforms in China. Only a relatively small shareof the studies focused on the effects of these reforms on forest management activities. The results ofthese studies are mixed. Yin and Newman (1997) examined the impacts of the first round of forestlandtenure reform on timber harvest and on the dynamics of forest resource during 1978–1989. Theyfound that timber harvest, timber inventory, and forest area in northern China increased dramaticallybetween 1978 and 1989. During the same time period, southern China experienced only moderateincrease in timber harvest and forest area, whereas the timber inventory decreased. Their results showthat the reform increased timber harvest in both regions, but had opposite effects on the developmentof forest resources in the two regions. Zhang et al. (2000b) examined timber harvest and forest coverduring 1978–1995 in the same regions as Yin and Newman (1997) and found that land tenure reformhad a positive effect on the increase of forest area in both the southern and the northern regions.

Rozelle et al. (2003) examined the impacts of forestland tenure and policy changes on forest areaand timber inventory in China during 1980–1993 using national forest inventory data. The resultssuggest that policy changes since the late 1970s had a positive effect on increasing the level of timberinventory but a negative effect on the change of forest area. Wang et al. (2004) examined the factors thatexplain tree planting in China during the period 1953–2001. They found that forest reform in generalhad a negative effect on the total area of afforestation/regeneration. The reason for the unexpectedresult was, according to the authors, the increase of illegal timber cutting triggered by the reform since1978 and that the reforms did not provide sufficient incentive for tree planting because of a series ofirrational institutional arrangements such as heavy tax burdens.

Xie et al. (2011) examined the effect of the recent reform on forestation at the village level using3 years data (2000, 2003, and 2005/2006) for 288 villages in eight provinces. They found that thereform resulted in a significant increase in the area of newly forested land in the year the reform wasimplemented. The reform also affected positively forestation in the subsequent years, but the effectwas much smaller. They also found that implementation of the reform did not cause an immediateincrease in timber harvest. Since timber harvest did not increase, afforestation reduced the area ofun-wooded land. This explains why the effect of the reform on forestation decreased with time. Italso implies that the reform would cause an increase in the total forest area, at least in the short-run.

In a recent study, Qin and Xu (2013) examined farmer’s forestry investment (labor input andfertilizer application) in Fujian province using household survey data collected in 2006. The studydistinguished among three types of forestland (timber forest, bamboo forest, and economic forest)and several different forms of land right arrangements. A general conclusion of this study is that farm-ers tend to invest less on forest plots for which the land tenure security is perceived to be low. Fortimber forests, they found that increasing harvest quota (the allowable harvest relative to the stand-ing timber stock) would increase the labor input per unit area. Moreover, the size of forest plot has asignificant and negative effect on the labor input per unit area for both timber and bamboo forests.

One consequence of the recent reform is parcelization of forestland – larger forest tracts weredivided into small plots and granted to different households. In general, small forest plots imply higherunit costs for harvesting, regeneration, and other silviculture activities, which usually lead to lowermanagement intensity (Zhang et al., 2005). In the United States and many European countries, forestparcelization is typically associated with reduction of holding size, which often reduces the incentiveand possibility for landowners to conduct intensive management of the forests (Mehmood and Zhang,2001; Zhang et al., 2005; Butler and Leatherberry, 2004; Haines et al., 2011; Hatcher et al., 2013). Manystudies have provided strong evidence that management intensity is positively correlated to the size

128 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

of holding (see, e.g. Straka et al., 1984; Butler and Leatherberry, 2004; Pan and Zhang, 2007; Hatcheret al., 2013).

Family forestry in China differs from that in Europe and North America in many aspects, such asforest owners’ dependency on forestry income and regulations that confine forest management activ-ities. Thus, the management behavior of family forest owners (farmers) in China is not necessarilysimilar to the behavior of NIPF owners in Europe and North America. The purpose of this paper is toexamine the factors that affect farmers’ forestry investment following the recent land tenure reform inChina. In particular, we want to determine whether or not and how forestland parcelization affects theintensity of forestry investment. Forestry investment is measured by the amount of capital and house-holds’ own labor input in afforestation, regeneration, and stand improvements (including weeding,fertilization, pruning, pre-commercial thinning). The sum of capital and labor input is also examinedto achieve a better understanding of forestry investment behavior.

The paper is structured as follows. The next section provides an overview of the collective forestlandtenure reform in China. Section “Conceptual model and econometric approach” presents a conceptualmodel of a representative household’s forestry investment decision and the econometric model. Sec-tion “Data and descriptive statistics” describes the data used in the analysis. The empirical results werepresented in section “Results”. The final section closes with conclusions and a discussion of the policyimplications.

Collective forestland tenure reform in China

Officially, collective forestland tenure reform in China started in 1981, flowing the promulgationof the “Decisions on a number of issues concerning forest protection and forestry development” bythe Central Committee of the Communist Party of China and the State Council. The initial reform,carried out in the early 1980s, involved clarification and confirmation of the ownership of forests andbarren hills, allocation of family hills to member households of the collectives, and establishmentof forest management responsibility system. A family hill entitles a household the ownership of thetrees and permanent use rights of the land. The forest management responsibility system impliesthat each member household is responsible for the management of some assigned forestlands (theso-called responsibility hills) with a benefit-sharing arrangement. Lu et al. (2002) estimated that by1986 nearly 70% of the collective forestlands in the eight southern provinces had been transferred tohousehold management.

The reform in the early 1980s provided rural households some freedom in managing the collectiveforestlands. This had contributed to the excessive timber harvesting and extensive forest destruc-tions following the liberalization of timber trade in southern China in 1985. The situation became soserious that, in 1987, the Government stopped continuing allocation of timberland to individual house-holds, tightened the regulations of the management of forestlands that had already been allocated tohouseholds as family hills or responsibility hills, and launched a harvest quota system.

