PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 KslelechresIClRC/Dateoue.indd

 

EXPLORING THE EFFICIENCY AND EFFECTIVENESS OF PAYMENTS FOR
ECOSYSTEM SERVICES IN CHINA’S WOLON G NATURE RESERVE

By

XIAODONG CHEN

A DISSERTATION

Submitted to ,
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Fisheries and Wildlife

2010

ABSTRACT

EXPLORING THE EFFICIENCY AND EFFECTIVENESS OF PAYMENTS FOR
ECOSYSTEM SERVICES IN CHINA’S WOLONG NATURE RESERVE

By
Xiaodong Chen

Conservation investments, including payments for ecosystem services (PES),
have been increasingly devoted to protecting and restoring ecosystems. However, the
efficiency and effectiveness of PBS programs depend on the program design, biological,
economic and social conditions, and dynamic trends in population and households. This
dissertation focused on the efficiency and effectiveness of payments for ecosystem
services. My study area is China’s Wolong Nature Reserve for giant pandas where human
interact with the environment under the Grain-to-Green program (GTGP) and the Natural
Forest Conservation program (N FCP). Specific objectives for this dissertation are to: (1)
evaluate the effects of social and economic factors on the reenrollment in the GTGP after
the current contracts expired, (2) target land for enrollment in the GTGP, (3) assess the
impacts of social capital and labor migration on the fuelwood consumption of indigenous
people, and (4) model the impacts of dynamics in population and households on the
effects of the NFCP. This research use interdisciplinary methods and tools for human-
environment interactions in a coupled human and natural system (CHANS).

I used stated choice method with main effects design to measure the effects of
conservation payment, social norms and program duration on the reenrollment in the
GTGP, controlling for household characteristics and land features. In addition to
conservation payment amounts and program duration, social norms had significant

impacts on program reenrollment. Farming income had a negative effect on program

reenrollment, while income from rural-urban labor migrants had a positive effect on
program reenrollment. I then explored cost-effective targeting of land for enrollment in
the GTGP. Environmental benefits of lands and opportunity costs for land enrollment
were estimated using land features and household characteristics. The efficiency of
investments in a discriminative payment scheme (payments differ according to
opportunity costs) was substantially higher than in a flat payment scheme (same price
paid to all participants). In addition, both optimal targeting and suboptimal targeting
achieved substantially more environmental benefits than random selection of land.

To assess the impacts of social capital and labor migration on the fuelwood
consumption, I used propensity score techniques. Results suggested that social capital in
the form of weak social ties to people in urban settings had significant impacts on rural-
urban labor migration. Following the chain of capital substitutions, labor migration
increased household income, which in turn reduced fuelwood consumption. Simulation
results from the systems model suggested substantial panda habitat can be obtained from
both cash payment and electricity payment scenarios. However, electricity payment, as a
more direct payment approach for reducing human impacts, can improve the efficiency of
conservation investments. The effects of conservation investments are non-linear due to
increases in human population and number of households. In addition, policy effects can
be uncertain due to uncertainties in the behavior of new households that have not been
included in the NF CP. The approach and methods in this research may also be applied in

many other CHANS.

Copyright by
Xiaodong Chen
201 O

ACKNOWLEDGMENTS

First, I thank indigenous people in China’s Wolong Nature Reserve for sharing
their insights and life experiences. Without their generous assistance this effort would
have been impossible.

I would like to thank my committee Sandra Batie, Ken Frank, and Ashton
Shortridge for providing guidance and support throughout my doctoral study at Michigan
State University. I thank Ken in particular for providing me opportunities to attend
cutting-edge workshops that have broadened my knowledge in environmental sociology.
I would like to offer special thanks to Frank Lupi and Tom Dietz for their continuing
efforts as mentors and collaborators. My doctoral experience would not have been
successful without these advisors.

I thank Jack Liu for providing me more opportunities than I could possibly take
advantage of. Jack’s own interdisciplinary scholarship, his collaborations with leading
scholars from several academic fields, and his interdisciplinary team made this
interdisciplinary work possible. He set an academic standard far higher than I thought
possible and then helped me achieve it. Thank you, Jack, for all your help, patience and
support.

I thank Shiqiang Zhou, Mingchong Liu, J inyan Huang, Weihong Tan, Shumin
Fan, Bing Yang, Kai Yang, Jian Yang, Yinchun Tan, Lun Wang, Xiaoping Zhou and
Hemin Zhang for their collaboration during fieldwork, and Bill McConnell, Zhiyun
Ouyang, Bill Taylor and J iaguo Qi for their support and help. I would like to thank
Andre's Vifia for his tireless help on remote sensing issues, and Li An, Marc Linderman,

Guangming He and Scott Bearer for their excellent research in Wolong that laid a

foundation for this effort. Thanks to all the members of the Center for Systems
Integration and Sustainability for their constructive inputs.

Finally, I thank my family and friends. Thanks to my parents, sisters and in laws
for their unconditional love and support. Thanks to Chuntian and all my friends including
“Keke bang” who enriched the social side of my doctoral study.

I would like to gratefully acknowledge the financial support from the National
Science Foundation, the National Aeronautics and Space Administration, the National
Institute of Health, the Michigan State University Environmental Research Initiative,
Michigan Agricultural Experimental Station, the Environmental Science and Policy

Program and Department of Fisheries and Wildlife at Michigan State University.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES ............................................................................................................. x
CHAPTER 1
INTRODUCTION ............................................................................................................... 1
1.1 BACKGROUND ...................................................................................................... 2
1.1.1 PES programs around the world ........................................................................ 4
1.1.2 PES programs in China ...................................................................................... 7
1.2 RESEARCH OBJECTIVES ................................................................................... 10
1.3 STUDY AREA ....................................................................................................... 11
CHAPTER 2
LINKING SOCIAL NORMS TO EFFICIENT CONSERVATION INVESTMENT IN
PAYMENTS FOR ECOSYSTEM SERVICES ................................................................ 16
2.1 INTRODUCTION .................................................................................................. 17
2.2 METHODS ............................................................................................................. 21
2.2.1 Household surveys ........................................................................................... 21
2.2.2 Stated choice .................................................................................................... 23
2.2.3 Econometric model .......................................................................................... 25
2.3 RESULTS ............................................................................................................... 27
2.3.1 Effects of social norms and conservation payment on re—enrollment .............. 27
2.3.2 Effects of program durations on re-enrollment ................................................ 29
2.3.3 Effects of household economic and demographic conditions .......................... 29
2.3.4 Effects of land plot features and respondent’s characteristics ......................... 30
2.4 DISCUSSION ......................................................................................................... 32
CHAPTER 3

USING COST-EFFECTIVE TARGETING TO ENHANCE THE EFFICIENCY OF
CONSERVATION INVESTMENT IN PAYMENTS FOR ECOSYSTEM SERVICES 36

3.1 INTRODUCTION .................................................................................................. 38
3.2 METHODS ............................................................................................................. 42
3.2.1 Modeling strategy ............................................................................................ 42
3.2.2 Household survey ............................................................................................. 43
3.2.3 GTGP land identification ................................................................................. 44
3.2.4 Quantification of environmental benefits ........................................................ 46
3.2.5 Opportunity-cost estimation ............................................................................. 49
3.2.6 Environmental benefits targeting approaches .................................................. 50
3.3 RESULTS ............................................................................................................... 52
3.3.1 Effects of household characteristics and plot features ..................................... 52
3.3.2 Cost-effective targeting of land for environmental benefits ............................ 53
3.4 DISCUSSION ......................................................................................................... 54

vii

CHAPTER 4
WEAK TIES, LABOR MIGRATION, AND ENVIRONMENTAL IMPACTS:

TOWARDS A SOCIOLOGY OF SUSTAINABILITY .................................................... 64
4.1 INTRODUCTION .................................................................................................. 65
4.1.1 Sustainability in a transforming economy ....................................................... 67
4.1.2 Environmental change in contemporary China ................................................ 69
4.1.3 Labor migration patterns in China ................................................................... 72
4.2 METHODS ............................................................................................................. 74
4.2.1 Household surveys ........................................................................................... 74
4.2.2 Causal inference ............................................................................................... 76
4.2.3 Analytical approach ......................................................................................... 80
4.2.4 Quantifying the robustness of the inference .................................................... 83
4.3 RESULTS ............................................................................................................... 85
4.3.1 Data summary .................................................................................................. 85
4.3.2 Determinants of labor migration ...................................................................... 86
4.3.3 Balancing covariates using the propensity score weights ................................ 89
4.3.4 Estimation of the effects of labor migration on fuelwood consumption .......... 89
4.3.5 The robustness of the inference ....................................................................... 91
4.4 CONCLUSIONS ..................................................................................................... 92
CHAPTER 5
MODELING THE EFFECTS OF PAYMENTS FOR ECOSYSTEM SERVICES IN A
COUPLED HUMAN-NATURE SYSTEM .................................................................... 104
5.1 INTRODUCTION ................................................................................................ 106
5.2 METHODS ........................................................................................................... 109
5.2.1 Model summary ............................................................................................. 109
5.2.2 Demographic submodel ................................................................................. 110
5.2.3 Policy submodel ............................................................................................. 111
5.2.4 Landscape submodel ...................................................................................... 113
5.2.5 Model validation ............................................................................................ 117
5.2.6 Simulation experiments ................................................................................. 118
5.3 RESULTS ............................................................................................................. 119
5.3.1 Model validation ............................................................................................ 119
5.3.2 Simulation experiments ................................................................................. 120
5.4 DISCUSSION ....................................................................................................... 121
CHAPTER 6
CONCLUSIONS ............................................................................................................. 136
LITERATURE CITED .................................................................................................... 143

viii

LIST OF TABLES

Table 2.1. Estimation of policy attributes and other characteristics and their marginal
effects on the program re-enrollment ................................................................................. 34

Table 3.1. Pooled logit estimation of conversion of Grain-to-Green program plots to
agriculture after contract expiration ................................................................................... 57

Table 3.2. Pooled logit estimation of re—enrollment of Grain-to-Green program plots after

expiration of current contract ............................................................................................. 58
Table 4.1. Descriptive statistics of individual level variables ........................................... 96
Table 4.2. Descriptive statistics of household level variables ........................................... 97
Table 4.3. Determinants of labor migration models .......................................................... 98

Table 4.4. Testing for balance between migration group and non-migration group with

and without propensity weighting ...................................................................................... 99
Table 4.5. Estimated effect of labor migration on fuelwood consumption (kilograms)
using general linear models (GLM) ................................................................................. 100
Table 5.1. Pooled Ordinary Least Squares of fuelwood use pattern ................................ 124
Table 5.2. Pooled logit estimation of forest loss .............................................................. 125
Table 5.3. Pooled logit estimation of forest recovery ...................................................... 126
Table 5.4. Comparison of model predictions of panda habitat, population size and
household number to observed values ............................................................................. 127
Table 5.5. Sensitivity tests for selected model parameters .............................................. 128

ix

LIST OF FIGURES

Figure 1.1. Locations and elevations of Wolong Nature Reserve, China, and households
in the reserve ...................................................................................................................... 15

Figure 2.1. Estimated program re-enrollment under different levels of payment and
neighbors’ reconversion behavior ...................................................................................... 35

Figure 3.1. Probability of land being enrolled in the Grain-to-Green program ................. 59

Figure 3.2. Percentage of soil benefit obtained with different benefit targets and budgets
using a discriminative payment scheme ............................................................................ 60

Figure 3.3. Percentage of habitat benefit obtained with different benefit targets and
budgets using a discriminative payment scheme. .............................................................. 61

Figure 3.4. Percentage of land benefit obtained with different benefit targets and budgets
using a discriminative payment scheme. ........................................................................... 62

Figure 3.5. Percentage of land benefit obtained with different budgets using a flat

payment scheme. Only the land-benefit (amount of land area enrolled in the PES program)
curve is shown because the curves for soil benefits and habitat benefits are almost
identical to the land-benefit curve. .................................................................................... 63

Figure 4.1. The effect of social capital on the environment as mediated by labor migration.
......................................................................................................................................... 101

Figure 4.2. Overlap in propensity scores between treatment group and control group. .. 102

Figure 4.3. The impact of a confounding variable on a regression coefficient. .............. 103
Figure 5.1. Conceptual framework of the model. ............................................................ 129
Figure 5.2. Panda habitat and household distribution in 2007. ....................................... 130

Figure 5.3. Panda habitat and household distribution in 2020 (a) and 2030 (b) under no-
payment scenario from one run. ...................................................................................... 131

Figure 5.4. Panda habitat and household distribution in 2020 (a) and 2030 (b) under cash
payment scenario from one run. ...................................................................................... 132

Figure 5.5. Panda habitat and household distribution in 2020 (a) and 2030 (b) under
electricity payment scenario from one run. ..................................................................... 133

Figure 5.6. Predicted panda habitat under cash payment, electricity payment, and no
payment scenarios. We did not draw confidence intervals because standard deviations
from 30 runs are small. .................................................................................................... 134

Figure 5.7. Predicted panda habitat under different fuelwood use patterns followed by
new households (households formed after 2001). ........................................................... 135

xi

CHAPTER 1

INTRODUCTION

1.1 BACKGROUND

Much of the world’s natural land cover has been transformed by human
activities (Foley et a1. 2005, Morton et al. 2006), resulting in ecosystem degradation
and biodiversity loss worldwide (Vitousek et al. 1997, Green et al. 2005). Human
alteration of land cover is not limited to human-dominated areas, as it is also very
common in many of the world’s protected areas, such as nature reserves (Liu et al.
2001 , Curran et al. 2004). Interventions have been used to counter this trend through
development activities such as stimulating community economies (e. g., ecotourism),
encouraging community-based natural resources management, providing social
benefits (e.g., education), and redirecting labor and capital from activities that harm
ecosystems. These indirect approaches have been referred to as “conservation by
distraction” (Ferraro 2001, Ferraro and Simpson 2002). Although billions of dollars
have been invested by governments, private sectors, and conservation non-
government organizations (N GOs) through these approaches, the deterioration of
ecosystems continues (James et al. 2001, Ferraro and Kiss 2002, Feamside 2005).

One approach to improve the efficiency of conservation investments is
through Payments for Ecosystem Services (PES) that provide incentives directly to
ecosystem services providers to undertake actions for desired environmental benefits
(Ferraro and Kiss 2002, Wunder 2007, Jack et al. 2008). In recent years, PES
programs, including land set aside and forestry contracting, have been implemented
in many countries (Bennett 2008, Claassen et al. 2008, Wunder 2008). Since current
conservation investments are far below the requirements for conserving ecosystems

globally (James et al. 1999, James et al. 2001), the efficiency and effectiveness of

conservation investments in these PES programs have been a great concern to
conservation practitioners.

From the perspective of policy design, many PES programs are short-term,
compared to outright purchases, with uncertainties in land use after the programs end
(Claassen et a1. 2008). One approach to sustain the conservation gains from PES
programs is through continued conservation investments. In addition, allocation of
scarce conservation fiinds is critically important. Previous studies suggested that the
efficiency of conservation investments is affected by heterogeneous biological values,
demographic, political and socioeconomic conditions (O'Connor et al. 2003, Polasky
et al. 2005, Wilson et al. 2006). Based on biological values and human impacts, 25
regions have been identified as global conservation priority (Myers et a1. 2000).
However, heterogeneous social contexts (e.g., social norms) have often been
neglected when allocating conservation resources. Efficient conservation investments,
therefore, need take heterogeneous human and environmental factors into
consideration. Sustainable conservation gains can be achieved only if participants
would maintain these conservation gains even after PES programs end (Uchida et al.
2005). For example, rural-urban labor migration may not only lower farmers’
dependence on the ecosystems that are targeted for conservation, but also reduce
human impacts on the places of origin. Moreover, both human and natural systems
are not isolated. The effectiveness of conservation investments may largely depend on
the dynamic interactions among human and natural components.