In the 1990s market mechanisms for transferring the use rights of forests and forestlands wereestablished, in line with the market oriented economic reform. Since 1993, allocation of responsibilityhills to member households of the collectives was replaced with lease of forestlands by all interestedhouseholds and private economic entities. First, the collectives were allowed to sell long-term userights of unwooded forestlands (the so-called four wastelands) through auctions. Later, trade of userights of all kinds of collective forestlands, as well as establishment of forestry cooperatives, waspromoted. The policy change in the 1990s enabled households and private economic entities to acquirelong-term use rights of fairly large areas of forestlands at very low prices (Zhang et al., 2000a; Yin andXu, 2002; Wang et al., 2004; Weyerhaeuser et al., 2006). Despite the increase in the scale of forestryoperation, the productivity of collective forests decreased in the 1990s compared to state ownedforests, mainly because the heavy tax burden and difficulties in obtaining harvest permits decreasedfarmers’ incentives to invest in forestry (Yin and Xu, 2002; Lu et al., 2002; Wen, 2009).

The recent round of collective forestland tenure reform was started in 2003. The aim of the reformis to increase farmer’ forest management incentives by increasing forestland tenure security, reduc-ing the tax burden, relaxing restrictions on timber harvest, strengthening public support (financialand technical) to forest farmers, and promoting the development of stumpage and land use rights

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 129

trade markets (Xie and Wen, 2009). With respect to forest property rights, this reform is in essencea combination of the reform policies adopted in the 1980s and 1990s: long-term use rights of collec-tively owned forestlands are granted to member households of the collectives, and the use rights offorestlands can be traded in open market or transferred in some other way.

The heterogeneity of forestland makes it difficult to fairly allocate the forestland owned by a col-lective to its member households. A common practice is dividing the land into a large number of smallparcels and allocating several scattered parcels to each household. There is a widespread concern thatparcelization of forestland reduces farmers’ incentives to invest in forest management (see, e.g., Zhangand Wen, 2008; FAO, 2009; Wen et al., 2011). There is, however, a lack of adequate empirical evidenceshowing that the recent reform has had negative effects on investment in forest management.

Conceptual model and econometric approach

Conceptual model

In order to identify the factors that could affect farmers’ forestry investment decision, we applya two-period model (Johansson and Löfgren, 1985; Max and Lehman, 1998; Amacher et al., 1999;Conway et al., 2003; Barua et al., 2010) to examine the managerial behavior of a representative house-hold. The representative household is endowed with certain amount of cropland, forest, and laborwhich can be used in agriculture production, in forest management, and/or in wage-earning activi-ties outside the farm. The household determines the optimal use of the resources that maximizes thepresent value of utilities of consumption in two periods. Timber harvest is bounded from above bya harvest quota h̄, and the maximum amount the household can borrow is constrained by a creditlimit D̄. Since we are interested in farmers’ management intensity, our model includes farmers’ ownlabor input and capital investment in forestry, in addition to timber harvest, as decision variables. Onthe other hand, we ignore the amenity values of the forest, because farmers in developing countriestypically pay little or no attention to such benefits (Cubbage et al., 2003; Zhang and Owiredu, 2007).By combining the model of a self-employed forest farmer (Johansson and Löfgren, 1985) and the opti-mal harvest model under credit rationing (Kuuluvainen, 1989), we developed the following decisionmodel.

max U(C1) + U(C2)1 + �

(1)

Subject to:

C1 = P1h + R(Aa, La) + WLw + D − K (2)

C2 = P2[Q0 − h + F(Q0 − h, K, Lf , Af , Z)] − (1 + r)D (3)

La + Lw + Lf = L (4)

0 ≤ h ≤ h̄ (5)

D ≤ D̄ (6)

K, La, Lw, Lf ≥0 (7)

where C1 and C2 are consumptions in period one and two, respectively, U (·) is the household’s utilityfunction, � is the rate of time preference. Q0 is the initial standing timber stock, h is the timber volumeharvested in the first period. P1 and P2 are the stumpage price (timber price net of harvesting costs)in period one and two, respectively,1 D is loan/saving in period one, K is the capital input in forestmanagement in the first period, and r is the market interest rate. R(Aa, La) denotes agricultural income,

1 In a two-period model, the second period includes all future time beyond the first period. Forestry income in the secondperiod includes not only the harvest revenue at the beginning of the second period, but also all revenues from subsequentharvests. Therefore, P2 can be interpreted as the maximum net present value of the forest at the beginning of the second perioddivided by the standing timber stock, or the shadow price of the standing timber stock at the beginning of the second period.

130 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

which is a function of the area of cropland (Aa) and labor input (La) used in agricultural production.Lw is the amount of time spend on wage-earning activities and W is wage rate. Lf and Af are thehousehold’s own labor input in forest management and the area of forestland, respectively.

The forest growth function F(Q0 − h, K, Lf , Af , Z) describes the volume increment in one period– this increment depends on the growing stock of timber left after harvesting in the first period(Q0 − h), capital input (K), labor input (Lf ), forestland area (Af ), and a set of variables (Z) related tothe household’s skills in forest management. In order to simplify the analysis, we employ a standardassumption in defining the budget constraint so that the option of working for others and getting wageexists only in the first period, while income from timber harvesting can occur in both periods.

Eq. (4) means that the representative household allocates a given amount of labor among agricul-ture production (La), forest management (Lf ) and non-farm work (Lw). L is the total labor force ownedby the household. The harvest constraint (5) says that the amount of timber to be harvested in periodone cannot be negative and is bounded from above by the harvest quota h̄. The credit constraint (6)means that the household cannot borrow more than an exogenously determined amount (credit limit)D̄.

Based on an analysis of the first-order conditions for the optimal solution of the decision prob-lem (1)–(7), we derived the following general behavior functions for the optimal timber harvest andoptimal inputs of capital and household’s own labor in forest management2:

K = K(P1, P2, r, Q0, Af , h̄, D̄, Aa, W, L, M), (8a)

Lf = Lf (P1, P2, r, Q0, Af , h̄, D̄, Aa, W, L, M), (8b)

where M is a vector consisting of the elements of Z and other household characteristics that affect thepreferences of the household.