Studies on the efficiency and effectiveness of PES programs need to integrate

human society with natural environment and use methods from both social and

natural sciences due to complex interactions among components in coupled human
and natural systems (Liu et al. 2007a, Liu et al. 2007b). This dissertation attempts to
address some of these critical issues using China’s two PES programs in Wolong
Nature Reserve. Understanding of these issues will not only help improve the
efficiency and effectiveness of existing PES programs, but also will have profound
implications for the allocation, design, and implementation of future conservation
investments such as the United Nation’s collaborative initiative on Reducing
Emissions from Deforestation and forest Degradation (REDD) in developing
countries.
1.1.1 PES programs around the world

The most famous PES program is probably the Conservation Reserve Program
(CRP) in the United States. The CRP is a subtitle of the Conservation Title of the
Food Security Act of 1985. The main objective of the CRP was to reduce soil erosion
caused by agricultural production, with secondary objectives of creating wildlife
habitat, improving water quality, controlling crop supply, and transferring income to
farmers (Smith 1995, Johnson et al. 1997). Enrolled farmers receive conservation
payments for converting highly erodible or environmentally sensitive cropland to
grass, trees, or other conservation uses through a contract that typically lasts 10 years.
From 1986 to 1992, about 36.4 million acres of cropland were enrolled in the CRP at
an average annual payment of $50 per acre (Skaggs et al. 1994, Cooper and Osborn
1998), which resulted in a cost of over $1 billion per year (Parks and Schorr 1997).
By the end of 2005, 35.9 million acres of land was enrolled in the CRP with an

annual cost of approximately $1.8 billion (Claassen et al. 2008).

Some other developed countries have also implemented ambitious PES
programs (OECD 1997). The Permanent Cover Program (PCP) in Canada began in
1989. The PCP was intended to conserve and improve soil productivity by retiring
cropland where annual cultivation was causing long-term soil damage and to generate
benefits for water, wildlife habitat, and landscapes. About 1.3 million acres of
marginal and erodible cropland was converted from grain production to pasture or
forests. In return, enrolled farmers received one-time conservation payments of $15
and $22 per acre for 10-year contracts or $36 and $47 per acre for 21-year contracts
for pasture and forests, respectively. The total cost of the PCP was around $51 million.
The European Union (EU) also has implemented conservation payment programs for
agricultural land conversion. As part of the reforms of the Common Agricultural
Policy, two land-conversion programs were introduced in 1992. The first one is part
of the agri-environmental regulation, a set of policies aimed to promote agricultural
production compatible with protection of the environment. Enrolled farmers receive
an annual conservation payment of up to $784 per ha for setting aside agricultural
land for at least 20 years to prevent soil erosion and improve water quality. Specific
implementations in member countries are different due to the diversity of
environmental conditions and agricultural structures. Another program of the EU is
an afforestation scheme, which pays for afforestation of agricultural land to reduce
the wood shortage in the EU. Enrolled farmers receive a payment covering the cost of
afforestation and new woodland maintenance along with an annual conservation
payment of up to $947 per ha for up to 20 years. By 1997, the afforestation scheme

had converted around 930 000 ha of land at a cost of about $2.6 billion.

In the developing world, Costa Rica’s Pagos de Servicios Ambientales (PSA)
program provides a well-known example of a payment for ecosystem service program.
Since 1997, the PSA has been implemented with three subprograms: reforestation,
sustainable forest management, and forest conservation (Zbinden and Lee 2005,
Sierra and Russman 2006, Pagiola 2008). The reforestation subprogram subsidizes
conversion of cropland to forest for 15 years. Enrolled farmers have to maintain a tree
survival rate of at least 85% to receive a total payment of approximately $550 per ha.
In the sustainable forest management subprogram, only valuable trees beyond a
threshold diameter are allowed to be cut during a contract of 10 years. As
compensation, enrolled forest owners receive a payment of approximately $327 per
ha. Moreover, access roads to forest plots are limited to reduce the disturbance due to
timber harvesting. The forest conservation subprogram rents forest land from owners
for 5 years with a payment of approximately $210 per ha, and enrolled owners were
not allowed to harvest timber or develop the land for other uses (e. g., livestock
breeding) during the contract. By 2001, the PSA had provided conservation payments
to more than 4400 farmers and forest owners, and the total area of land enrolled in the
PSA was more than 284 000 ha, which is about 5.5% of Costa Rica’s national
territory (Zbinden and Lee 2005). Many other ecosystem protection programs have
also recently been launched at regional and national levels, such as the payment for
forest protection programs in Los Negros of Bolivia and Pimarnpiro of Ecuador
(Asquith et a1. 2008, Wunder 2008, Wunder and Alban 2008) and Payment for
Hydrological Environmental Services program (Pago de Servicios Ambientales

Hidrologicos, PSAH) in Mexico (Munoz-Pina et al. 2008).

In addition to their main objectives of converting land cover and protecting
forests (Smith 1995, Zbinden and Lee 2005), many of these programs have also
achieved objectives in conserving/creating wildlife habitat and in restoring
ecosystems (Dunn et al. 1993, Johnson and Schwartz 1993, McMaster and Davis
2001, Sierra and Russman 2006, Asquith et al. 2008). However, there may be
uncertainties in land use afier these programs end. For example, studies of the CRP in
the United States have shown that most of the enrolled land (60% or higher) is likely
to be reconverted to crop production when contracts end (Johnson et a1. 1997, Cooper
and Osborn 1998). Since 1996, expired CRP contract holders could apply for re-
enrollment. Even with continued payments with similar prices, only about 55% of
previous CRP land was re-enrolled in new contracts by the end of 2001 (Claassen et
al. 2008).

1.1.2 PES programs in China

Over the past three decades, China’s economy has grown faster than any
major nations, fueling unprecedented ecosystem degradation that has caused
devastating socioeconomic impacts (Liu and Diamond 2005). For instance, the severe
droughts in 1997 and the major floods in 1998 has been recognized at least partially
as a result of farming on steep slopes and excessive deforestation (World Bank 2001).
To mitigate the impacts of the degraded ecosystems, China has been implementing
two nation-wide PES programs, the Grain-to—Green Program (GTGP, also referred to
as the Sloping Land Conversion Program) and the Natural Forest Conservation
Program (N FCP, also referred to as the Natural Forest Protection Program) (Xu et al.

2006, Liu et al. 2008). The GTGP converts sloping cropland to forest or grassland by

providing participating farmers with conservation payments, whereas the NFCP
protects natural forests through logging bans and afforestation by providing
incentives to forest enterprises and rural communities.

The GTGP has been implemented since 1999. Due to its main objective of
reducing soil erosion by increasing vegetative cover, the criterion for land conversion
in the GTGP is for the slope of cropland in southwestern China to be >25° and
cropland in northwestern China to be >15°. Although cropland with slopes above the
threshold receives priority in enrollment, many cropland plots with slopes below the
threshold can also be enrolled (Uchida et a1. 2005). Participating farmers receive
conservation payment for a maximum of 8 years. The government offers farmers an
annual payment of 2250 kg and 1500 kg of grain or cash payments of 3150 and 2100
yuan per ha (as of June, 2010, 1 USD = 6.8 yuan) of enrolled cropland in the upper
reaches of the Yangtze river basin and in the middle—upper reaches of the Yellow
river basin, respectively. In addition, annual miscellaneous expenses of 300 yuan per
ha and a one-time subsidy of 750 yuan per ha for seeds or seedlings were provided.
By the end of 2006, the GTGP had converted about 9 million ha of cropland (Liu et al.
2008). Studies have shown that the GTGP has substantially improved ecosystem
services such as increased forest cover, reduced water surface runoff and soil erosion,
reduced river sediments and nutrient loss for maintaining soil fertility, and reduced
desertification (Liu et al. 2002, Ma and Fan 2005, Li et al. 2006, Liang et a1. 2006,
Long et al. 2006, Xu et al. 2006, Wang et al. 2007). While these conservation gains

are encouraging, the cost of the GTGP is also tremendous. By the end of 2005, over

90 billion yuan had been invested in the GTGP, and it is expected that the total
investment in the GTGP will reach 220 billion yuan by 2010 (Liu et al. 2008).

The NFCP has been implemented as a pilot program in 1998 with full
implementation since 2000. The aims of the NF CP are to (1) protect and restore
natural forests through logging bans in the upper reaches of the Yangtze river basin
and the middle-upper reaches of the Yellow river basin by 2000 and reduction in
harvesting elsewhere; (2) construct plantation forests through aerial seeding and
artificial planting to increase the capacity for timber harvesting from plantation
forests; (3) create alternative employment for traditional forest enterprises (Zuo 2002a,

Liu et al. 2008). The NFCP planned to reduce timber harvests in natural forests from

32 million m3 in 1997 to 12 million m3 in 2003, and afforest 31 million ha by 2010.

Conservation payments are provided for 10 years. Specific payments for different
conservation actions are 750 yuan/ha for aerial seeding; 1,050 yuan/ha for forest
regeneration through mountain closure; 3,000 and 4,500 yuan/Ira for artificial planting
in the Yangtze and Yellow river basins, respectively; and 10,000 yuan per worker for
protecting 340 ha of forests (Xu et al. 2006). By the end of 2005, about 61 billion
yuan has been invested through the NFCP (Liu et al. 2008). Although the NF CP has
been shown to have produced substantial ecosystem services such as increased carbon
sequestration (Hu and Liu 2006) and reduced soil erosion (Zhang 2006), China’s
timber imports from other countries has increased at least in part due to the NF CP
(State Forestry Administration of China 2005-2007).

Most of the GTGP contracts matured in 2008. To sustain the conservation

gains from the GTGP, the program was extended for another cycle of up to 8 years. It

is expected that the conservation payments under the NFC P will also be extended
when the initial contracts end in 2010. However, another cycle of these programs may
not guarantee the sustainability of their conservation achievements into the future. For
instance, some land enrolled in the GTGP may be reconverted to agriculture when the
payments cease (Uchida et al. 2005). These two programs may provide especially
important opportunities to avoid deforestation and restore degraded ecosystems in
many biologically significant regions such as nature reserves (Loucks et al. 2001).
Given the tremendous investments in these two programs, it is important to evaluate
the efficiency of conservation investments in these programs and the effects of these
programs on ecosystem recovery.
1.2 RESEARCH OBJECTIVES

The overall goal of this dissertation is to explore the efficiency and
effectiveness of PES programs in China’s Wolong Nature Reserve. In Chapter 2, I
evaluated the effects of social norms, together with other household level
socioeconomic factors and land features, on the reenrollment of lands that have been
enrolled in the GTGP after the current contracts end. In Chapter 3, I demonstrated the
environmental benefits that can be obtained through cost-effective targeting of lands
that have been enrolled in the GTGP using household characteristics and land features
under different payment schemes. In addition to the direct effects from the GTGP,
such as increased forest cover, the GTGP may also produce indirect effects on the
environment. Studies on the GTGP found that part of the labor force has been
released fiom agriculture and has boosted the trend of rural-urban labor migration

(Bao et al. 2005, Liu 2005, Ge et al. 2006, Hu et al. 2006, Uchida et al. 2009); hence

10

human population pressure on the ecosystem has been reduced (Liu et al. 2007a). In

Chapter 4, I assessed the impacts of social capital and rural-urban labor migration on

fuelwood consumption. By linking social capital to labor migration and then to the

use of firelwood (one form of natural capital), I examined the substitution among

different forms of capitals for sustainability. For Chapter 5, I developed a spatially

explicit model to study the effects of the NF CP and alternative policy scenarios and

dynamics in population and households on panda habitat. In Chapter 6, I summarized

the results of previous chapters and discuss their implications for policies that aim to

achieving sustainability. Specific objectives include:

1. Linking social norms to efficient conservation investment in payments for
ecosystem services (Chapter 2)

2. Using cost-effective targeting to enhance the efficiency of conservation
investment in payments for ecosystem services (Chapter 3)

3. Understanding the impacts of weak ties and labor migration on the environment
(Chapter 4)

4. Modeling the effects of payments for ecosystem services in a coupled human-
nature system (Chapter 5)

1.3 STUDY AREA

Wolong Nature Reserve was established in 1963 with an area of 200 km2 and

was expanded to 2000 km2 in 1975 (Figure 1.1). Located in southwest China, within

one of the 25 global biodiversity hotspots (Myers et al. 2000), Wolong Nature
Reserve is one of the largest reserves for the protection of endangered giant pandas

(Ailuropoda melanoleuca). The wild panda population in the reserve represents about

11

10% of wild pandas in China, the only country with wild pandas in the world. In
addition to bamboo, which is the pandas’ main diet, conifer and broadleaf forests are
important components of panda habitat by providing shelter and cover (Schaller et al.
1985). Wolong Nature Reserve also provides habitat to more than 6,000 plant and
animal species. As a coupled human and natural system (Liu et al. 2007b), the reserve
is also home to about 4,500 human residents in about 1200 households distributed
between two townships (Wolong and Gengda). People in the reserve engage in
diverse economic activities such as fuelwood collection, deforestation for agricultural
land, road construction, and supporting tourism. Commercial timber harvesting has
been prohibited in the reserve. Previous studies in this reserve have demonstrated
rapid degradation in panda habitat due to these human impacts, especially fuelwood
collection and deforestation for agricultural land (An et al. 2005, Vina et al. 2007).
Almost all of these impacts were within 6 km (Figure 1.1) of households (Linderman
et al. 2005). Moreover, the rapidly increasing human population and even more rapid
increase in the number of households have produced increasing human impacts on the
ecosystem in the reserve (Liu et al. 2003a).

Although enormous amounts of time and energy were needed for fuelwood
collection due to extremely rugged terrain and the difficulty of fuelwood collection
was increasing due to shrinks in forested area, indigenous people in the reserve still
rely much of their energy requirements on fuelwood for cooking and heating. The
reserve administration had limited the amount of fuelwood collection, however, it
was difficult for the administration to monitor and enforce due to the complex terrain

and broad scale of the reserve (Figure 1.1). Electricity, as an alternative to fuelwood,

12

is available in Wolong, but it had been used mainly for lighting and some electronic
appliances (e. g., TV) because electricity was expensive for them to afford, and the
voltage and stability of electricity were not reliable (An et al. 2002). To encourage the
use of electricity as a replacement of fuelwood, the electricity networks in Wolong
were reconstructed in 2001 leading to greatly improved voltage and stability of
electricity. However, these conservation policies and efforts were not effective
without providing conservation payments to indigenous people or substantial
investments in monitoring (Liu et al. 2007a).

The GTGP enrollment took place in Wolong Nature Reserve in 2000, 2001,
and 2003. Farmers were encouraged to enroll their cropland plots with slopes over 25
degrees, but many cropland plots with slopes below 25 degrees were also allowed to
enroll. The GTGP may generate a number of positive impacts for the protection of
panda habitat in Wolong Nature Reserve. The most immediate observable impact, for
instance, is that part of the labor force has been released from agriculture and has
boosted the trend of rural-urban labor migration. In the long run, degraded panda
habitat may recover because the GTGP increased forest cover, which is an important
component of panda habitat (Liu et al. 1999, Liu et a1. 2001). In addition, GTGP land
may generate substantial fuelwood, therefore alleviating further degradation of panda
habitat due to fuelwood collection in natural forests. However, removal of trees for
fuelwood from GTGP land may compromise the GTGP’s potential for restoring
panda habitat.

The NF CP enrollment took place in Wolong Nature Reserve in 2000. All

households that existed in 2000 were enrolled in the NFCP for 10 years. No

13

 

183

..u\

additional enrollment of NF CP has been offered after 2000, and new households
(households formed after 2000) were not included in the program. Under the NFC P
contract, every 5~10 households have been allocated a natural forest parcel to monitor
to prevent illegal harvesting. Illegal harvesting refers to cutting trees in natural forests.
However, people are allowed collect branches of trees near their households. Each
participating household is provided an annual payment of about 850 yuan, which
accounts for 14% of average annual income of households in Wolong in 2001 (He
2008). If illegal timber harvesting is found in a natural forest parcel, the monitoring
households will lose part or all of their NFCP payment of the year depending on how
heavy the parcel gets harvested. Illegal harvesters, if get caught, will lose their NFCP
contracts. Participating households are encouraged to use the payment to purchase

electricity to replace fuelwood for conservation.

l4

A

N
D 6 km Buffer

Elevation (m)

' 1200 - 2000
E “g 2000 - 3000
- 3000 - 4000
- 4000 - 5000
- 5000 - 8200

5 0 5

l

10 km

  

I . Household
Locations

 

, I/ .
I /
/‘.. ’
r .
, , :1 7/ A, - Forest Cover
L. V. - i: in 2007
/ 4 8 km

Figure 1.1. Locations and elevations of Wolong Nature Reserve, China, and

households in the reserve.

CHAPTER 2

LINKING SOCIAL NORMS TO EFFICIENT CONSERVATION

INVESTMENT IN PAYMENTS FOR ECOSYSTEM SERVICES

In collaboration with

Frank Lupi, Guangrning He, and J ianguo Liu

l6

ABSTRACT

An increasing amount of investment has been devoted to protecting and
restoring ecosystem services worldwide. The efficiency of conservation investments,
including payments for ecosystem services (PES), has been found to be affected by
biological, political, economic, demographic, and social factors, but little is known
about the effects of social norms at the neighborhood level. As a first attempt to
quantify the effects of social norms, we studied the effects of a series of possible
factors on people’s intentions of maintaining forest on their Grain-to-Green Program
(GTGP) land plots if the program ends. GTGP is one of the world’s largest PES
programs and plays an important role in global conservation efforts. Our study was
conducted in China’s Wolong Nature Reserve, home to the world-famous endangered
giant pandas and more than 4,500 farmers. We found that in addition to conservation
payment amounts and program duration, social norms at the neighborhood level had
significant impacts on program re—enrollment, suggesting that social norms can be
used to leverage participation to enhance the sustainability of conservation benefits
from PES programs. Moreover, our results demonstrate that economic and
demographic trends also have profound implications for sustainable conservation.
Thus, social norms should be incorporated with economic and demographic trends for
efficient conservation investments.
2.1 INTRODUCTION

Current investments are far below the requirements for conserving ecosystems
globally (James et al. 1999, James et al. 2001). Moreover, most of these investments

are spent within wealthy countries, whereas places with rich biodiversity under threat

l7

are often poor (James et al. 1999, Brooks et a1. 2006). To minimize biodiversity loss
with limited conservation resources, priorities for conservation investments have been
placed on areas where biodiversity and human impacts are highest, e. g., global
biodiversity hotspots (Mittermeier et al. 1998, Myers et al. 2000, Brooks et al. 2006).
However, priority settings based on biological values and threats to these values alone
may not guarantee the efficiency of conservation investments.

Efficient conservation investments need to incorporate biological values with
heterogeneous demographic, political, and socioeconomic conditions (O'Connor et al.
2003, Polasky et al. 2005, Wilson et al. 2006). The high human population and
household density and growth rates in the biodiversity hotspots indicate that human
population is and will remain an important factor in global biodiversity conservation
(Cincotta et al. 2000, Liu et al. 2003a), and the uneven distribution of human
population and households should be considered in conservation investments (Luck et
al. 2004). Political conditions (e. g., political corruption and government stability) in
targeted regions also have a pronounced effect on the efficiency of conservation
investments (Smith et al. 2003). Like human population, per unit area costs of
effective conservation also vary enormously across different places (Balmford et al.
2003). The efficiency of conservation investments can be improved by considering
economic conditions, such as land prices, at global, regional, and local scales (Ando
et al. 1998, Balmford et a1. 2000, Odling—Smee 2005, Arrnsworth et al. 2006).
Although much has been learned about the effects of these socioeconomic factors on
the efficiency of conservation investments (Ando et al. 1998, Balmford et al. 2000,

Cincotta et al. 2000, Balmford et al. 2003, Liu et al. 2003a, O'Connor et al. 2003,

18

Smith et al. 2003, Odling-Smee 2005, Wilson et al. 2006), little is known about the
effects of social norms at the neighborhood level (Ehrlich and Levin 2005).

Social norms are shared understandings of how individual members should
behave in a community under a given circumstance, and members within the
community reward or punish people for their behaviors in following or breaking the
norms (Coleman 1990, Bendor and Swistak 2001). More generally, social norms may
also be sustained by the feelings of reputation and self-esteem through conforming to
social norms, or shame and guilt through detachment from the norms even in the
absence of third-party punishment (Elster 1989, Coleman 1990, Cialdini and
Goldstein 2004). In this chapter, we study social norms in the more general context,
which may be sustained by both self-enforced psychological feelings and/or third-
party enforced punishment. Specifically, we examine when an individual’s behavior
is directly influenced by the behavior of other members in the community, and
substantial change in aggregate behavior of the community can change an
individual’s behavior (Manski 2000, Dietz 2002). Social norms also have been
important in the collective actions of natural resources management (Sethi and
Somanathan 1996, Ostrom 2000, Dietz et al. 2003, Pretty 2003) but have received
little attention in studies of conservation investments.

One approach to conservation investments is through Payments for Ecosystem
Services (PES) (Smith 1995, Daily 1997, Ferraro and Kiss 2002, Zbinden and Lee
2005), such as land set aside and forestry contracting in the United States and
European Union (OECD 1997). In contrast to outright purchase of land or permanent

easements, short-term PES programs may result in only temporary conservation

19

benefits, with uncertainty about land use after the programs end. Past studies have
focused on the program participation of landowners (Langpap 2004, Zbinden and Lee
2005), but much less is known about the impacts of subsequent policies on land use
when a PES program ends. Subsequent PES programs are very important for the
sustainability of conservation benefits from initial PES programs.

Past studies have suggested that post-program land use of farmers who
participated in conservation payment programs can be determined by their
sociodemographic conditions (Johnson et al. 1997, Cooper and Osborn 1998).
Agricultural income has been shown to have positive impacts on the reconversion of
enrolled land (Cooper and Osborn 1998). Furthermore, farmers tend to enroll
marginal land into conservation programs (Zbinden and Lee 2005). In regard to the
characteristics of respondents, older contract holders tend not to reconvert their
enrolled land (Cooper and Osborn 1998).

People’s decisions to participate in a PES program are made in a social
context. Studies indicate that both economic incentives and social norms are
important in an individual’s behavior (Lindbeck 1997) in terms of common resources
management (Levin 2006, Vincent 2007). Individuals whose land-use decisions differ
from the majority in the community may be exposed to social pressures from the
community. Studies of individuals’ participation in PES programs have focused on
the incentives provided by conservation payments (Smith 1995, Cooper and Osborn
1998); little is known about the impacts of social norms at the neighborhood level on
the sustainability of conservation, although substantial conservation benefits (e. g.,

through land enrolled in conservation contracting programs) may be produced with a

20

relatively small change in policy or other exogenous factors due to social norms
(Lindbeck et al. 1999, Nyborg and Rege 2003). To illustrate the impacts of social
norms on the sustainability of conservation, we studied the impacts of subsequent
policies on the land plots that have been enrolled in the Grain-to-Green Program
(GTGP) in China’s Wolong Nature Reserve.

In this chapter, we focus on the re-enrollment intentions of local inhabitants
regarding their GTGP land plots that are likely to be reconverted to agriculture when
the program ends, given different PES policy scenarios following the GTGP. Our
policy scenarios were combinations of three attributes: conservation payment,
program duration, and neighbors’ behavior (the percentage of neighbors reconverting
their enrolled land plots to agriculture). We used stated-choice methods (Louviere et
al. 2000, Naidoo and Adarnowicz 2005) to relate these attributes to the re-enrollment
of those GTGP land plots that are likely to be reconverted when the GTGP ends. In
addition, controls were set for household economic and demographic conditions,
features of the GTGP land plots, as well as characteristics of respondents.

2.2 METHODS
2. 2.1 Household surveys

We conducted household surveys in Wolong from May to August of 2006.
We chose household heads or their spouses as our interviewees because they are
usually the decision makers of household affairs. Our questionnaire was iteratively
pre-tested and revised using qualitative interviews with 54 randomly chosen local
households (Presser et al. 2004). The finalized survey was implemented on a sample

of 321 households, which represent 226.8% of households in the reserve, randomly

21

chosen from the Wolong Household Registration list for 2006. The sample frame
included all households regardless of whether they had enrolled in GTGP. After 5
revisits, 11 households did not have an eligible interviewee and 5 households refused
resulting in 305 respondents and a 95% response rate. Of these 305 households, only
one did not participate in the GTGP and was removed from this study. Similarly high
rates of participation in GTGP (> 85%) have been found in other places in China (Xu
and Cao 2002, Ge et al. 2006, Tao et al. 2006). The elicited information includes
household economic and demographic status, characteristics of enrolled GTGP land
plots, and expected sustainable, annual fuelwood production from that land.
Interviewees were asked if they plan to reconvert each of their GTGP land plots to
crop production if the program ends in 2008 assuming that the prices of crop products
wi l 1 be the same as they were in 2005 and people will be allowed by the government
to reconvert their enrolled plots if they want.

Among the households in our sample, 98 of them (32.2%) planned to
reconvert at least some of their GTGP land plots to crop production after the GTGP
ends and payment ceases. The number of land plots for reconversion (166) accounts
for 22 .6% of a total of 735. This low reconversion rate is not unique. Even lower
planIIed reconversion rates (<20%) have been found in several other places in China
(B 30 et al. 2005, Liu 2005, Ge et al. 2006). Although the reconversion rate is low, the
land plots chosen for reconversion are important to ecosystem services, such as the

connectivity of panda habitat in Wolong, because they are often scattered among

euro 1 1 ed land plots.

22

2. 2.2 Stated choice

Respondents who would reconvert all or some of their enrolled plots if the
program ends were further questioned about their potential actions in the face of
similar PES programs (e. g., extensions of the GTGP). Three contingent behavior
questions were asked about their plans to re-enroll their land plots under different
policy scenarios. Since actual behaviors in response to these scenarios cannot be

observed, we asked respondents’ intentions under these scenarios.

The proposed policy scenarios consisted of three attributes: conservation
payment, program duration, and neighbors’ behaviors. Each of these attributes had
three levels. The amount of annual conservation payment ranged from 100 to 300
yuan/mu with an intermediate value of 200. After the first quarter of the survey, the
high payment level was adjusted to 250 yuan/mu because almost all respondents
would re-enroll all of their GTGP land plots under the annual payment of 300
Wall/mu, and changing the value to 250 yuan/mu allowed more variation in responses
The duration of proposed policy scenarios could be 3, 6, or 10 years. Neighbors were
referred to as households who were located in the same group]. There were 26 groups
Within 6 villages within 2 townships in the reserve containing 21200 households, and
each group contained from 14 to 89 households (Wolong Nature Reserve 2005). We
defined households in the same group as neighbors because our respondents clearly
know who are in the group, and households in the same group tend to have more
interactions among each other, e. g., in collaborative planting and harvesting, which

are important for social norms to be formed and sustained (Elster 1989, Coleman

\

l I
In rural China, a group is a well-defined administrative unit within a village, and a village is an

adm ‘ - . . . . .
lnlstl‘atrve unit wrthrn a township.

 

23

l 990, Manski 2000). For the neighbors’ behaviors, respondents were told that 25%,
50%, or 75% of households in the same group would reconvert part or all of their
enrolled land plots. Therefore, there were 27 possible combinations of attribute levels.
In stated choice models, it is generally impractical and statistically inefficient
to include all possible combinations of attribute levels within an experimental design
(Louviere et al. 2000). Instead, a subset of the attribute combinations that maintains
independent variation among the attributes is usually used in the choice questions. To
understand the main effect of each scenario attribute on the program re-enrollment
choices, we used a main effects design in which each of the three attribute arrays are
orthogonal to one another (Hedayat et al. 1999). Each of the attribute combinations
from the main effects plan then represents one of the “scenarios” presented in the
stated choice question. In this study for each household before the interview, the
scenarios were randomly drawn without replacement from the 9 scenarios from the
main effects plan, and stated-choice methods (Louviere et al. 2000) were used to
query people’s re-enrollment intentions for GTGP land plots under different policy

scenarios,
In the statistical analysis of the stated choice responses, both conservation

Payment and neighbors’ behaviors entered as continuous variables (see econometric
model below). This is common in stated choice models (Louviere et al. 2000) and
allows model-based inferences of respondents’ land use plans at attribute levels other
than the design levels. For instance, given different levels of neighbors’ reconverting

ra . .
te, Conservation program re-enrollment was evaluated across different levels of

24

payment (0 ~ 300 yuan) where all other explanatory variables were set as their mean
values as in Figure 2.1.

As suggested by two very influential theories of intention-behavior
relationship, namely the theories of reasoned action and planned behavior (Fishbein
and Ajzen 1975, Ajzen and Fishbein 1980, Ajzen 1985), intention is very often the
strongest predictor of actual behavior (Madden et al. 1992, Schultz and Oskamp 1996,
Terry and Hogg 1996). These theories specify three conditions that affect the
magnitude of the relationship between intention and behavior: the degree of
correspondence of specificity between the measure of intention and behavior; the
degree of an individual’s volitional control of carrying out the intention; and stability
of intention during the time of measurement and performance of the behavior
(F ishbein and Ajzen 1975, Madden et al. 1992). To improve the consistency between
people’s intentions and actual behaviors, our policy scenarios specify conservation
payment and program duration, which are also attributes of the current GTGP as well
as many other PES programs, and therefore are familiar to respondents. In addition,
We selected household heads or their spouses as our interviewees because they are

usual ly the decision makers for household affairs, and have the most volitional

control over participation in PES programs.

2' 2° 3 Econometric model
We assume that farmers are willing to re-enroll their GTGP land plots in a

renevVed program if the utility of re-enrolling the plot is greater than the utility of the
p101; Without re-enrollment. That is, U,1 > U ,0 , where U 21 and Ui0 are the utilities of

p lot i being re-enrolled and not re-enrolled in the new program, respectively. The

25

utility function U(.) is unobservable; however, there is a probability of re-enrolling

Pr(Yl- =1) = Pr(Ui1 > U 10 ) where Yi=l if planned to re-enroll and 0 otherwise, and a

farmer’s participation plan regarding the plot 1', Y1 , can be observed.

Empirically, the program re-enrollment under different policy scenarios was

modeled with a random-effects probit model (Wooldridge 2002):

Pr(enrolll.jk =1 l Hi’lij’sik’uij) = <D(Hia +PijIB+Sik7 +l‘ij) , (2.1)

where Pr(enrolll.j k =1) is the probability of the ith household enrolling its jth GTGP

land plot under the kth scenario; (1)(.) is the cumulative normal distribution; H i

represents household economic and demographic conditions as well as characteristics

of the respondent associated with the ith household; Pz'j represents the features of the

j th land plot of the ith household; Si k is the kth scenario that household 1' is exposed

to ; a , ,6 , and 7 are parameter vectors associated with household, plot, and policy

scenario factors, respectively; and at]. represents the unobserved random effects
associated with the jth land plot of ith household. We did not find multicollinearity

anion g independent variables. Since our goal was to obtain a relatively accurate
estimation of the effects of independent variables, especially social norms, on the re-
enrollment, the predictive power of the model is less important (Wooldridge 2003).

We used marginal effects to interpret the changes in the probability of re-
enrollment in response to per unit change in explanatory variables. In cases where one
is interested in the percentage changes in outcomes in responses to a percentage

charlge in explanatory variables, elasticity should be estimated. In the probit model,

26

the marginal effects of continuous variables are obtained from the formula (Greene
2003):

(3‘ Pr(enroll == 1)
6X

 

= ¢(Xfl)fl . (2.2)

where X represents all model variables; ¢(.) is the standard normal density function;
and the derivative is calculated at the mean of the explanatory variables. The marginal
effect for a dummy variable (d) is given by

Pr(enroll =1 I 7r(d),d = l) — Pr(enroll =1 l f(d),d = O) , (2.3)
where 7c(d) represents the means of all other variables in the model.