Empirical model specification

Based on the analysis presented above we formulated the following reduced form for capitalinvestment (K), own labor input (Lf), and the aggregate input of capital and own labor force (I):

Yi = Fi(age, education, labfor, province, cropland, nonfarm, loan, fampr, areaplot,

quota, techser) (9)

where i = 1, 2, or 3, and Y1 = ln(K + 1), Y2 = =ln(Lf + 1), Y3 = ln(l + 1).Capital investment (K) is defined as the expenditure on forest management, including the cost of

hiring people. Own labor input (Lf) refers to the amount of time (person-days) household member(s)spend on managing their own forestland. The aggregate input (I) is the sum of capital investment andthe cost of own labor input (calculated using the average local wage rate). The dependent variables aremeasured in terms of the average amount per unit area forestland. Obviously, the aggregate input perunit area forestland is an adequate measure of the management intensity. The purpose of estimatingthe capital investment and own labor input separately is to examine to what extent the householduses its own labor as a substitute for cash input when facing a binding credit limit.

Household characteristic variables include household’s location, i.e., which province the sampledhousehold comes from (denoted as province), age of household head (denoted as age), education levelof household head (denoted as education), and the number of labor force in the family (denoted aslabfor). Zhang and Owiredu (2007) found that the age of household head had positive impacts on plan-tation establishment. However, Romm et al. (1987) reported that higher age reduces the probability ofsilviculture investment, and Zhang and Flick (2001) found that age has no influence on tree planting.It is likely that the qualitative impact of the age of household head on labor and capital input in forestmanagement changes as the age increases. To account for this possibility, we include the square ofthe age of household head in the empirical models (Shi et al., 2007). Education level is expected to

2 Derivation of these behavior functions is available from the author upon request.

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 131

have a positive influence, for some studies have found that higher education favors forestry activity(Zhang and Owiredu, 2007; Joshi and Arano, 2011). The more labor force a family owns, the more laborforce is expected to be used in forestry, but less capital is invested due to the substitution relationshipbetween labor force and capital. Thus, the influence of number of labor force on the aggregate inputis ambiguous.

Future timber price is deduced to have a positive effect on labor input and capital investment.Regarding our cross section data, the impact of timber price is embedded in the effect of householdlocation. Due to the lack of accurate data, timber price is excluded in the regression analysis.

Previous studies show that capital market imperfection affects forest management decisions (see,e.g., Uusivuori and Kuuluvainen, 2001). We include in the empirical models a dummy variable (denotedby loan) which indicates whether a household can easily obtain a loan from bank. Intuitively, thedifficulty in obtaining loan would reduce capital investment in forest management. The reductionin capital investment would be compensated partly by increasing own labor input. Therefore, it isexpected that the easiness in obtaining loan would have a positive effect on capital investment and anegative effect on labor input. The effect on the aggregate input is expected to be positive.

The area of cropland (denoted as cropland) has both negative and positive impacts on forestryinvestment. One the negative side, a larger area of cropland would require more time for agricultureproduction and thus less time would be available for forest management. On the positive side, thosewho spend more time in agriculture production are less likely to be absent in ownership, and thenwill have higher intention on forest management (Romm et al., 1987; Conway et al., 2003). Moreover,a larger area of cropland also means a larger income from agriculture production, which strengthensthe financial position of the household and thereby would favor the capital investment in forestry.

The term W in the conceptual model refers to the constant marginal return of labor input in non-farm work. In reality, farmers can get non-farm income not only by working for others (wage income),but also from doing some business, which makes the marginal return difficult to measure. To overcomethis difficulty, we use the proportion of nonfarm income to the total income of a household (denotedas nonfarm) in the empirical model to represent the opportunity cost of labor. A larger nonfarm incomeimplies that the return on labor input in nonfarm activities is high and/or a larger amount of time isspent in such activities, which means that the opportunity cost of working in the forest is higher and/orless time is available for forest management. Therefore, we expect that the variable nonfarm affectslabor input in forest management negatively. However, this variable is expected to have a positiveeffect on capital investment in forestry. A household with a larger nonfarm income has greater financialpotential to invest in forestry. It is also possible that such a household would invest more capitalto compensate the lack of its own labor input (e.g. by hiring someone to conduct the managementactivities). The opposite effects of nonfarm income on own labor input and on capital investmentimply that the effect of nonfarm on the aggregate input is ambiguous.

Two variables are used to describe the forestland managed by each household. The first is theaverage area per plot of forestland (denoted by areaplot). The forestland managed by each householdtypically consists of several scattered plots. A primary concerns about the recent tenure reform of col-lective forestland is that the small forest plots could significantly reduce farmers’ incentives to managethe forests. Our hypothesis is that the average area per plot affects forest management investmentpositively. Our data show that the total area of forestland managed by each household is stronglycorrelated to the average area per plot forest land (the coefficient of correlation is 0.83). Therefore,the total area of forestland is not included in the empirical model. The second variable, fampr, is theshare of the so-called family hills in the total area of forestland managed by each household. Familyhill is an arrangement which grants individual households the ownership of resources on delimitatedforestlands and exclusive rights of using the land over an indefinite period of time. Traditionally, thishas been a relative stable arrangement and family hills are in general perceived as more secure thanforestland acquired through contracts. For this reason, we use the share of family hills in the total areaof forestland managed by each household as an indicator of land tenure security. The variable fampris expected to have positive effects on forest management investment (Zhang and Pearse, 1997; Placeand Otsuka, 2000; Zhang and Owiredu, 2007; Qin and Xu, 2013).

The policy variables included in the empirical models are harvest quota (denoted as quota) andtechnical service (denoted as techser). It is argued that the harvest quota system limits management

132 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

Table 1Description of the study area at provincial level.

Zhejiang Liaoning Shaanxi Chongqing

Total forest area (million ha)a 5.84 5.12 7.68 2.87Forest coverage rate (%)a 57.4 35.1 37.2 34.9Area of forestland per household (ha)b 6.1 3.3 8.2 0.1Rural household income per capita (CNY)a 10 007 5958 3437 4478Proportion of forestry income to total household income (%)a 22.6 37.6 14.7 4.6

a China Statistical Year Book (2010). http://www.stats.gov.cn/tjsj/ndsj/2010/indexch.htm.b Data collected by the authors.

incentives (e.g., Miao and West, 2004). The need for harvest quota to harvest timber may reduce theexpected return of forestry investment. Farmers would have lower propensity in forest investment ifthey feel that it is difficult to obtain large enough harvest quota through official application process.3

Technical service is expected to have positive impacts on forest management intensity (e.g., Hybergand Holthausen, 1989; Boyd, 1984).