2.3 RESULTS
2. 3.1 Effects of social norms and conservation payment on re-enrollment

Both social norms and conservation payments had significant impacts on the
respondents’ intentions of re-enrolling their GTGP land plots in PES programs (Table
2.1). It was estimated that an additional 10% of neighbors’ reconverting at least part
of their GTGP land plots to agriculture reduced the respondents’ intentions of re-
enrollment by 6.4% on average. In other words, people’s re-enrollment intentions can
be affected by the re-enrollment decisions of their neighbors and tend to conform to
the majority. With a decreasing proportion of neighbors’ reconverting at least part of
their GTGP land plots to agriculture, an individual’s probability of program
participation will increase. For instance, with an annual payment of 200 yuan/mu,
25% more land plots will be re-enrolled if the percentage of neighbors reconverting
their GTGP land plots is changed from 75% to 25%.

The proposition of higher conservation payments increased the number of

land plots intended for re-enrollment. Specifically, an additional yuan in the payment

27

will increase the probability of re—enrolling in the PES program by 0.8%. Among the
GTGP land plots that are likely to be reconverted to agriculture when the GTGP ends,
more than half can be prevented from being reconverted under a PES program
offering an annual payment of 200 yuan/mu. If the current GTGP can be renewed
with the same payment (250 yuan/mu), more than 90% of GTGP land plots could be
saved fi'om reconversion. This finding is quite different from that in studies of the
Conservation Reserve Program (CRP) in the United States, where maintaining
enrolled land was much more expensive than the original cost (Cooper and Osborn
1998). Compared to the land set aside in the CRP, the GTGP land plots have high
costs of reconversion due to reforestation in the land plots. Moreover, the GTGP land
plots may provide additional ecosystem services, such as fuelwood production, to
participants.

Intentions of re-enrollment were also influenced by the interactions between
the conservation payment and neighbors’ re-enrollment behavior (Figure 2.1). For
instance, offering an annual payment of 200 yuan/mu with 75% of neighbors’
reconverting at least part of their GTGP land plots had similar effects on the total re-
enrollment as offering an annual payment of 158 yuan/mu with only 25% of
neighbors’ reconverting their GTGP land plots. Re-enrollment of 50% of land plots
that will be reconverted when the GTGP ends would require an annual conservation
payment of 184 yuan/mu or 142 yuan/mu if 75% or 25% of local residents were to
reconvert at least part of their GTGP land, respectively. If the cost of program re-
enrollment over multiple years and across all involved regions is considered, the

differences in conservation cost under different social norms are substantial.

28

The impact of social norms on program re-enrollment was nonlinear across
different levels of conservation payments. Social norms had the largest impact on the
re-enrollment rate when the payment was intermediate, whereas the effects of social
norms were smallest with the highest and lowest payments, where almost all or none
of the respondents would participate (Figure 2.1).

2. 3.2 Effects of program durations on re-enrollment

Program durations also had nonlinear effects on re-enrollment. As shown in
Table 2.1, a 3-year program re-enrolled 23% fewer GTGP land plots than a 6-year
program. However, re-enrollment for a 10-year program was not significantly
different from a 6-year program (Table 2.1). Presumably, farmers made tradeoffs
among stability, total payment, risks, and flexibility. Compared to short-term
programs, longer-term programs provide more stable income and larger cumulative
payment, but also bring more risks and less flexibility by limiting farmers’ ability to
adapt to changing conditions in markets of crop products.

2.3.3 Effects of household economic and demographic conditions

We found that sources of household income had different effects on program
re-enrollment (Table 2.1). Farming income had a significant, negative effect on
people’s re-enrollment intentions. It was estimated that 1,000 more yuan of farming
income reduced the probability of re-enrollment by 2.9%. However, income from off-
farrn employment outside of Wolong significantly increased the number of GTGP
land plots to be re-enrolled in the PES program: 1,000 more yuan of income from
employment outside of Wolong increased the probability of re-enrollment by 9.7%,

whereas the incomes from off-farm employment within Wolong (tourism

29

employment, temporary off-farm employment, and permanent employment) did not
have such an effect. Although there are conflicts of time allocation between off-farm
employment and farming, off-farm employment within Wolong is much more
flexible in terms of time allocation compared to off-farm employment outside of
Wolong, and therefore cause less conflicts with farming. Thus, not all off-farm
income may increase the participation in PES programs, and different types of off-
farrn employment should be treated differently.

Households with more cropland tended to re-enroll their GTGP land plots in
the PES program (Table 2.1) because GTGP land plots are usually marginal for
growing crops, and people would not reconvert them to agriculture as long as they
already have adequate land for farming. One extra mu of cropland increased the
probability of re-enrollment by 13.8% (Table 2.1). In contrast to other studies
(Cooper and Osborn 1998), livestock breeding did not affect people’s re-enrollment
intentions. Moreover, no effects of household size and total area of GTGP land plots
on program re-enrollment were found (Table 2.1).

2.3.4 Effects of land plot features and respondent’s characteristics

The respondents’ perception of fuelwood that can be sustainably produced2 by
land plots had a positive effect on the program re-enrollment. Fuelwood is one of the
most important energy sources for local people in Wolong (An et al. 2002). Since
fuelwood collection will be allowed in the mature GTGP land, the prospect of more
fuelwood production can increase the number of land plots to be re-enrolled. An

expectation that the land plot will annually produce an additional 10 kg of fuelwood

 

2 .
The amount of fuelwood that can be generated in land plots in the long run was estimated by
respondents based on their past experiences of fuelwood collection.

30

in the long run increased the probability of re-enrolling the land plot by 1% (Table
2.1). Among households, the average distance from each household to its land plots
had a negative effect on the program re-enrollment, probably because the average
distance was correlated to some unmeasured variables, such as social status, of
households. But within a household, the deviation of plot-household distance from
the average distance (difference between each plot-household distance and the
average distance of each household) was not significant in determining the GTGP
land plots to be re-enrolled (Table 2.1). No effects of other plot features on the re-
enrollment were found (Table 2.1).

For the characteristics of respondents, older people were more likely to re-
enroll their GTGP land plots. One additional year of a respondent’s age increased the
probability of re—enrollment by 3.0% (Table 2.1). Since farming and reconverting
GTGP land plots to agriculture are labor intensive, re-enrolling these land plots in the
PES program would be a convenient way for older people to reduce labor demand
(N agubadi et al. 1996, Zbinden and Lee 2005). Respondents’ gender also affected the
program re-enrollment. Male respondents were 30.0% less likely to re-enroll their
GTGP land plots than female respondents (Table 2.1). Combining gender effects with
a respondent’s age, on average a 50-year-old man had the same likelihood of re-
enrollment as a 40-year-old woman. A respondent’s education level was not found to

affect program re—enrollment (Table 2.1).

31

2.4 DISCUSSION

Our findings suggested that the aggregate impacts of social norms at the
neighborhood level on the cost of PBS programs can be substantial. If most people in
a community were to enroll their land in a conservation payment program, the extra
cost for conserving an additional unit of land would be low due to social norms. Even
in communities where most people would initially not participate in a PES program,
social norms can be leveraged with increased conservation investments toward
participation. Thus, incremental cost of conserving an additional unit of land can be
reduced when social norms are leveraged.

A sustainable gain from PES programs can be achieved only if participants are
willing to maintain conservation benefits, even after programs end (Uchida et al.
2005). As an alternative source of income to farming, off-farm employment through
rural-to-urban labor migration not only lowersfarmers’ dependence on the enrolled
land, but also reduces their ecological impacts (Liu et al. 2007a). Numerous off-farm
employment opportunities have been generated by the transitional economy in urban
areas of China (Yang 2000, Li and Zahniser 2002) and many other developing
countries (Korinek et al. 2005). The trend of rural-to-urban migration is expected to
continue over the next several decades (United Nations 2004). These labor and
income trends provide a great opportunity for PES programs to lower costs and
sustain conservation.

Economists have recognized that individuals’ preferences over alternatives
may depend on the actions of others (Manski 2000), suggesting that not only

economic incentives but also social norms may be analyzed by means of utility theory

32

(Lindbeck 1997). Observed outcome data typically have limited power to distinguish
the inference of social norms from other processes (Manski 2000). With the main-
effects design of our stated-choice model, however, the inference of social norms can
be relatively easily distinguished from the effects of other factors.

In conclusion, the results of our study suggest that the efficiency of
conservation investments can be improved by integrating social norms at the
neighborhood level with demographic trends, economic conditions, and biological

values.

33

Table 2.1. Estimation of policy attributes and other characteristics and their marginal
effects on the program re-enrollment.

 

 

 

 

 

 

 

 

 

Independent Variables Parameters SE Marginal
Effects
Social norms and Neighbors’ behavior -1 .662*** 0.581 -0.636
conservation payment Conservation pflment (flran) 0020*” 0.003 0.008
3-year duration -0.598** 0.277 -0.230
Program durations (dummy, reference = 6 years)
lO-year duration -0.270 0.28 1 -0.104
(dummy, reference = 6 years)
Farming income (1000 yuan) -0.075* 0.042 -0.029
Off-farm income (1000 yuan)
- Labor migration to outside of 0.253“ 0.127 0.097
Household economic Wolong
and demographic - Tourism employment in Wolong 0.046 0.071 0.018
conditions - Temporary employment in Wolong 0.063 0.062 0.024
- Permanent employment in Wolong 0.054 0.047 0.021
Cropland after GTGP (mu) 0.361 *** 0.127 0.138
Livestock (dummy) 0.406 0.520 0.157
Household size -0.127 0.176 -0.049
Total land enrolled in GTGP (mu) 0.025 0.085 0.010
Area of land plot (mu) -0.1 10 0.246 -0.042
Fuelwood production (kg) 0003* 0.002 0.001
Average walking distance from each -0.038*** 0.012 -0.015
household to its land plots (minutes)
Deviation of plot-household distance 0.015 0.013 0.006
Land 1310t features from the average distance (minutes)
Elevation ( 1000 m ASL.) 0.050 2.099 0.019
Slope (degrees) -0.038 0.024 -0.015
Aspect -0.007 0.006 -0.003
(180 = north-facing; 0 = south-facing)
Labor cost of reconversion 0.002 0.003 0.001
(persons*days)
Geographic location (dummy) 0.498 0.816 0.185
Respondent Age (years) 0077*" 0.023 0.030
characteristics Gender (reference = female) -0.841* 0.474 0300
Education (years) 0049 0.072 -0.019
Constant -3.812 4.902

 

Significance: * p 50.1; ** p $0.05; *** p $0.01.
Observations = 498; Number of plots = 166; Log Likelihood = -219.209
Significant parameters for a,“ = 1.836 (p < 0.01) and p = 0.771 (p < 0.01) suggest

the random-effects model is appropriate, and the test statistic 12 =80.59 (p < 0.01)
indicates the random-effects model is preferred to the model without random effects.

34

100 "

   
     

 

 

l - ”TIN—ojfghb‘dfs' iieco—rprgve—rting_r_afltem—= 75h
— - Neighbors' reconverting rate = 50% ,
A ., —N°_i9hb°.r§' rfiowmmreteflfk ’
.\° 75 r * * , "
E / , '
2 x x
= 50 , ’
2 .
i
a:
25 r
o a " - t r n—
0 50 100 150 200 250 300

Payment (Yuan/mu)

Figure 2.1. Estimated program re-enrollment under different levels of payment and
neighbors’ reconversion behavior.

35

CHAPTER 3

USING COST-EFFECTIVE TARGETING TO ENHANCE THE EFFICIENCY
OF CONSERVATION INVESTMENT IN PAYMENTS FOR ECOSYSTEM

SERVICES

In collaboration with

Frank Lupi, Andres Vifia, Guangrning He, and Jianguo Liu

36

ABSTRACT

Ecosystem services are being protected and restored worldwide through
payments for ecosystem services. The efficiency of such investments depends on the
design of payment scheme. Land features have been used to measure the
environmental benefits of and cost for land enrollment in cost-effective targeting of
land obtained through payments for ecosystem services. Household characteristics of
program participants, however, may also be important in the targeting of land for
enrollment. We used the characteristics of households participating in China’s Grain-
to-Green program and features of enrolled land to examine the targeting of land
enrollment in that program in Wolong Nature Reserve. We compared levels of
environmental benefits that can be obtained through cost-effective targeting of land
enrollment for different types of benefits under different payment schemes. The
efficiency of investments in a discriminative payment scheme (payments differ
according to opportunity costs, i.e. landholders’ costs of forgoing alternative uses of
land) was substantially higher than in a flat payment scheme (same price paid to all
participants). Both optimal targeting and suboptimal targeting of land enrollment for
environmental benefits achieved substantially more environmental benefits than
random selection of land for enrollment. Our results suggest that cost-effective
targeting of land using discriminative conservation payments can substantially
improve the efficiency of investments in the Grain-to-Green program and other

payment for ecosystem services programs.

37

3.1 INTRODUCTION

Conservation programs in which landholders are paid to alter land
management to achieve environmental benefits have been implemented in many
countries (OECD 1997, Wunder 2008). These programs have reduced soil and wind
erosion (Osborn et al. 1993), restored desirable attributes of ecosystems (Sierra and
Russman 2006), and maintained habitat for native plants and animals (Johnson and
Schwartz 1993, McMaster and Davis 2001). We refer to these desired changes or
maintenance as environmental benefits. The efficiency of the investment in payment
for such environmental benefits, often called payments for ecosystem services (PES),
however, depends on the program’s design.

To induce landholders to participate in PBS programs, incentives should be
greater than the cost of forgoing certain other uses of the land (i.e., opportunity costs).
Landholder opportunity costs and the level of environmental benefits a parcel of land
offers will vary. In practice, flat payments (all participants paid the same price) and
discriminative payments (participants paid different prices according to opportunity
costs) have been used in PBS programs (Claassen et al. 2008, Pagiola 2008). At first
glance, flat payments appear equitable because every participant is paid the same
price. However, flat payments are not equitable when landholders bear different
opportunity costs and their lands supply different levels of environmental benefits
(Ferraro 2008). Additionally, discriminative payments where participants are paid
their opportunity costs will cost less than flat payments. As such, society gains more

environmental benefits for any given investment (J ack et al. 2008).

38

To maximize environmental benefits, PES programs must be implemented on
land that provides the desired environmental benefits with the least cost which is
referred to as cost-effective targeting or optimal targeting (Babcock et al. 1996). In a
cost-effective targeting approach, a benefit-to-cost ratio (level of environmental
benefits provided: cost) is used to rank plots of land from high to low. The lands with
the highest benefit-to-cost ratio are enrolled in the PES program first so that a
maximum amount of environmental benefits can be obtained with a fixed budget.

Lands enrolled in PES programs often supply multiple environmental benefits.
Cost-effective targeting for one type of environmental benefit, however, usually does
not maximize the provision of the other types of environmental benefits under a fixed
budget unless the benefits are perfectly and positively correlated (Babcock et al.
1996). Therefore, the targeting approach that is optimal for a given environmental
benefit is usually a suboptimal targeting approach for other types of environmental
benefits (Babcock et al. 1996, Ferraro 2003). Nevertheless, where different types of
environmental benefits are positively correlated, cost-effective targeting for achieving
one environmental benefit will increase the level of other types of environmental
benefits.

The environmental benefits provided by a particular parcel depend on the
biological and physical features of the land and on the landholder’s actions. In many
cases, however, direct measurement of environmental benefits may be impossible or
prohibitively expensive. Other researchers have used site-specific proxies of
environmental benefits as measures of environmental benefits of land within PES

programs. These proxies include a single biological or physical feature of land parcels

39

(Babcock et al. 1997, Siikamaki and Layton 2007) or combinations of biological and
physical features (Babcock et al. 1997, Khanna et al. 2003, Ferraro 2004, Alix-Garcia
et al. 2008).