Data and descriptive statistics

The study area

Our data were collected in Zhejiang, Liaoning and Shaanxi provinces, and Chongqing municipalcity (Table 1). Zhejiang province is located in southeast China. It has the third highest forest coveragerate, and has a relatively well developed economy. Liaoning province is located in northeast China,and is the main distribution zone of collective forests in northern China. Shaanxi and Chongqing arelocated in northwest and southwest China, respectively, and are less developed in terms of rural house-hold income. In terms of forest resources per household, Shaanxi ranks the first, but the importanceof forestry income ranks the third among the four provinces. Farmers in Chongqing have relativelylittle forest resources, and a low proportion of forestry income. It is noted that northern China dif-fers significantly from southern China in terms of climate, forest rotation, and tradition of forestmanagement.

Data collection

We applied stratified sampling method to select households to be interviewed. Following China’sadministrative system, we selected three counties from each province, three townships from eachcounty, two villages from each township, and 10 households from each village. Altogether, we inter-viewed representatives of 720 households that were randomly selected from the list of the householdsprovided by the village committees. The interviews were conducted in the selected villages in 2010.

The questionnaire used in the interviews for collection of primary data included sections on basichousehold variables, cropland resources, off-farm activities, forest plots, forest management andunderstanding about the institutional arrangements in forestry production. Detailed input-outputdata were collected for the households. Household’s income was computed by adding income fromcrops, livestock, timber and non-timber forest products, off-farm wage income, and income from otherbusiness activities. The investment in forest management was calculated by accounting for inputson planting, tending and thinning. Information about the forestlands was extracted from the forestproperty right certificates provided by the interviewees. The primary data covered the entire year of2009.

3 From forest owners’ perspective, the so-called harvest quota is in fact a harvesting permission. Before conducting a harvest,a forest owner (farmer) needs to apply for a harvest quota of a desired amount. Unfortunately, such applications often are nothandled in a transparent and fair manner. For those who do not have the right contact, the approved quota is usually lowerthan their needs. Obviously, this situation means that there are “unofficial” (and costly) ways to obtain the need harvest quota.

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 133

Table 2Descriptive statistics of the data.

Variable Definition Mean Standarddeviation

Dependent variablesK Logarithm of average capital investment divided per unit area

plus one (USD per ha)1.12 2.09

Lf Logarithm of average own labor input per unit area offorestlands managed by family plus one (person-days per ha)

1.01 0.80

I Logarithm of average aggregate input of capital and labor forcedivided by area of forestlands managed by family plus one(USD per ha)

3.39 2.67

Independent variablesage Age of head of household (years) 50.78 11.17education Education level of head of household (0 = illiteracy, 1 = primary

school, 2 = middle school, 3 = high school, 4 = university orabove)

1.79 0.82

labfor Number of persons in work in the family (persons) 2.46 1.20cropland Area of cropland managed by the household (ha) 0.35 0.38nonfarm Proportion of income from nonfarm sources to total income 0.59 0.32loan Difficulty in obtaining loan (1 = easy to obtain loan,

0 = otherwise)0.18 0.38

fampr Proportion of family hills to the total forestland managed byeach household

0.42 0.46

areaplot Average forestland area per plot (ha/plot) 1.25 2.32quota Harvest quota (1 = easy to obtain quota; 0 = otherwise) 0.37 0.48techser Attitude to technical service (1 = satisfied, 0 = otherwise) 0.43 0.50ln Liaoning province (1 = yes, 0 = otherwise) 0.27 0.44zj Zhejiang province (1 = yes, 0 = otherwise) 0.27 0.44sx Shaanxi province (1 = yes, 0 = otherwise) 0.23 0.42

Data description

We used data for 647 of the 720 sampled households in the analysis. 42 of the households didnot have any forest at the time of the interview, while the forestland area for 31 households wasabnormally large. These households were excluded from the analysis.

Descriptive statistics of the data are presented in Table 2. Among the 647 households, there were164 (25%) and 417 (64%) with positive capital investments and own labor input in forest management,respectively. The number of households that made an input of capital and/or labor was 441 (68%). Onaverage, each household spent about 14 person-days and $73.824 in forest management during 2009.The aggregate input was the sum of capital investment and the cost of own labor input calculatedusing average local wage rates (120, 90, 70 and 60 CNY per day in Zhejiang, Liaoning, Chongqing andShaanxi provinces, respectively).

Each household’s total inputs of labor and capital were divided by its total area of forestland toobtain the inputs per unit area. In order to avoid serious skewness, we made logarithmic transforma-tions of the per ha input plus one. After logarithmic transformation, the skewness values of the threeindependents of K, Lf and I respectively fall from 6.30 to 1.60, 4.40 to 0.15, and 3.84 to −0.11.

Most of the household heads were close to 50 years old. Most of the household heads had finishedmiddle school, some of them were illiterate (29 heads, accounting for 4%), and only 14 of them hadeducation in university (accounting for 2%). 21 percent of the households had members working invillage committees.

The villages surveyed were all located in mountainous areas where the proportion of arable land islow. On average, each household had 0.35 ha (0.081 ha per person) croplands, which is slightly lowerthan the national average of 0.092 ha per person. Variation among the 647 households in the area of

4 1 USD = 6.82 CNY as of December 31, 2009.

134 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

cropland is smaller (standard deviation = 0.38) than the variation in forestland area (mean = 4.42 ha,standard deviation = 19.09). The average area per plot forestland is 1.25 ha. The proportion of familyhill to total forestland managed by each household ranges from zero to 100%, and the mean value is42%.

The absolute value of household’s nonfarm income ranges from zero to more than 40 thousandUSD. On average, 59% of the total income comes from nonfarm sources. Due to undeveloped capitalmarket, which is quite common in rural area of developing countries, 82% of the households in thesample indicated that it was difficult to obtain loan. 37% of the households indicated that “it is difficultto get harvest quota”, which means that it is costly to get the needed harvested permission, and 43%were satisfied with the technical services.

Results

Because the observed data on forest management inputs are censored from below at zero, modelsextended from function (9) could be estimated by using a censored regression or Tobit model (Tobin,1958). In such a model, each independent variable affects the probability of the dependent variablebeing positive and the magnitude of the dependent variable in the same direction, which is not rea-sonable in some practical situations (Greene, 2010). In this case, estimation of a Tobit model shouldalways be accompanied by estimation of a Cragg (1971) specification, which is a more general doublehurdle model consisting of a Probit model and a Truncated normal model (Smith and Brame, 2003).Whether a Tobit model or rather a double hurdle model should be employed can be determined byconducting a likelihood ratio test that compares the Tobit and the sum of the log likelihood func-tions of the Probit and the Truncated models (Greene, 2010). The likelihood ratio test statistics forcapital investment, labor input, and aggregate input are respectively 2[683.7 − 297.3 − 260.8] = 260.8,2[712.6 − 348.1 − 358.7] = 11.6, and 2[1328.7 − 319.1 − 721.6] = 1040.7 with the same 15 degree offreedom. The first and the third statistics exceed the critical value of Chi-square distribution at signif-icance of 1% (which is 30.58), but not the second one. The test suggests that the double hurdle modelis preferable to the Tobit model for the analysis of capital investment as well as aggregate input. Withrespect to labor input, however, we cannot reject the Tobit model. The results for the Tobit, Probitand truncated regression models are all presented below. But we will focus on the results of the dou-ble hurdle model in the following discussion. The Chi-square test shows that the parameters in themodels are jointly significant at the 1% level. The Skewness/Kurtosis tests for normality show that theresiduals for the truncated models are normally distributed.

Determinants of capital investment

The regression results (Table 3) show that seven factors have statistically significant impacts onthe probability of investing capital in forest management. These are the age of household head, area ofcropland, nonfarm income, difficulty in obtaining loan, proportion of family hills, technical assistance,and geographical location of household. Factors that have significant effects on the investment inten-sity in the truncated regression model include nonfarm income, difficulty in obtaining loan, averagearea per plot forestland, and geographical location of household. The other factors included in theregression models (i.e. education of household head, labor force, and harvest quota) do not have anysignificant impact either on the probability or on the size of capital investment.

The probability of making capital investment in forest management increases as the age of thehead of household increases up to about 56 years. Further increases in the age would lead to lowerprobability of capital investment. For those households that invested capital in forest management,the age of the head of household has a negative effect on the size of the investment, but the effectis not significant. Cropland area has a positive impact on the probability of capital investment. Itseffect on the size of the investment is negative, though insignificant. The effects of the proportionof nonfarm income to total income on the probability and the size of investment are both negative.As expected, households that can easily obtain loan as well as those that are satisfied with technicalservice have significantly higher probability of making capital investment in forest management. Theresults also show that, in contrast to our expectation, the proportion of family hills to total forestland

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 135

Table 3Results of the Tobit model and the double hurdle model on capital investment.

Independentvariables

Tobit Probit Truncated Regression

Coefficient t-Ratio Coefficient t-Ratio Coefficient t-Ratio

age 0.500** 2.22 0.106** 2.32 −0.0163 −0.18age2 −0.005** −2.11 −0.0009** −2.16 −0.00004 −0.05education 0.122 0.32 0.0359 0.46 −0.129 −0.98labfor −0.281 −1.13 −0.0594 −1.16 0.0198 0.22cropland 2.299** 2.47 0.517*** 2.74 −0.302 −0.88nonfarm −2.399** −2.21 −0.442** −1.96 −0.749** −1.97loan 2.475*** 2.99 0.500*** 2.93 0.144 0.52areaplot −0.268* −1.96 −0.0214 −0.77 −0.253*** −6.25fampr −3.072*** −3.76 −0.578*** −3.50 −0.199 −0.62quota 0.664 1.04 0.106 0.80 0.240 1.11techser 2.435*** 3.97 0.539*** 4.33 −0.138 −0.67ln −6.465*** −5.03 −1.243*** −4.86 −0.620 −1.14zj 2.215** 2.33 0.464** 2.37 −0.224 −0.66sx 2.068** 2.06 0.540*** 2.60 −1.027*** −3.12Constant −15.62*** −2.64 −3.539*** −2.97 6.959*** 2.91

Statistics diagnosisChi-squared 136.5 137.9 135.0Log lik. −683.7 −297.3 −260.8pseudo R2 0.091 0.188N 647 647 164

* Statistical significance at the 10% level.** Statistical significance at the 5% level.

*** Statistical significance at the 1% level.

managed by the family has negative impacts on the probability and intensity of capital investment.This may have resulted from the fact that the family hills are usually of low quality and has less valuefor investment. With respect to the effect of geographic location, households in Liaoning are less likelyto invest capital, while those in Zhejiang and Shaanxi provinces have higher investment probability,compared with households in Chongqing. Households in Liaoning, Zhejiang, and Shaanxi provinceshave lower investment intensity than those in Chongqing. But the difference is statistically significantonly between households in Shaanxi and those in Chongqing.

The average area per plot forestland has negative effects on the probability and the size of capitalinvestment in forest management. Thus, households with smaller plots of forestland tend to investmore per unit area. The effect on investment intensity is significant at 1% significance level.

Determinants of own labor input

Table 4 presents the regression results for household’s own labor input in forest management. Fac-tors that have significant effects on the probability of inputting labor in forest management includenonfarm income, proportion of family hills, technical assistance, and geographical location of house-hold. Only three factors (the average area per plot forestland, technical service, and geographicallocation of household) have significant effects on the input level.

The proportion of nonfarm income to total income affects both the probability and the amount ofown labor input negatively. Satisfaction with technical service has significant and positive impactson the probability and amount of own labor input. The proportion of family hill to total forestlandarea affects the probability of own labor input negatively, and it affects the amount of labor inputpositively. The average area per plot forestland has a negative effect on the probability as well as onthe amount of own labor input, the latter effect is statistically significant. Moreover, the results showthat the probability of having positive own labor input in forest management is significantly higherwhile the amount of own labor input is significantly smaller for households in Liaoning, Zhejiang andShaanxi provinces than those in Chongqing.