Opportunity costs of landholders participating in PES programs are often
difficult to measure because they are only known to landholders. However,
landholder’s opportunity costs are often correlated with the location and features of
the land and with household characteristics (Cooper and Osborn 1998). Researchers
have estimated the value of a parcel on the basis of its biological and physical
features (Ferraro 2003, Khanna et al. 2003, Alix-Garcia et al. 2008) as proxies for the
opportunity costs of landholders. Even though households are often the basic unit on
which land-use decisions are based (Liu et al. 2003a), household characteristics of
landholders usually have not been included in determination of opportunity costs
(N aidoo et al. 2006, Siikamaki and Layton 2007). Despite these measurement
difficulties, targeting in PES programs can substantially improve the efficiency of
investments, especially when the level of environmental benefits and the costs to
obtain the benefits are heterogeneous across the parcels within a landscape (Osborn et
al. 1993, Babcock et al. 1996, 1997, Chan et al. 2006).

In actual implementation of PES programs, it may not be feasible to collect
the information on households and land parcels needed to determine opportunity costs

for cost-effectively targeting of lands to enroll. Another approach to enrolling lands in
PES programs is to use competitive auctions in which potential enrollees submit bids
(the payment they require) to provide environmental benefits. The cost-revelation

mechanism in most competitive bidding processes makes auctions a powerful tool for

40

inducing potential participants of PES programs to submit bids equal to their
opportunity costs (Latacz-Lohmann and Van der Hamsvoort 1997).

China is implementing several large-scale conservation programs (Liu et al.
2008). Among these is the Grain-to-Green program (GTGP), which was implemented
in 1999 and is the largest PES program in the developing world. Participating farmers

receive payments in grain or cash for a maximum of 8 years to convert cropland to
forest or grassland. Because the main objective of GTGP is to reduce soil erosion by
increasing natural land cover (forest and grassland), the slope of enrolled land should
be above 15° in northwestern China and above 25° elsewhere. Although croplands
with slopes above the thresholds receive priority for enrollment, some croplands with
slopes lower than the thresholds have been enrolled (Uchida et al. 2005). By the end
of 2006, GTGP had converted about 9 million ha of cropland into forest and
grassland (Liu et a1. 2008). (In the United States, about 14.5 million ha of cropland
are enrolled in a similar program, the Conservation Reserve Program (Claassen et al.
2008) In addition to its main objective of restoring natural vegetation cover, GTGP
aims to generate other environmental benefits, such as restoration of habitat for
certain animals and plants (Zuo 2002b).

The GTGP has only 2 payment levels nationwide which operate as flat
payments within each region. On an annual basis payments are 2250 kg of grain or
3450 yuan per ha of enrolled cropland in the upper reaches of the Yangtze River
basin and 1500 kg of grain or 2400 yuan in the middle-upper reaches of the Yellow
River basin. The different regional payment levels are used in part to account for the

regional differences in opportunity costs of landholders because land in the upper

41

reaches of the Yangtze River basin is usually more productive than in the middle-
upper reaches of the Yellow River basin (Uchida et al. 2005). The payments for most
participating farmers exceed cultivation income from the enrolled land (Uchida et a1.
2009), which indicates similar environmental benefits may be obtained at lower cost.
By the end of 2005, more than 90 billion yuan had been invested in GTGP (Liu et al.
2008). When contracts started expiring in 2008, they were extended for up to 8 years.
In the future the program’s budget is likely to be reduced (Liu et al. 2008). Given its
large scale and heterogeneities in opportunity costs of landholders and environmental
benefits, the cost-effectiveness of GTGP payments may be improved greatly if
payments are made to landholders whose lands can provide environmental benefits at
lowest cost (i.e., cost-effective targeting of lands to be enrolled). We examined the
GTGP in China’s Wolong Nature Reserve to determine the efficiency of investments
made through cost-effective targeting using flat and discriminative payments to
landholders. We used features of specific parcels as proxies of environmental benefits
and physical features of the parcels and household characteristics of landholders to
estimate the opportunity costs of participating in GTGP. The results of our study can
be used to maintain the environmental benefits from GTGP after the expiration of
current contracts with a reduced budget.
3.2 METHODS
3. 2.1 Modeling strategy

To study cost-effectiveness of alternative GTGP targeting and payment
schemes, we modeled enrollment in and environmental benefits from GTGP in

Wolong Nature Reserve. We determined the locations, environmental benefits, and

42

opportunity costs for all GTGP plots in the reserve. Because we did not know GTGP
plot locations of all households, we distributed all GTGP plots across the landscape
through stochastic simulations and then calculated environmental benefits provided
by these plots. We then modeled the conversion to croplands of GTGP plots that were
not re-enrolled in the program after cessation of payments and re-enrollment of plots
in the program. We used these models to identify the enrollment probabilities and
opportunity costs for each GTGP plot. We modeled the environmental benefits
provided by the enrolled GTGP plots under different conservation budgets and
compared cost-effectiveness among the different targeting approaches and payment
schemes.
3.2.2 Household survey

We interviewed heads of households in Wolong Nature Reserve in the
summer of 2006. We used the govemment’s household registration list of 2006 to
randomly select 321 of the 1200 households for interviews. Of those 321, 304 (95%)
completed an interview. For each plot enrolled in GTGP, we collected information on
the landholder’s land-use plans after expiration of their GTGP contact. These plans
were used to estimate the probability a household planned to reconvert their GTGP
plot after the current program ends, P(reconvert). Surveyed landholders planned to
convert 166 (22.6%) of their 735 GTGP plots to crop production after GTGP
payments ceased (Chen et al. 2009a).

For those respondents that planned to reconvert their GTGP plots, stated
choice methods (Louviere et al. 2000) were used to elicit whether they would re-

enroll in a new round of GTGP under various payment amounts, i.e.,

43

P(reenroll j | pay > 0, reconvert). The proposed annual conservation payment had

three levels: 1,500, 3,000, and 4,500 Yuan per ha. After the first quarter of the survey,
the highest level of payment was adjusted to 3,750 Yuan per ha because almost all
respondents would re-enroll all of their GTGP plots under an annual payment of
4,500 Yuan per ha, and changing the value to 3,750 Yuan per ha allowed more
variation in responses. Varying payment prices across scenarios and respondents
allowed us to statistically model re-enrollment as a function of payments, thus to
identify opportunity costs of re-enrollment.

3.2.3 GT GP land identification

For all households in the reserve, we obtained information on characteristics
such as household size and age and gender of the household head from the local
govemment’s 2006 household registration list. The geographic location of each
household in the reserve was recorded in 2006 with global positioning system (GPS)
receivers. Government data for the reserve showed that 2470 plots (total of 367.5 ha)
belonging to 969 households were enrolled in GTGP in 2003.

Although information on the number of plots each household had enrolled in
the program and the area of each plot was available, information on the geographic
location of the plots was not available. Since we required information on the locations
of all plots that were enrolled in the GTGP, we developed a map of the probability
that each grid cell (i.e., pixel) is under the GTGP. The 304 surveyed households
enrolled a total of 735 plots comprising a total of 110.4 ha. The locations of these 735
plots were measured using a GPS receiver. To map the probability that each grid cell

is under the GTGP, we used a fuzzy classification algorithm based on the principle of

44

maximum entropy (J aynes 195 7). The algorithm was applied to multi-spectral and
topographic data in grid format using the software MaxENT (Phillips et al. 2006).
Multispectral data consisted of two Landsat Thematic Mapper (TM) images (28.5m x
28.5m / pixel) acquired on April 19 and September 18, 2007. Topographic data (with
the same pixel resolution as the Landsat TM imagery) consisted of elevation, slope
and aspect derived from a digital elevation model generated for the study area from
topographic maps (Liu et al. 2001). We randomly selected two-thirds of the
geographic locations of the 735 GTGP plots that we measured to calibrate the fuzzy
classification algorithm, and one-third to validate the output map. Although the area
of some GTGP plots is smaller than the area comprised by a Landsat TM pixel, if at
least one GTGP plot fell within a pixel, the entire pixel was considered as a GTGP
plot, and used for model calibration and validation. This constitutes an approximation
since not necessarily 100% of a pixel is under the GTGP, however it is a common
procedure in many pixel-based imagery classification methods (Lu and Weng 2007).
We then resampled the resolution of the GTGP probability map to 10 meters using
nearest-neighbor interpolation so that each GTGP plot occupied at least one pixel.
Since all of these 735 plots were located within 6 km of their corresponding
households, the probability map (Figure 3.1) was developed in a 6 km buffer around
all household locations.

The GTGP probability map was validated by means of a receiver operating
characteristic (ROC) curve (Hanley and Mcneil 1982). The ROC curve is a plot of the
sensitivity values (i.e., true positive fraction) vs. their equivalent l-specificity values

(i.e., false positive fraction) for all possible probability thresholds. The area under the

45

ROC curve (AUC) is a measure of model accuracy, with AUC values ranging from 0
to l, where a score of 1 indicates perfect discrimination, a score of 0.5 implies a
prediction that is not better than random, and lower than 0.5 implies a worse than
random prediction. We used the validation data set (one-third of the 735 GTGP plots
that we measured) together with 10,000 randomly selected pixels (Wiley et al. 2003,
Phillips et al. 2006) for deriving the AUC value. The GTGP probability map
exhibited high accuracy (AUC = 0.98).

We then stochastically distributed all 2470 GTGP plots across the landscape
on the basis of the GTGP probability map and the probability distribution of distances
between the 735 GTGP plots and their corresponding households. For each GTGP
plot, we first randomly chose its distance to its corresponding household based on the
probability distribution of the distances between the 735 GTGP plots and their
corresponding households. We then randomly chose a pixel as the central pixel of the
GTGP plot from all the pixels on the GTGP probability map that are at the specified
distance from the household based on these pixels’ probability of being GTGP land.
Finally, the neighboring pixels, with a positive probability of being GTGP land, of the
central pixel were treated as part of the GTGP plot until the area of the GTGP plot
was reached. The simulation was conducted using Java programming language (J DK
1.4.2, Sun Microsystems).

3.2.4 Quantification of environmental benefits

To examine the different targeting approaches and payment schemes for

GTGP, we constructed three possible indices of environmental benefits. Because

slope was the only available measure of the reduction of soil erosion through GTGP

46

(Uchida et al. 2005), we used plot-specific slope (measured by the mean slope of
pixels in the plot) as a proxy for the environmental benefit of reduction in soil erosion.
The soil-benefit index is the square of the standardized (Ferraro 2004) slope of the
plots:

slopel. —slopemin )2 (3 l)
slopemax —slopemin ’ '

 

soil benefit index i =(

where slopei is the slope of a GTGP plot and slopemi and slopemax are the

n
minimum and maximum slopes among all GTGP plots, respectively. This index
measures the relative steepness of a plot relative to the minimum and maximum

slopes among all GTGP plots in the reserve. The higher the soil-benefit index of the

ith plot (soil benefit index i ), the greater the probability the ith plot will have less soil

erosion if enrolled in the PES scheme than if used to grow crops. Because land with
steeper slopes was given priority for enrollment in GTGP, we used the square of the
standardized slope to place more weight on plots with steeper slopes. It should be
pointed out that slope-based proxies of soil benefit do not represent all the factors,
such as rainfall, soil type, and ground cover, which determine soil benefit. As these
types of information become available, more robust measures of soil benefit, such as
the Revised Universal Soil Loss Equation (RUSLE) (US. Department of Agriculture
2001), may be used.

Besides reducing soil erosion, GTGP also aims to restore habitat for many
plant and animal species. Distance to patches of habitat that existed prior to PES
programs has been used as a measure of habitat quality (Babcock et al. 1996).

However, distance-based proxies of habitat quality do not represent all the factors

47

important in determining habitat quality. We used the distance between GTGP plots
and the nearest patch of natural forest as a measure of the habitat quality of the GTGP
plot. Using protocols described in (Vifia et al. 2007), we determined the distribution
of natural forest (Figure 1.1) in the reserve by classifying remotely sensed imagery
acquired on 18 September 2007. The habitat-benefit index is

drstl- —drstmin 2

habitat benefit index i = (I - (3.2)

 

distmax — dist min
where dist z“ is the distance between a GTGP plot and the nearest natural forest patch,
dist min and dist max are the minimum and maximum distances to the nearest natural

forest patches among all GTGP plots, respectively. Here we used a subtraction from
unity so that a higher index value would correspond to a smaller distance to the
nearest forest patch. Therefore, the higher the habitat-benefit index of the ith plot

(habitat benefit index i ), the higher the habitat quality the ith plot is presumed to have

for certain animals and plants. As with the soil-benefit index, we used the square of
the standardized distance to place more weight on those plots that were closer to
patches of natural forest.

We measured the amount of each type of environmental benefit of a plot by
multiplying the benefit index by the area of the plot. For comparison purposes, we
also measured the amount of land area enrolled in the PES program, defined as land
benefit, when we examined the effectiveness of the different approaches to targeting

environmental benefits and the different payment schemes.

48

3.2.5 Opportunity-cost estimation

Given that GTGP plots were still under contract when our data were collected,
we used landholders’ plans for their GTGP plots after their contracts expired to model
the probability of land holder re-enrollment in GTGP. For those plots for which there
were no plans to convert the land to crops after the contract expired, we assumed the
plots would be re-enrolled under any positive payment for participation. For GTGP
plots landholders planned to convert, there was a probability that the plot would be
re-enrolled if any positive payments were offered. Thus, the probability of a GTGP
plot being re-enrolled is

P(re - enroll j) = 1 — P(convertj) +

, 3.3
P(convertj) * P(re - enroll j | pay > 0, convert) ( )

where P(convertj) is the probability of the jth GTGP plot being converted to crop

production after contract expiration, 1— P(convertj) is the probability the jth GTGP

plot will not be converted to crop production after contract expiration (and thus the

plot will be re-enrolled at any positive payment), and P(re - enroll j I pay > 0, convert)

is the probability of re-enrolling the jth GTGP plot under a new payment program for
plots that will be converted to crop production after contract expiration, which must

then be weighted by the probability that the plot will be converted, P(convertj).

In logistic regression models, we used proposed conservation payments,
features of GTGP plots, and household characteristics to explain the probability a
GTGP plot will be re-enrolled (Eq. 3.3). We corrected for dependencies among plots

of the same landholder and among responses to different proposed alternative

49

payments for the same plot with Huber’s variance correction (Wooldridge 2002). We
did not find multicollinearity among independent variables. We applied these models
to all GTGP plots in the reserve and calculated the probability of each GTGP plot

being re—enrolled ( P(re - enroll j) in Eq. 3.3). We determined the per hectare

opportunity cost of each plot with a Bernoulli trial, which determined re-enrollment
of plots as a function of the payment. The rate parameter of the Bernoulli distribution

was, P(re - enroll j) , and we estimated it for different payment amounts (Cooper and

Osborn 1998). The per hectare opportunity cost of a plot was the payment level at or
above which the plot would be enrolled. The opportunity cost of a plot was the per
hectare opportunity cost of the plot multiplied by its area.
3. 2.6 Environmental benefits targeting approaches

For each of the three types of environmental benefits, we examined the
amount of that environmental benefit that can be obtained with cost-effective
targeting of the lands to enroll in the PES program. We also illustrated, how much of
each type of environmental benefit would be obtained had one of the other two types
of benefits been the target of the PES program (i.e., sub-optimal targeting). We
examined targeting using both flat payment and discriminative payment schemes. We
conducted the initial analysis only on those GTGP plots that would be converted to
crop production after contract expiration. Under the discriminative payment scheme,
we determined the cost-effective enrollment of plots for each type of environmental
benefit by ranking all GTGP plots from high to low according to the benefit that
could be obtained for each unit of cost (i.e., ratio of benefit to cost) and enrolling

plots with the highest benefit-to-cost ratio first. For the land-benefit maximization

50

approach, where the goalis to maximize the area of land enrolled, we based GTGP
plot enrollment on per hectare cost; thus, less expensive GTGP plots had enrollment
priority. In addition to determining cost-effective targeting for each type of
environmental benefit, we also calculated the amount of each environmental benefit
obtained and the amount of land enrolled in GTGP under suboptimal targeting (i.e.
when plot benefit-to-cost ratios are ranked on the basis of the non-targeted
environmental benefits). For instance, maximizing the amount of land enrolled is the
optimal approach for land acquisition, but it is usually suboptimal for acquiring either
of the other environmental benefits. Maximization of soil benefits, however, is the
optimal approach to achieve soil benefits, but it is suboptimal for improving habitat
quality for some species and for land acquisition.