136 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

Table 4Results of the Tobit models and the double hurdle model on input of labor force (person-days).

Tobit Probit Truncated Regression

Coefficient t-Ratio Coefficient t-Ratio Coefficient t-Ratio

age −0.0214 −0.82 −0.0525 −1.29 −0.00423 −0.22age2 0.000211 0.84 0.000458 1.18 0.0000661 0.36education 0.0149 0.30 0.0202 0.27 0.00727 0.19labfor 0.0449 1.39 0.0738 1.52 −0.000368 −0.02cropland −0.102 −0.87 −0.0875 −0.47 −0.0273 −0.33nonfarm −0.403*** −2.94 −0.865*** −4.01 −0.0882 −0.88loan 0.00838 0.08 −0.169 −1.01 0.124 1.53areaplot −0.100*** −4.97 −0.0192 −0.73 −0.114*** −7.81fampr −0.479*** −4.76 −0.610*** −4.34 0.0289 0.35quota 0.0882 1.01 0.128 1.00 0.0169 0.26techser 0.194** 2.44 0.200* 1.69 0.118*** 2.01ln 0.419*** 3.14 1.393*** 7.15 −0.922*** −8.39zj 0.468*** 3.52 1.594*** 8.19 −1.063*** −9.38sx 0.498*** 3.49 1.199*** 5.86 −0.711*** −6.15Constant 0.747 1.11 1.276 1.21 1.709*** 3.51

Statistics diagnosisChi-squared 76.19 146.0 234.2Log lik. −712.6 −348.1 −358.7Pseudo R2 0.051 0.173N 647 647 417

* Statistical significance at the 10% level.** Statistical significance at the 5% level.

*** Statistical significance at the 1% level.

The set of factors that have significant effect on labor input differ from those that significantly affectscapital investment. A comparison of Table 4 with Table 3 shows that the age of household head, laborforce in household, area of cropland, and easiness in obtaining loan affect the probability of usingown labor and the probability of investing capital in forest management in opposite directions. Theeducation level of household head, nonfarm income, per plot forestland area, proportion of family hill,harvest quota, and technical assistance affect these probabilities in the same direction. This observationsuggests that the factors influencing farmers’ forestry decisions can be divided into two groups: thosethat affect the decision on whether to make an input in the forest management or not, and those whichaffect the choice between the two types of inputs (own labor and capital).

For those households that made an input in forest management, we could also identify two cate-gories of influencing factors. One category consists of those that affect the level of total input, i.e. thosefactors that affect the amount of labor input and the size of capital investment in the same direction.The area of cropland, nonfarm income, difficulty in obtaining loan, per plot forestland area, and har-vest quota belong to this category. The other category includes those that affect the type of input,i.e. the factors that affect the amount of labor input and the size of capital investment in oppositedirections. These include the education level of household head, labor force, proportion of family hill,and technical assistance.

Because of the opposite effects of some of the factors on own labor input and on input of capital, withrespect both the probability and the size of the input, it is necessary to examine the aggregate inputof labor and capital in order to examine how these factors affect the intensity of forest management.

Determinants of aggregate input of capital and labor

The regression results for the aggregate input of capital and labor are presented in Table 5. Notethat, in our data set, the own labor input of a household contributes, on average, twice as much to theaggregate input as the capital investment does. Therefore, the effect of each independent variable onthe aggregate input depends to a larger extent on its effect on labor input than on its effect on capital

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 137

Table 5Results of the Tobit model and the double hurdle model on aggregate input.

Tobit Probit Truncated Regression

Coefficient t-Ratio Coefficient t-Ratio Coefficient t-Ratio

age −0.0375 −0.39 −0.0339 −0.81 0.0196 0.48age2 0.000294 0.32 0.000258 0.65 −0.000142 −0.36education 0.124 0.68 0.0330 0.42 0.0144 0.18labfor 0.0768 0.65 0.0339 0.68 −0.0268 −0.52cropland −0.116 −0.27 0.0734 0.36 −0.112 −0.64nonfarm −2.098*** −4.22 −0.985*** −4.29 −0.413* −1.95loan 0.248 0.62 0.0226 0.13 0.297* 1.75areaplot −0.314*** −4.38 −0.0252 −0.91 −0.340*** −10.91fampr −1.902*** −5.22 −0.680*** −4.73 0.0670 0.39quota 0.392 1.24 0.118 0.90 0.164 1.17techser 0.965*** 3.35 0.404*** 3.23 0.186 1.50ln 2.977*** 6.14 1.261*** 6.33 −1.122*** −4.87zj 4.179*** 8.65 1.743*** 8.60 −1.090*** −4.61sx 3.009*** 5.80 1.319*** 6.13 −1.231*** −5.08Constant 2.640 1.07 1.004 0.93 5.935*** 5.67

Statistics diagnosisChi-squared 157.8 171.4 291.3Log lik. −1328.7 −319.1 −721.6Pseudo R2 0.056 0.212N 647 647 441

* Statistical significance at the 10% level.*** Statistical significance at the 1% level.

investment. The regression results (see Table 5) show that the variables that have significant impactson the probability of inputting capital and/or labor are the same as those that have significant effectson the probability of using own labor force in forest management.

These variables (i.e. nonfarm, fampr, techser, ln, zj and sx), with the exception of the dummy variableln for Liaoning province, are also those that affect the probability of investing capital in the same wayas they affect the probability of using own labor in forest management. A change in each of thesevariables increases (or decreases) both the probability of investing capital and the probability of usingown labor in forest management, thereby increases (or decreases) the probability of the aggregateinput being greater than zero.

There are three more variables (age, cropland, and loan) that have significant impacts on the prob-ability of investing capital in forest management. However, these variables affect the probability ofusing own labor and the probability of investing capital in opposite directions. The effects on theprobability of investing capital are partially canceled by the effects on the probability of using ownlabor force, which makes the effect of these three variables insignificant in the Probit model for theaggregate input.