To understand the relation between each environmental benefit and
expenditure, we calculated the total level of an environmental benefit that can be
obtained within a budget that varied from zero to the cost of obtaining all the
environmental benefits possible. Because our spatial distribution of GTGP plots and
enrollment decision were stochastic processes, we calculated the mean values of
environmental benefits from 300 simulations for each targeting approach to facilitate
relatively robust relations between environmental benefits and expenditure. Results
from a total of 400 simulations are almost identical to those from 300 simulations.
We also drew a 45° line (Babcock et al. 1996) in each of the benefit-budget planes to
show the amount of environmental benefit that could be obtained through random

selection of plots constrained within a particular budget.

51

In addition to the discriminative payment scheme, we explored the
environmental benefits obtained through a flat payment scheme. Under the flat
payment scheme, all plots with per hectare opportunity costs less than or equal to the
per hectare payment were enrolled. When all landholders are paid the same flat price
for their plots, each increase in the number of plots enrolled requires that a higher per
hectare payment be made to all plots, not just to the plots with higher opportunity
costs. Thus, all plots that would have enrolled at a lower payment level (because their
opportunity costs were lower) receive a surplus equal to the difference between their
opportunity costs and the amount of the flat payment. The magnitude of this surplus
defines the difference in costs between discriminative and flat payment schemes.

3.3 RESULTS
3. 3.1 Effects of household characteristics and plot features

Household size had significant positive effects on the probability of
conversion of a GTGP plot to cropland (Table 3.1). The more land the household had
enrolled in GTGP, the less likely the household planned to convert any of the plots to
agriculture. In addition, households in Gengda were less likely to have plans to
convert plots than households in Wolong.

The higher the payment the more likely landholders were to participate in
GTGP (Table 3.2). Households with more members were less likely to re—enroll their
plots. Probability of re-enrolling increased as age of household head and area of
cropland increased. The distance between plots and the household reduced the
probability of re—enrollment, perhaps because distance was correlated with some

unmeasured variables such as the household’s social status.

52

3.3.2 Cost-effective targeting of land for environmental benefits

Our simulations showed that about 78% of GTGP land in the reserve would
not be converted to agricultural uses even after the expiration of contracts. We
conducted our analyses of re-enrollment of plots under new payments by including all
plots and by including only those plots that would be converted to agriculture when
contracts expired; in the figures that follow we show the results for the latter (Figure
3.2 — 3.5). The approach that optimized soil benefits (i.e. cost-effective targeting for
soil benefits) obtained 82% of the soil benefits when the budget for payments was
100,000 yuan and 97% of the benefits when the budget was 200,000 yuan (Figure
3.2). Cost-effectively targeting for habitat benefits obtained 81% of the habitat
benefits (Figure 3.3) when the budget for payments was 100,000 yuan, whereas cost-
effectively targeting for land benefits obtained 75% of the land benefits (Figure 3.4)
when the budget was 100,000 yuan.

Even though cost-effective targeting achieved more of the targeted
environmental benefit for any budget amount than when suboptimal targeting was
used, suboptimal approaches were far superior to random selection of plots. In all
cases, differences in the amount of environmental benefits obtained between optimal
and suboptimal targeting approaches were much smaller than differences between any
of the targeting approaches and random selection of plots. When the budget fer
payments was 100,000 yuan (Figure 3.2), cost-effectively targeting for soil benefits
obtained 82% of the soil benefits compared to 76% and 69% of the soil benefit from
the two suboptimal approaches, but only 29% of the soil benefit was obtained when

plots were randomly selected for enrollment (45° line).

53

The amount of environmental benefits obtained with discriminative payments
and flat payments were quite different. It cost 92,000 yuan with discriminative
payments to obtain 80% of soil benefits (Figure 3.2). To obtain the same amount of
soil benefit with flat payments (Figure 3.5), it cost 298,000 yuan. The difference
between the cost of discriminative payments and the cost of flat payments increased
as the percentage of environmental benefits increased. For instance, in terms of land
acquisition, to obtain 30%, 60%, or 90% of the land with flat payments would cost
29,000, 128,000, and 585,000 yuan, which is about 1.7, 2.1, and 3.4 times the cost of
discriminate payments, respectively.

These differences demonstrate how efficiency of investments can be improved
by switching from the most cost-effective flat payment approach to the most cost-
effective discriminative payment approach. Results presented in the graphs illustrate
the effectiveness of targeting specific levels of benefit and assume that no payments
would be made to landholders who did not plan to convert lands to crop production
upon contract expiration. As such, even flat payments are to a small degree
discriminative payments. When we included in the payment scheme the GTGP plots
that would not be converted to crop production after contacts expired, the efficiency
of the payments improved by more than 10 times when we switched from flat
payments to discriminative payments.

3.4 DISCUSSION

Substantially greater environmental benefits were obtained when lands were

optimally or suboptimally targeted for enrollment than when enrollment of land was

random. When suboptimal targeting approaches are used in PES schemes, the

54

efficiency of the program depends on correlations among the types of environmental
benefits (Babcock et al. 1997). When different environmental benefits of plots are
highly and positively correlated, as in our case, similar amounts of environmental
benefits can be obtained with suboptimal targeting as can be obtained with cost-
effective targeting. More generally though, targeting the desired environmental
benefit can be critical to achieving conservation objectives if the environmental
benefits of plots are not highly and positively correlated.

The differences in cost-effectiveness between the payment schemes was
substantially larger than the differences among environmental benefit targets. In all
cases, discriminative payments were more efficient (up to 10 times) than flat
payments. The reason for the difference is that flat payments pay all enrollees the
same price regardless of opportunity costs.

Household characteristics were also significant determinants of opportunity
costs of landholders participating in GTGP. For instance, a plot that has little
agricultural value for a household with a small labor supply can be much more
valuable for a household with a larger labor supply. In addition, we found substantial
regional differences in landholders’ willingness to continue participating in GTGP.
One of the main differences between the 2 townships in our study was that Gengda
was closer to more urbanized regions outside the reserve. Thus, PES programs are
more likely to cost-effectively achieve their objectives if household characteristics
and regional differences, as well as biological and physical features, are incorporated
in the planning of PES programs, especially in areas without robust land markets.

Other household characteristics (e. g., off-farm income) were also significant

55

determinants of opportunity costs (Chen et al. 2009b), but were not included in this
study because such information was not available for all households in the reserve.
Opportunity costs of landholders are typically private information that is not
available to the public, which results in an information gap between landholders and
conservation practitioners (Ferraro 2008). Competitive auctions can reduce this
information gap substantially (Latacz-Lohmann and Van der Hamsvoort 1997).
Moreover, competitive auctions have been applied successfully in some PES
programs and have improved the efficiency of conservation investments (Kirwan et al.
2005, Claassen et al. 2008). Cost-effective targeting for environmental benefits
coupled with competitive auctions could greatly improve the efficiency of
investments in PES programs, especially in programs, such as GTGP, that are
relatively large and have substantial heterogeneities in opportunity costs and
environmental benefits. Competitive auctions and cost-effective targeting may
increase transaction costs of PES programs, but our results suggest that the improved
efficiency from cost-effective targeting will far outweigh likely increases in
transaction costs in GTGP. Although cost-effective targeting using competitive
auctions will likely improve the efficiency of conservation investments in PES, it is
important to maintain low transactional costs. The growing demand for conservation
resources globally (Ferraro 2008, Jack et al. 2008) makes it increasingly important to

improve the efficiency of investments in PES and other conservation programs.

56

Table 3.1. Pooled logit estimation of conversion of Grain-to-Green program plots to

a
agriculture after contract expiration.

 

 

Independent Description Parameters Marginal

variables (robust SE) effects

Household size no. of people in the 0.250* (0.103) 0.039
household

Cropland cropland of the -0.963 (1.022) -0.151
household (ha)

GTGP land land enrolled in GTGP -1.734** (0.633) -0.273
(ha)

Age of household years -0.003 (0.012) -0.001

head

Gender of l,female; 0,male 0.400 (0.376) 0.069

household head

Township 1, Gengda township; -1.182* (0.515) -0.199
0, Wolong township

Area ha 0.015 (1.018) 0.002

Slope degree -0.004 (0.016) -0.001

Elevation 100 m (asl) -0.033 (0.104) -0.005

Distance 100 m -0.050 (0.030) -0.008

Constant 0.396 (2.346)

12 44.41 ***

 

a P(convertj) in Eq. 3.3; number of plots 735.

Significance: * p $0.05; **p $0.01; *** p $0.001.

57

Table 3.2. Pooled logit estimation of re-enrollment of Grain-to-Green program plots

a

after expiration of current contract.

 

 

Independent variables Parameters Marginal
(robust SE) effects
Ln(payment in yuan) 1.816*** (0.300) 0.453
Household size -0.372** (0.141) -0.093
Cropland 3.668** (1.169) 0.914
GTGP land 0.340 (0.839) 0.085
Age of household head 0.034** (0.013) 0.008
Gender of household head -0.330 (0.469) 0082
Township 0.102 (0.521) 0.025
Area 0.994 (1.203) 0.248
Slope 0.016 (0.020) 0.004
Elevation -0.003 (0.134) —0.001
Distance -0.096* (0.041) -0.024
Constant -9.985*** (3.071)
2,2 59.83***

 

a P(re - enroll j I pay > 0, reconvert) in Eq. 3.3; observations 498; number of plots I66.
Significance: * p $0.05; ** p $0.01; *** p 30.001.

58

  

Probability of enrollment in GTGP

' 0.99

0.01

420 4 8km
E

Figure 3.1. Probability of land being enrolled in the Grain-to-Green program.

59

 

 

 

 

 

 

 

 

 

 

100 _ .-__—v./
/ .
80 ~ ‘
Is
I; 60 .
‘8
c
o
2 40- , .
‘2 / x . ' Soil benefit maximization i
20 7 , ' l — — — -Habltat beneflt maximization .
x’ ’ I ------ Land benefit maximization
." —----45°Ilne g
o I L If 7 ‘— f ”It WET—J L _ -7"
O 50 1 00 1 50 200 250 300 350
Budget (thousand yuan)

Figure 3.2. Percentage of soil benefit obtained with different benefit targets and
budgets using a discriminative payment scheme.

60

 

 

 

 

 

 

 

 

 

 

 

 

 

100
80
S?
if; so »
c
o
.D
5 4o . _ _-_ _ _ ,
3 4~ / . ' / f Soil benefit maximization l
E 20 ° , ' — — - -Habitat beneflt maximization i
I ’x x ------ Land benefit maximization .
,/ -----45°llne j
0 ‘ ‘ L‘ 4’ 1 4
0 50 100 150 200 250 300 350
Budget (thousand yuan)

Figure 3.3. Percentage of habitat benefit obtained with different benefit targets and
budgets using a discriminative payment scheme.

61

100

 

 

 

 

 

 

 

 

80 ~
.1;
If: 60 ~
:
3
U 40 t . ' __ gm -fl, ,____
I .
F. / . ' Soil beneflt maximization l
" 20 .’ / — ' l - — - -Habitat benefit maximization ’
I x” ' L ------ Land benefit maximization .
— - - - -45°line 4
o . . . , ,__ . . . ,l
0 50 100 1 50 200 250 300 350
Budget (thousand yuan)

Figure 3.4. Percentage of land benefit obtained with different benefit targets and
budgets using a discriminative payment scheme.

62

100 ”3--.-.. - __ ______

 

 

 

 

80
§
5;; 60 e
c
3
1: 4° ’
c
to
.1
20
o I r . I r I
0 300 600 900 1200 1500 1800 2100

Budget (thousand yuan)

Figure 3.5. Percentage of land benefit obtained with different budgets using a flat
payment scheme. Only the land-benefit (amount of land area enrolled in the PES
program) curve is shown because the curves for soil benefits and habitat benefits are
almost identical to the land-benefit curve.

63

CHAPTER 4

WEAK TIES, LABOR MIGRATION, AND ENVIRONMENTAL IMPACTS:

TOWARDS A SOCIOLOGY OF SUSTAINABILITY

In collaboration with

Kenneth A. Frank, Thomas Dietz, Guangrning He, and Jianguo Liu

64

ABSTRACT

Debate about the substitutability of manufactured, natural, human and social
capital is at the heart of sustainability theory. Sociology can contribute to this debate
by examining the processes and mechanisms by which one form of capital is
substituted for another. We examine the substitution among different forms of
capitals at China’s Wolong Nature Reserve where the consumption of an important
aspect of natural capital, fuelwood, has serious consequences for the local
environment. We found that social capital in the form of weak social ties to people in
urban settings had significant impacts on rural-urban labor migration. Following the
chain of capital substitutions, labor migration then significantly affected income,
which in turn affected fuelwood consumption. Quantification of the validity of
inferences suggests the inferences are robust with respect to concerns about omitted
confounding variables.
4.1 INTRODUCTION

At least since the Bruntland Commission published its historical report “Our
Common Future” (World Commission on Environment and Development 1987),
international policy has been concerned with the practices that affect sustainability
(see also International Union for the Conservation of Nature 1980, US. National
Research Council 1999, Rockwood et al. 2008). Sustainability also has spurred a
vibrant literature in resource, environmental, development and ecological economies
with some contributions from political science and philosophy (Becker and Ostrom
1995, Kates et al. 2001, Clark and Dickson 2003, Norton 2005, Henry 2009).

However, aside from some critiques of the term as used in environmental politics

65

(Bltihdorn 2007, Bliihdorn and Welsh 2007), we cannot point to a sociological
approach to sustainability. Tire challenge for sociology is to develop an approach to
sustainability that moves past the focus on human action alone as in the human
exceptionalism paradigm (Catton and Dunlap 1978, Dunlap 1994) to embrace the
study of coupled human and natural systems (Liu et al. 2007a, Liu et al. 2007b).

The debate in sustainability theory over the degree to which “natural capital”,

defined as the goods and services humans derived from ecosystems (Costanza et al.

1997, Daily et a1. 2000), can be replaced by “manufactured capital?” in the form of
increased affluence provides an entry point for sociology. Neoclassical economic
theory suggests that the factors of production—land (natural resources/ natural
capital), manufactured capital and labor—can be substituted for one another to a
substantial degree (Hubacek and van den Bergh 2006) with extensions to include
human capital (Arrow et al. 2004, Dietz et a1. 2008, Engelbrecht 2009, Dasgupta in
press). If these substitutions are possible then communities can be sustained if their
total capital is constant, regardless of the distribution of capital across forms.

But a full sociological analysis must emphasize that the substitution of one
form of capital for another is not a mechanical or automatic process but an active one
(Bourdieu 1986) that often involves use of social capital, defined as the resources
people access through social relations/ties (Portes 1998, Lin 2001). A sociology of
sustainability, in tracing the processes of capital substitution, examines the tensions

between agency and structure, as individuals, households and more aggregate actors

 

3 . . . . .
Unfortunately, there are multiple terms for what we are calling manufactured capital, including

physical capital and financial capital. For the micro-level analysis we conduct, distinctions among
these terms and the attendant conceptualizations need not be explored but this is clearly an area in need
of theoretical development.

66

develop strategies to use caprtal and face constraints in realrzmg those strategies.

Attention to micro-level substitutions complements the macro-level approach that

dominates the current sustainability literature.5
4.1.1 Sustainability in a transforming economy

We focus on substitution of capital in Wolong Nature Reserve in the rapidly
changing economy of China. The reserve is a source of natural capital for its human
inhabitants because they make extensive use of fuelwood for heating and cooking.
But that practice has substantial adverse effects on the local ecosystem, and especially
the habitat of giant pandas (An et a1. 2005). Electricity is available locally and can
displace fuelwood use, but there are few opportunities in this very rural area to obtain
the income necessary to use electricity, so a direct substitution of income for natural
capital is not feasible for most households. However, the households in the Reserve
also possess human and social capital and can use those resources as a basis for labor
migration and wage income. This leads to the possibility of what we term “chain
substitution.” Local residents may be able to deploy social capital to obtain jobs that
allow a return to income from their human capital, and use the income to modify their

use of natural capital.