For households that had made a positive input of own labor force and/or capital in forest manage-ment, the amount of the aggregate input is significantly affected by nonfarm income, the easiness inobtaining loan, and the average area per plot forestland. The effect of nonfarm income and the averagearea per plot forestland are negative whereas the effect of the easiness in obtaining loan is positive.Moreover, the amount of aggregate input is significantly lower for households in Liaoning, Shaanxi,and Zhejiang than those in Chongqing.

Conclusions and discussion

This study reveals that there are significant geographic variations in farmers’ investments in forestmanagement in China. Farmers in Liaoning, Zhejiang, and Shaanxi province have significantly higherprobabilities of investing in forest management, but their investments are significantly smaller, com-pared with farmers in Chongqing. There are also significant differences among different provinces

138 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

with respect to farmers’ choices of inputs in forest management. Farmers in Liaoning province areless likely to make a capital investment than farmers in the other provinces. But farmers in Liaoninghave a higher probability of using own labor force in forest management than those in Shaanxi andChongqing.

Apart from the geographic variations, there are five factors (nonfarm income, easiness in obtainingloan, average area per plot forestland, proportion of family hill, and technical assistance) that havesignificant impacts on the aggregate input in forest management. The proportion of nonfarm incometo a household’s total income as well as the average area per plot of forestland has negative impactson both the probability of making an investment and the size of the investment in forest manage-ment. Easiness in obtaining loan and technical assistance affect positively the probability of makingan investment and the size of the investment. The proportion of family hill to the total area of forestlandmanaged by a household has a negative effect on the probability of investing in forest management,but it has a positive effect on the size of the investment for those who choose to make an investment.

Our results do not support the hypothesis that fragmentation of forest land leads to reduced invest-ments in forest management (e.g. Zhang and Wen, 2008; Wen et al., 2011). On the contrary, the resultsimply that an increase in the area per plot forestland would lead to lower investment intensity. Thisresult is consistent with the findings of Qin and Xu (2013). A likely reason for this effect is that theamount of investment a household can afford is limited, because of imperfection of capital market.Recall that there is a strong positive correlation between the total area of forestland a householdmanages and the average area per plot forestland. With a more or less fixed amount of investment,a larger area of forestland implies nothing but a smaller investment per unit area, i.e. lower invest-ment intensity. The effect of the difficulty in obtaining loan on the aggregate input indirectly supportsthe argument above. Households that can easily obtain loan invest significantly more per unit areaforestland than those that had difficulties to obtain loan.

There is a clear tendency that some farmers choose to use their own labor force to substitute forcapital investment in forest management and vice versa. Younger farmers, as well as those who havesmaller area of cropland or have difficulties in obtaining loan, are more likely to have a positive inputof their own labor but less likely to have a positive input of capital in forest management. The capitalinvestment in this study includes the cost of hiring people. Thus, technically, capital can be regardedas a perfect substitute of household’s own labor force. Given this, the deliberate choices made by thefarmers between the two types of inputs indicate that the opportunity cost of using a household’s ownlabor force in forest management differs from the cost of hiring others to do the work, or the amountof capital a household can invest in forest management is limited.

A majority of the farmers still found it difficult to obtain harvest quotaharvest quota, which led tolower investment in forest management. Consistent with earlier findings (e.g., Miao and West, 2004;Qin and Xu, 2013), our results show that relaxing the constraint of harvest quotaharvest quota wouldhave a positive effect on forestry investment. However, the effect is statistically insignificant.

The focus of this paper is the determinants of farmers’ investment in forest management. Becauseof the lack of information about the management activities each household needs to carry out inits forest, we cannot draw any conclusion about the adequacy of the investment. However, if it isnecessary to stimulate the investment in forest management, our results suggest that policy makersshould first consider to provide better technical services to farmers and to improve their possibilityof getting loan to finance the investment.

Acknowledgments

The authors acknowledge the National Natural Science Foundation of China (Grant No. 71003007)and the Fundamental Research Funds for the Central Universities (Grant No. RW2010-2) for financialsupport. They want to thank the Department of Forest Economics of Swedish University of AgricultureSciences for hosting the first author as visiting researcher from April 2011 to March 2012. Dr. CaiZhen from University of Missouri, and Dr. Li Qiang from Beijing Forestry University provide beneficialsuggestion on data analysis. Two anonymous referees and the editor of the journal have contributedto improving the paper.

Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140 139

References

Amacher, G., Brazee, R., Koskela, E., Ollikainen, M., 1999. Bequests, taxation, and short and long run timber supplies: anoverlapping generation problem. Environ. Resour. Econ. 13, 269–288.

Barua, S., Kuuluvainen, J., Laturi, J., Uusivuori, J., 2010. Effects of forest taxation and amenity preferences on nonindustrial privateforest owners. Eur. J. For. Res. 129 (2), 163–172.

Boyd, R., 1984. Government support of non-industrial production: the case of private forests. Southern Econ. J. 51, 89–107.Butler, B.J., Leatherberry, E.C., 2004. America’s family forest owners. J. For. 102, 4–9.Conway, C., Amacher, G., Sullivian, J., Wear, D., 2003. Decisions nonindustrial forest landowners make: an empirical examination.

J. For. Econ. 9, 181–203.Cragg, J., 1971. Some statistical models for limited dependent variables with application to the demand for durable goods.

Econometrica 39, 829–844.Cubbage, F., Snider, A., Abt, K., Moulton, R., 2003. In: Sills, E.O., Abt, K. (Eds.), Private Forests Management and Policy in a Market

Economy. Kluwer Academic Publishers, Dordrecht, The Netherlands.FAO, 2009. Have decollectivization and privatization contributed to sustainable forestry management and poverty alleviation

in China? Forestry Policy and Institutions Working Paper No. 23. FAO, Rome.Greene, W., 2010. Econometric Analysis, 7th ed. Macmillan, New York, pp. 892–898.Haines, A., Kennedy, T., McFarlane, D., 2011. Parcelization: forest change agent in northern Wisconsin. J. For. 109, 101–108.Hatcher Jr., J.E., Straka, T., Greene, J.L., 2013. The size of forest holding/parcelization problem in forestry: a literature review.