 

One way of enhancing well-being over the longer term is by deploying capital resources to enhance
one’s power and to use power to change the substitutability of one form of capital for another and thus
their relative value. Because power differentials are not prominent in the context we examine here, we
will not elaborate these linkages but clearly the relationships among the four capitals and the role of
agency and structure in shaping access to and use of them remain undertheorized and should be a key
element of a sociology of sustainability. For example, Braverrnan’s (1974) analysis of the deskilling
of labor can be viewed as a strategic effort by those with control of manufactured capital to reduce the
value of human capital and thus the cost of replacing human capital with manufactured capital.

5 . . . .
Sociology has much to contribute to the exrstrng macro-level literature as well but those
considerations are beyond the scope of this Chapter.

67

A refraining of conventional sustainability theory is necessary to examine the
processes of capital substitution in this context. At the micro-level in a capitalist

economy, most individuals are not “producers” in the sense that they use

manufactured capital directly to enhance their well-being.6 Rather the dominant
mode of economic activity for most households and individuals is to exchange labor
based on their human capital (i.e. formal education, skills, health, physical strength)
for income and use the income to purchase goods and services to support well-being.
Thus at the micro-level income and wealth become the operational equivalent of
manufactured capital at the macro-level. But while access to markets in which to sell
labor can be taken as given for most individuals in an advanced capitalist economy,

access to labor markets that allow human capital to generate income can be

problematic in economics in transition.7

A further elaboration of our framing is the idea of chain substitution of
capitals and in particular the use of social capital to gain employment and thus
income (Granovetter 1973, 1985). As in many other developing regions, the local

labor market in Wolong provides few opportunities for converting labor to income so

 

6 At least from Becker (Becker 1976) the concept of a household production function has been
deployed to explain key aspects of human behavior. It is true that households use some manufactured
capital in producing well-being (e.g. utensils in cooking, vehicles or animals in transportation). But a
fundamental aspect of the shift to a capitalist economy is a shift from the household as a nexus of
production and consumption to, for most households, the exchange of labor for income which in turn is
used to procure the requisites of well-being. That is the shift towards a capitalist economy entails a
shift towards a primacy of consumption over production for most members off the society. In recent
history this process has continued with the commodification of many goods and services that were
formally produced in the household, e.g. food preparation, child care, cleaning.

In most labor markets, there are income rewards to human capital in the sense that individuals with
more human capital tend to receive more income for their labor. However, as Wright (1979)
demonstrated, the relationship between human capital and income depends upon social class. Again, a
sociology of sustainability could profitably explore the role of class in the use of manufactured, natural,
human and social capitals in the generation of well-being at both the micro and macro levels.

68

migration to better developed labor markets becomes critical in the chain of capital
substitutions. The implications of rural to urban migration for the rural environment
have been examined in a number of studies over the last two decades and most
emphasize the tradeoff between using human capital locally to convert natural capital
to income and moving to an urban area where labor can generate income. As young
people migrate to urban areas, the supply of labor for forest clearing and the local
demand for forest clearing to support agriculture both drop off (Allen and Barnes
1985, Rudel 1989, Rudel and Roper 1997, Ehrhardt-Martinez 1998, Tole 1998).

Our analysis elaborates how social capital can be used to facilitate generating
income from human capital via migration to urban labor markets. We engage with
the larger literature on job seeking and migration to examine the importance of social
networks and in particular of “weak ties” (relationships characterized by low intimacy
or infrequent interaction) in shaping migration. Thus we posit that a sociological
approach to sustainability will emphasize how micro-level action to deploy capital is
embedded in a social context (Granovetter 1985). As a result, social contexts can
generate profound differences in how humans use natural resources (Frank et al.
2000). Indeed, one of the compelling counterargument to Hardin’s (Hardin 1968)
“Tragedy of the Commons” is that communities can organize in ways that prevent the
overexploitation of common pool resources (Ostrom et al. 2002, Dietz et al. 2003,
Dietz and Henry 2008).

4.1.2 Environmental change in contemporary China
Several recent analyses document the impact of human behavior on the

environment in China (Economy 2004, Liu and Diamond 2005, Liu and Diamond

69

2008). While problems of industrialization, such as air and water pollution, are the
most visible, local communities are placing serious strains on several critical habitats.
For instance, the use of wood for cooking and heating can have substantial impact on
local environment, such as our study site, the Wolong Nature Reserve (Liu et al.
2003b). More than 95% of the inhabitants are farmers living in isolated farmsteads.
Their traditional livelihood depends heavily on natural capital and includes farming,
fiielwood collection and livestock breeding. F uelwood collection has been
demonstrated to have an especially pronounced impact on panda habitat because the
amount of fuelwood collected by inhabitants is very substantial, with a mean of over
6000 kilograms per household per year, resulting in the removal of forest canopy that
provides shelter and cover for pandas (Liu et al. 1999, An et al. 2002, An et al. 2005).
As a result, the panda habitat has suffered from serious degradation (Liu et al. 2001).
The inhabitants of the reserve have an alternative to obtaining energy from
wood. Due to China’s recent investment in hydroelectric power, electricity has
become more available and more reliable. In fact, all households in the reserve have
access to electricity. In our interviews, the residents of the reserve indicated they

preferred electricity to fuelwood because it is more convenient, cleaner, and required
. 8 . . .
less labor for gathering. In contrast to electrrcrty, fuelwood IS free except for the

. . 9 . . .
labor required for extraction. However, the cost of electrrcrty has increased recently,

in large part to offset the large government investments in producing electricity. Thus

 

Fuelwood is not sold at the local market, and farmers in the reserve collect fuelwood mainly in
winter for their own use in the following year.

9 . . . . .

Although the Wolong reserve administration has developed several polrcres to reduce fuelwood
collection, monitoring and enforcement of these policies are a problem because the settlements and
fuelwood collection are very geographically dispersed.

70

the primary obstacle to the use of electricity is economic, and a key factor for
improving economic status is through taking off-farm employment. So local
residents have potential to substitute income for the use of natural capital, and would
prefer to do so, but their ability to use human capital as labor to generate income
locally is limited.

The move towards a market economy and urbanization forces that have
affected the rest of China since the 1990’s are just beginning to affect the Wolong
Nature Reserve. However, some inhabitants now complement their traditional
economic activities based on the use of natural capital by working in urban settings
through temporary rural-urban labor migration. Previous research on labor migration
has suggested that the remittance from migrants may substantially improve the
livelihood of their rural households (Koc and Onan 2004, Airola 2007).

In the Wolong Nature Reserve, labor migration may have substantial impacts
on the local ecosystem in several ways. First, remittances from labor migrants may
be used to shift rural energy consumption from fuelwood to electricity. Second, labor
migrants may also contribute to rural household economy through sending materials
(e. g., food, clothes, and electronic appliances) back home, which may free up income
for purchasing electricity. Third, the reduction of the local human population due to
labor migration may reduce energy needs (both fuelwood and electricity) of rural
households (An et al. 2001). Fourth, labor migration may reduce the labor supply for
collecting fuelwood (Allen and Barnes 1985, Rudel 1989, Rudel and Roper 1997,
Ehrhardt-Martinez 1998, Tole 1998). Thus the relationships among labor migration,

fuelwood consumption and electricity consumption may be complex.

71

4.1.3 Labor migration patterns in China

For the most part, the labor migrants of the Wolong Nature Reserve, do not
migrate to urban areas to settle permanently. Instead they seek temporary
employment in urban settings and return to their home villages whenever needed (e. g.,
in planting or harvesting seasons). In this sense, they take advantage of the rapid

economic development in China in seeking temporary jobs, but are not permanent

urban residents in the larger urbanization process. 10 Such temporary migration is

very common in China as well as many other parts of the world (United Nations 2004,
Korinek et al. 2005). In the case of the Wolong Nature Reserve, in 2004, 162 people
worked in cities through temporary labor migration. Although the proportion of labor
migrants is small (accounting for about 6.0% of eligible laborers in the reserve), it is
substantial compared with many other rural areas in China (Li and Zahniser 2002),
and is increasing rapidly (Liu Mingchong, 2005, personal communication).

The determinants of labor migration and the relationship of such labor
migration to macro political and economic changes have been carefully studied in
contemporary China (Goldstein et al. 1997, Yang and Guo 1999, Yang 2000, Liang
2001, Li and Zahniser 2002, Fan 2003). The standard model of labor migration
examines how households use human capital (e. g., gender, age and education) to
generate income via labor migration (Becker 1985, Shelton and John 1996, Angrist
and Evans 1998, De Jong 2000). Yet few studies have explored the impact of social

capital in the context of Chinese internal migration (Zhang and Li 2003).

 

In some areas, income flowing from relatively permanent urban migrants back to rural villages may
have important environmental consequences. This is an intriguing issue worthy of further explanation
but is beyond the scope of our analysis.

72

It is well known that social capital may affect migration decisions (Massey
1990, Hugo 1998, Palloni et al. 2001) and facilitate migration processes (e.g., help
migrants settle down and become familiar with places of destination) so that costs and
risks of migration may be alleviated (Korinek et al. 2005). Social capital is also
important for accessing employment information and influence (i.e. influential
persons in particular labor subsectors) (Lin et al. 1981, Granovetter 1995, Bian 1997,
Yakubovich 2005). It is important to differentiate the strength of social ties as ties
with different strengths may have different roles in facilitating labor migration. For
instance, relatives may be perceived as stronger ties than fiiends, while friends may
be perceived as stronger ties than acquaintances (Granovetter 1995, Bian 1997).
Strong social ties may be more reliable in facilitating migration processes such as
transportation and settlement, while weak social ties may expand information about
employment opportunities (Massey and Espana 1987, Massey and Espinosa 1997,
Wilson 1998). Moreover, weak ties may provide direct access to influence, while
strong ties are usually indirectly associated with influence (Granovetter 1995, Bian
1997, Yakubovich 2005).

Granovetter (1973; 1995) explained how ties, in particular weak ties, might
affect employment seeking. His arguments are very salient for the reserve. The
Wolong Nature Reserve is in a mountainous rural area, and is far from any urban
areas (>100 km), which makes communication between the reserve and the outside
difficult. Without any government institutions or other formal organizations

providing employment information, social capital is an important source of such

73

information. Without social capital it may be very difficult for a household to use
human capital to generate income.

To understand the environmental impact of labor migration in the reserve, we
must retrace a causal path that starts with the use of fuelwood, from there back to the
economic and demographic impacts of labor migration, and finally from labor
migration to the social capital that facilitates such migration. Our model is
summarized in Figure 4.1. Building on the classic effects of social ties on labor
outcomes, our first hypothesis is that access to social capital, especially weak social
ties, facilitates labor migration. The second hypothesis is that labor migration reduces
household fiielwood consumption, as income is substituted for the use of natural
capital. Thus the resources individuals access through social relations indirectly
affect fuelwood consumption. Our analysis elucidates the links between social
relations, labor migration and fuelwood consumption and thus shows the processes by
which one form of capital is substituted for another.

4.2 METHODS
4. 2.1 Household surveys

Our in-person interviews were conducted fi'om May to August 2005 in the
Wolong Nature Reserve. We chose household heads or the spouses of household
heads as interviewees because they are usually the decision makers on household
affairs and know the most about other household members’ information (e.g.,
employment and income). From the govemment’s household registration list
containing about 1200 households in all the groups in the reserve (groups are nested

within villages within townships in rural China), households with temporary labor

74

migrants were identified by group heads (farmers who are elected by their group
members to coordinate some group affairs such as recruiting laborers for group
infrastructure work). There were 138 households with temporary labor migrants in
2004. No eligible respondent in 7 of these households could be reached within 5
revisits and data from 2 households were not complete, which resulted in 129
households corresponding to 152 labor migrants. For the purpose of comparison, we
also interviewed 215 households out of 223 households randomly selected from 1018
households that were not identified by group heads as households with labor migrants.
Our overall response rate for interviews was 95%.

We collected socio-demographic information on individual members and
economic, social ties and fuelwood consumption data for households. We asked the
average amount (weight) of daily fuelwood consumption in the previous year for both
the winter season when more fuelwood is needed and the summer season when less
fuelwood is needed. Household fuelwood consumption was therefore measured as a
summation of daily consumption across the year. As noted above, previous studies in
this reserve have identified fuelwood collection as one of the main reasons of the
degradation in the local ecosystem (Liu et al. 1999, Liu et al. 2001, An et al. 2002, An
et al. 2005).

In households without any labor migrants, we asked respondents about their
social ties with people who were living or working (including temporary migrants) in
cities outside the reserve. Since labor migration could lead to social ties, in
households with labor migrants we asked respondents to recall their social ties before

migration. We measured the strengths of social ties with relatives taken as strong ties,

75

acquaintances as weak ties, and friends as ties of moderate strength. We also asked if
each type of their ties includes people holding leadership positions. Our measures of
social ties in households with labor migrants are retrospective of the pre-migration
social network. Although accurate recall of social ties is difficult (Bernard et al. 1984,
Marsden 1990), people tend to report social ties with whom they have more
interactions (Neyer 1997, F eld and Carter 2002) and hence are more important for
activities such as labor migration. We used dummy variables to denote the
availability of various social ties because measures of network size tend to be biased
in retrospective studies (Brewer 2000). Wolong Nature Reserve is a relatively
isolated area where inhabitants do not have many ties to the outside, so dichotomous
measures of social ties still capture most of the variation in social resources among
households. In households with labor migrants, we asked how much remittance labor
migrants send back home.
4. 2.2 Causal inference

Because of self-selection into labor migration, the relationship between
migration and fuelwood consumption may be confounded with other factors. In the
absence of a randomized or natural experiment assigning people to migrate or not,
any estimated effect of labor migration on an outcome may be spurious. This is
reflected in the fundamental counterfactual question: “How much fuelwood would a
household with labor migrants have consumed if the household member(s) had not
temporarily worked outside of the reserve?” This question is counterfactual because
we cannot observe the fuelwood consumption of households with labor migrants

under the condition of no one working outside of the reserve. Neglect of this self-

76

selection process can result in invalid inferences (Winship and Morgan 1999, Hirano
and Imbens 2002).

We approximate counterfactual conditions using propensity score weighting
(Rosenbaum and Rubin 1983, Robins and Rotnitzky 1995, Hirano and Imbens 2002,
Hirano et al. 2003, Morgan and Harding 2006). Propensity score techniques use the
logic of comparing individuals in the treatment group (in our case, the treatment
group is composed of the households with temporary labor migrants) to individuals in
the control group (households without labor migrants) with a similar propensity score
(likelihood of working outside). The propensity score is defined as (Rosenbaum and
Rubin 1983):

e(x) = Pr(m = l | x), (4.1)
where m is a dummy variable indicating treatment (i.e. 1 if one or more members of a
household were working outside the reserve; 0 otherwise); e(x) is the propensity for
receiving the treatment and can be estimated using a logistic regression model using
covariates x (e. g., household level human capital and economic conditions,
geographical information, and social capital).

We use weights based on the propensity scores in estimating the average
causal effect of labor migration on fuelwood consumption (Robins and Rotnitzky
1995, Hirano and Imbens 2002, Hirano et al. 2003). The weights are defined by

m 1 — m
+ .
e(x) l — e(x)

 

a)(m,x) = (4-2)

Therefore, a household with migrants is weighted by 1/ e(x) and a household without
migrants is weighted by l/(l - e(x) ). In other words, the lower the propensity of

having migrants for those households with labor migrants, the greater weight they are

77

given. Similarly, the higher the propensity of having migrants for those households
without migrants, the more weight they are given. In this way, the estimation of the
average causal effect focuses mainly on the strongest overlap in pr0pensity, those
with lower propensity in the treatment group and those with higher propensity in the
control group (Figure 4.2).