Resources 2, 39–57.Hyberg, B., Holthausen, D., 1989. The behavior of non-industrial private forest landowners. Can. J. For. Res. 19, 1014–1023.Johansson, P., Löfgren, K., 1985. The Economics of Forestry and Natural Resources. Basil Blackwell Inc., New York, pp. 292.Joshi, S., Arano, K., 2011. Determinants of private forest management decision: a study on west Virginia NIPF landowner. For.

Policy Econ. 11, 118–125.Kuuluvainen, J., 1989. Nonindustrial private timber supply and credit rationing: microeconomic foundations with empirical

evidence from the Finnish case. Swedish University of Agriculture Sciences, Department of Forest Economics, Report 85.Umeå, Sweden.

Lu, W., Landell-Mills, N., Liu, J., Xu, J., Liu, C., 2002. Getting the Private Sector to Work for the Public Good—Instruments forSustainable Private Sector Forestry in China. International Institute for Environment and Development, London.

Max, W., Lehman, D., 1998. A behavioral model of timber supply. J. Environ. Econ. Manage. 15, 71–86.Mehmood, S.R., Zhang, D., 2001. Forest parcelization in the United States: a study of contributing factors. J. For. 99, 30–34.Miao, G., West, R., 2004. Chinese collective forestland: contribution and constrains. Int. For. Rev. 6 (3–4), 282–298.Pan, Y., Zhang, Y., 2007. Trends among family forest owners in Alabama, 1994–2004. Southern J. Appl. For. 31 (3), 117–123.Place, F., Otsuka, K., 2000. Population pressure, land tenure, and tree resource management in Uganda. Land Econ. 76 (2),

233–251.Qin, P., Xu, J., 2013. Forest land rights, tenure types, and farmers’ investment incentives in China: an empirical study of Fujian

Province. China Agric. Econ. Rev. 5 (1), 154–170.Romm, J., Tuazon, R., Washburn, C., 1987. Relating investment to the characteristics of nonindustrial private forestland owners

in northern California. For. Sci. 33 (1), 197–209.Rozelle, S., Huang, J., Benziger, V., 2003. Forest exploitation and protection in reform China: assessing the impact of policy,

Tenure, and Economic Growth. In: Hyde, W., Belcher, B., Xu, J. (Eds.), China’s Forest Policy: Global Lessons from MarketReforms. Resources for the Future, Washington, DC, pp. 109–134 (Chapter 6).

Shi, X., Heerik, N., Qu, F., 2007. Choices between different off-farm employment sub-categories: an empirical analysis for JiangxiProvince, China. China Econ. Rev. 18, 438–455.

Smith, D., Brame, R., 2003. Tobit models in social science: some limitations and a more general alternative. Sociol. Methods Res.31 (3), 364–388.

Straka, T., Wisdom, H., Moak, J., 1984. Size of forest holding and investment behavior of nonindustrial private owners. J. For. 82,495–496.

The State Forestry Administration, 2010. China Forestry Statistics Yearbook. China Forestry Publisher, Beijing.Tobin, J., 1958. Estimation of relationships for limited dependent variables. Econometrica 26, 24–36.Uusivuori, J., Kuuluvainen, J., 2001. Benefits of forest-owner collaboration and imperfect capital markets. For. Sci. 47 (3),

428–436.Wang, S., van Kooten, G., Wilson, B., 2004. Mosaic of reform: forest policy in post-1978 China. For. Policy Econ. 6, 71–83.Wen, T., 2009. Understanding the Economic Environments of the Three Reforms of Forest Property Rights,

http://www.forestry.gov.cn/Zhuanti/content gzhy/267078.htmlWen, Z., Guo, M., Lu, L., 2011. Jiangsu provincial forest fire prevention after forest tenure reform policy research support. For.

Econ. 5, 89–92 (in Chinese).Weyerhaeuser, H., Kahrl, F., Su, Y., 2006. Ensuring a future for collective forestry in China’s southwest: adding human and social

capital to policy reforms. For. Policy Econ. 8, 375–385.Xie, Y., Wen, Y., 2009. Empirical study on the influence factors of farmers’ forestland and stumpages transfer. J. Beijing For. Univ.

(Soc. Sci.) 8 (4), 48–54 (in Chinese).Xie, L., Berck, P., Xu, J., 2011. The Effect on Forestation of the Collective Forest Tenure Reform in China.,

http://dx.doi.org/10.2139/ssrn.1782289.Yin, R., Newman, D., 1997. Impacts of rural reforms: the case of Chinese forest sector. Environ. Dev. Econ. 2 (3), 291–305.Yin, R., Xu, J., 2002. A welfare measurement of China’s rural forestry reform during the 1980s. World Dev. 30 (10), 1755–1767.Zhang, D., Flick, W., 2001. Sticks, carrots, and reforestation investment. Land Econ. 77, 443–456.Zhang, D., Owiredu, E., 2007. Land tenure, market and the establishment of forest plantation in Ghana. For. Policy Econ. 9,

602–610.Zhang, D., Pearse, P., 1997. The influence of the forms of tenure on reforestation in British Columbia. For. Ecol. Manage. 98,

239–250.

140 Y. Xie et al. / Journal of Forest Economics 20 (2014) 126–140

Zhang, L., Wen, C., 2008. Impacts of collective forest tenure reform on farmer households’ livelihood – an empirical analysisBased on Jiangxi, Fujian, Liaoning and Yunnan Provinces. Sci. Silvae Snicae 44 (7), 73–78 (in Chinese).

Zhang, A.L., Wu, Z., Zhu, F., 2000a. Push forward forestland allocation by market mechanism and stipulate sustainable forestmanagement: report of forestland transfer in Sichuan province. For. Econ. 4, 15–22.

Zhang, B.Y., Uusivuori, J., Kuuluvainen, J., 2000b. Impacts of economic reforms on rural forestry in China. For. Policy Econ. 1,27–40.

Zhang, Y., Zhang, D., Schelhas, J., 2005. Small-scale non-industrial private forest ownership in the United States: rationale andimplications for forest management. Silva Fenn. 39 (3), 443–454.