The weighting in (Eq. 4.2) is informative for policy considerations because it
reflects individual responses to incentives. If policies focus on changing incentives
and resources for labor migration, then estimates of effects should focus on those
most likely to respond to changes in policies: those who were employed outside the
reserve but who had low propensity for doing so and, therefore, might not have
become employed outside the reserve if there were fewer incentives for doing so; and
those who were not employed outside the reserve but who had high propensity for
doing so and therefore might respond to increases in incentives. Thus, the estimate
using the weights in (Eq. 4.2) is referred to as the effect of the treatment for people at
the margin of indifference (EOTM) (Heekrnan 2005).

Propensity scores can also be used as a basis for matching or defining strata
(Rosenbaum and Rubin 1983, Morgan and Harding 2006). We prefer the weighting
approach because (1) the weighting scheme is relatively simple and intuitive; (2)
estimates using the weights are easy to obtain (e. g., using weighted least squares) and
can be implemented within the context of simple or more complex models; (3)
because the estimand is a smooth function of the data (as in the weighted regression),
bootstrapping techniques can be employed to calculate standard errors that reflect

uncertainty in estimating the propensity; and (4) all subjects contribute to the analysis

78

(though not equally, by definition). In fact, all matching estimators can be considered
examples of weighting approaches (Morgan and Harding 2006), but only the Robins’
approach we use here has been proven to improve the efficiency of estimation
(Hirano et al. 2003). The greatest concern about weighting is that extreme weights
could exert undue influence on the estimates. This is easily addressed by examining
the distribution of weights and trimming extreme values.

Under some circumstances, separate causal effects for the migration group
and the non-migration group are of interest. These estimates can be obtained with
minor changes to the weights. In particular, to estimate the effect of labor migration
for those households in which a member was working outside of the reserve, the

following weights can be used:

_ _ e(X)
(amt-(mgr) — m + (l m) 1_ e(x) . (4.3)

 

Thus those working outside of the reserve are weighted with a value of one, and
members of the comparison group are given more weight if they have a higher
propensity to migrate. As a complement, to estimate the effect of temporary labor
migration for those households in which no one was working outside the reserve, the

following weights can be used:

wnonmi

(m,x)=m1—;%+(l—m). (4.4)

Here those households in which no one was working outside the reserve are assigned
a weight of one, and those with labor migrants are given more weight if they have a

lower propensity to migrate.

79

4. 2.3 Analytical approach

We first model the propensity for labor migration as a function inter alia, of
social ties. Then we estimate the effect of labor migration on fuelwood consumption.
All laborers (912 people) from the 344 households that we interviewed are used in
logistic regression models to estimate the propensity for labor migration. Based on
past studies of labor migration in China (Goldstein et al. 1997, Yang and Guo 1999,
Yang 2000, Liang 2001, Li and Zahniser 2002, Fan 2003, Zhang and Li 2003), we
chose both individual level and household level factors as potential determinants of
temporary labor migration. At the individual level, we chose gender, age, marital
status, education level, number of children younger than 15 years of age and
availability of extended household member. At the household level, we chose
amount of land, non-migration income (measured by excluding migration income
from total household income), number of laborers (18~60 years of age, people
beyond this range usually do not work outside) and indicator of township the
household is located. We extend the model specifications suggested in the literature
by adding social capital to these individual and household level human capital and
income and wealth factors. We did not find multicollinearity among independent
variables.

The first model includes three dummy variables denoting the availability of
relatives, friends and acquaintances living or working in cities outside the reserve. In
models two through four, we isolate effects of each particular tie as well as the extent
to which the ties hold leadership positions. The fifth model controls for the

availability of any type of social ties (aggregation of the availability of three types of

80

social ties) and any ties to people holding leadership positions (aggregation of the
availability of three types of social ties holding leadership positions) outside the
reserve. We use the last model to calculate the propensity weights because it has the
best fit according to AIC criterion (the model having the lowest AIC is the best).
Moreover, because this model includes the primary factors predicting labor migration
described in the literature as well as measures of social capital we use it as a basis of
causal inference in the absence of a randomized experiment which would have been
ethically and logistically difficult to conduct in this context (Shadish et al. 2002).
After calculating the propensity weights, we confirm that the weights achieve balance
on our covariates by testing for differences between households with and without
temporary labor migrants using the weighted and unweighted data. Reduction in
differences when employing the weights suggests that selection bias has been
adjusted via the weights (Morgan and Harding 2006).

Next, we use the estimated propensities to weight a standard regression of the
effect of labor migration on fuelwood consumption. As noted above, the weights take
into account factors affecting the propensity of labor migration and allow us to focus
on the estimated effect of those on the margin of deciding whether work in an urban
context (Heckman 2005). In addition to the migration status of households (measured
with a indicator of whether a member of the household had engaged in temporary
labor migration in 2004), we control for household size, availability of senior
members (people over 60 years of age), household income, amount of land, number
of pigs the household breed, and indicator of township the household is located as

covariates in fuelwood consumption models as suggested by past studies in this

81

reserve (An et al. 2001, An et al. 2002). Moreover, a few determinants of labor
migration that are not well balanced through the propensity weighting are also
controlled as covariates. We hypothesize that there are effects of labor migration on
fuelwood consumption beyond the direct economic returns from labor migration
because migration reduces household labor for gathering fuelwood and decreases
demand as a result of the absence of a household member. In addition, migrants may
also send materials (e. g., food, clothes, and electronic appliances) home. To reflect
the potential indirect effects of labor migration, total household income, as an
alternative to non-migration income, is accounted for in some fuelwood consumption
models.

We use all working age individuals (household members from 18 to 60 years
of age) as units of analysis in estimating the propensity model because it is
individuals who choose whether work outside the reserve. Since we also use some
household predictors (e. g., amount of land owned by the household) in this model, we

report robust standard errors in the propensity models to account for the resulting lack

of independence among observations.11 We analyze fuelwood consumption at the
household level because fuelwood is consumed by households. The highest
propensity score of any individual in the household is assigned to the household in
that few households had more than one labor migrant.

Because there is uncertainty in the estimates of the propensity scores that are
used in weighting our fuelwood consumption models, we use case based

bootstrapping to calculate standard errors (Efron and Tibshirani 1993). For each

 

ll . . .
We corrected for dependencres among members of the same household wrth Huber‘s variance
correction in STATA 8.0 (STATA Corp., College Station, TX, USA).

82

estimate of the propensity and fuelwood consumption models, we calculate standard
errors from 500 bootstrap replicates that are then the basis for t—ratios.
4.2.4 Quantifying the robustness of the inference

Although we have attempted to reduce bias in our estimate by controlling
(through propensity score weighting) for many well recognized factors affecting labor
migration and fiielwood consumption as well as drawing on our own understanding of
the phenomenon in the reserve, we may have omitted confounding factors that could
bias our estimates. Therefore, we explore the robustness of our inferences to the
possibility of omitted variables.

Our approach to quantifying robustness can be considered an extension of
sensitivity analysis (Rosenbaum and Rubin 1983, Holland 1989, Copas and Li 1997,
Robins et al. 2000, Scharfstein and Irizarry 2003). Sensitivity analyses consider a set
of possible estimates given a broad set of alternative conditions. As in sensitivity
analysis, we consider how violations of assumptions could affect estimates. But rather
than reporting how violations of assumptions produce a range of estimates, we focus
on exactly how much an assumption must be violated to invalidate an inference. As a
result, the indices quantify the robustness of the original inference.

Classically, internal validity can be expressed in terms of confounding
variables that are correlated with both the predictor of interest and the outcome
(Shadish et al. 2002). We express the robustness of our inferences to these two
relationships by employing the impact threshold (Frank 2000, Pan and Frank 2003).
Frank (2000) defines the impact of a confounding variable on an estimated regression

coefficient as rvy * rvm , where rvy is the correlation between a covariate, v, and the

83

outcome y; and rvm is the correlation between v and m, the predictor of interest (for

example, m is an indicator of the status of labor migration of the household -- see
Figure 4.3).

To obtain the impact necessary to invalidate an inference, define r# as a

quantitative threshold for making inferences. Note that there is a direct relationship
between r (the observed correlation between the predictor of interest and the outcome)

and the statistical significance (t-ratio) of the predictor of interest (Cohen and Cohen

1983), t= ’1 d'f'

2

 

, where df is the degree of freedom in the regression analyses. In
1 — r

particular, defining tam-ca, as the critical value of a t-distribution (e. g., for p 50.05),

t . .
then r# = M defines a threshold based on statistical significance. That is,

,ld.f.+t2

rmy (the correlation between the predictor of interest, or, and the outcome y) will be

statistically significant if and only if it is greater than r#.

Given the definition of r# , a simplification of Frank (2000) shows that the

#

r — r

original inference from a bivariate regression is invalid if rvy * rvm > LIF' Thus
l-lr l

r _ #
the quantity _r_ny__

# defines the impact threshold for a confounding variable for
l— r ‘

 

simple linear regression. That is, if the impact (rvy * rvm ) of a confounding variable is

#

"my
l-r#

greater than the original inference is not valid. The corresponding threshold

 

 

84

,#
)

 

’.
. . . . . mv :
for the multivariate case With covariates 2 rs '

2 2
1 (la—r...» (I—rya .

,#
I

 

 

Critically, because the impact is defined by correlation coefficients it can be readily
understood by social scientists comfortable with correlation and the general linear
model. This makes it an ideal complement to our use of propensity score weighting
applied in a general linear model.
4.3 RESULTS
4.3.1 Data summary

There were 152 labor migrants from 129 households, and they worked in the
construction (31.6%), transportation (11.8%), industry (18.4%), service (29.6%) and
business (7.9%) sectors. About 74.3% of them worked in cities within the Sichuan
province, and 25.7% of them worked in cities in other provinces in China. Summary
statistics of individual level variables are presented in Table 4.1. About 52% of 912
laborers in our sample of 344 households were male with an average age of 36 years,
and 79.1% of the laborers were married. The mean number of years of education was
6. On average, each of these laborers had less than one child under 15 years, and
about half the laborers lived with extended household members such as parents or
parents-in-law. Labor migrants accounted for 16.7% of 912 laborers in our stratified
sample of 344 households.

At the household level, the mean household size was 4.7 people, while the
mean number of laborers was 2.7 (Table 4.2). About 1/3 of these households had
senior members. The mean non-migration income was 10.253 thousand Yuan, and

the mean total household income was 11.377 thousand Yuan. On average, each

85

household owned 0.282 hectares of cropland, breed about 3 pigs, and about 60% of
these households were located in the Gengda township. About half of the households
had relatives working or living in urban areas, but less than half of these relatives held
leadership positions (Table 4.2). Only about 19% and 24% of the households had
fiiends and acquaintances working or living in urban areas respectively, and few of
these ties held leadership positions. More strong ties (i.e. relatives) were reported
than weak ties (i.e. acquaintances) presumably because people tend to report social
ties with whom they have more interactions (N eyer 1997, Feld and Carter 2002). By
combining different types of social ties, 66.3% of the households had social ties in
urban areas and 30.2% of the households had ties with people holding leadership
positions. The proportion of households with labor migration in the overall study
area was 11.9% but in our stratified sample 37.5% of households had labor migrants.
On average, each household in our sample consume 6325 kilograms of fuelwood.
4. 3.2 Determinants of labor migration

Models of the determinants of temporary labor migration are presented in
Table 4.3. Model 1 shows that households with weak ties (i.e. acquaintances) were
significantly (p50.001) more likely to have labor migrants than were other
households—this form of social capital facilitates being able to use human capital to
produce income. The effect of relatives, representing strong ties, is not statistically
significant. These results are consistent with Granovetter’s “the strength of weak
ties” hypothesis (Granovetter 1973, 1995). Holding all other factors constant, the
availability of an acquaintance increases the odds of labor migration by 2.54, while

the effects of the availability of relatives and friends on labor migration do not

86

significantly differ from zero (see model 1 in Table 4.3). When exploring different
types of social ties separately controlling for demographic and economic factors as
covariates (see models 2 through 4), the availability of relatives and friends working
or living in urban areas does not have significant effects on labor migration, while the
availability of acquaintances still has a significant (p50.01) positive effect on labor
migration with a similar magnitude as that in model 1. Moreover, the insignificance
of ties holding leadership positions indicates that leadership ties were not more

helpful than non-leadership ties. When different types of ties were combined (model

5; AIC = 587.629, pseudo R2 = 0.319), the availability of social ties in cities outside

of the reserve significantly (p50.01) increased people’s probability of labor migration.
Holding all other factors constant, the availability of social ties increased the odds of
labor migration by 2.21. Thus social capital is, as expected, very important in
obtaining income from labor.

Human capital and economic conditions had similar effects across the 5
models. It is not surprising that human capital was very important. Men were more
likely (p50.001) to work outside the community than women because men are usually
expected to assume economic responsibilities for households in rural areas of China.
The odds of labor migration for men is about 2.82 times higher than that for women
(see model 5). Both age and its quadratic term had significant (p50.05) effects on
labor migration. The quadratic relationship between age and migration shows that the
probability of migration increases until 30 years and then declines as age increases.
The odds of labor migration for married people is only about 0.21 times of that for

unmarried people. Education increases the probability of labor migration

87

significantly (p50.001). Holding all other factors constant, the odds oflabor
migration increased by a factor of 1.20 for each additional year of education. No
effects of extended household member(s) and the number of children under 15 years
were detected. The number of laborers in the household had a significant (p50.05)
positive effect on labor migration. Each additional household laborer increased the
odds of labor migration by 1.39 holding other factors constant (Model 5 in Table 4.3).
Amount of cropland of the household did not have a significant effect on labor
migration. Non-migration income had a significant (p50.01) negative effect on labor
migration. Holding all other factors constant, the odds of labor migration decreased
by a factor of 0.93 with an increase in non-migration income of one thousand Yuan.
These effects are consistent with the fact that labor migration is a way of finding
alternative opportunities for those with the most limited opportunities in the reserve to
balance the inequalities in economic status and demographic conditions among
households. Finally, residing in Gengda township had a significant (p50.01) positive
effect on labor migration. The odds of labor migration for people in the Gengda
township was 2.29 times of that for people in the Wolong township. This result
reflects the fact that the Gengda township is geographically closer to urban areas
outside the reserve so its inhabitants have access to more information and material
exchanges with the outside than those living in the Wolong township. Our results of
the determinants of temporary labor migration are consistent with many other
empirical studies at regional or national levels in China (Goldstein et al. 1997, Yang

2000, Li and Zahniser 2002, Fan 2003).

88

4. 3.3 Balancing covariates using the propensity score weights

We tested for differences between those households with labor migrants and
those without labor migrants on household level variables that are used in the
propensity model. Note that we tested only for the household level characteristics
because our next model of fuelwood consumption is defined at the household level.
Test statistics with and without using propensity weighting as in equation (2) are
presented in Table 4.4. Propensity weighting reduced the differences between the
migration group and the non-migration group on almost all the household level
variables except the Gengda township indicator. Land, non-migration income, and
number of laborers in the household were not significantly different between the
migration and non-migration groups after weighting. Although the availability of
social ties outside of the reserve and that of ties to those holding leadership positions
were still higher for the migration group than those for the non—migration group in the
weighted analysis, there was less difference between the two groups after weighting.
These two social ties covariates and the Gengda township indicator were still
significantly different between migration and non-migration groups after weighting,
and therefore were controlled in estimating the effects of labor migration on fuelwood
consumption.
4.3.4 Estimation of the effects of labor migration on fuelwood consumption

Estimates of the effect of labor migration on fuelwood consumption with
propensity weighting are presented in Table 4.5. Fuelwood consumption of
households with labor migrants was significantly less than fuelwood consumption of

households without migrants. When non-migration income, together with other

89

covariates, was included in the model, households with migrants consume 1827
kilograms less fuelwood (~28.9% of average annual household fuelwood
consumption in the reserve) on average than those without migrants ((750001). In
contrast, the effect of labor migration without using weights was estimated to be 1647
kilograms (p_<_0.001).

We also estimated the effects of labor migration separately for the migration
groups and non-migration groups using estimate-specific weights (see equations 3

and 4). Labor migration had less effect on reducing fuelwood consumption for those

households in the migration group (the 3rd row in the lSt column of Table 4.5), while

. . . . h .
the effect rs strongest for those in the non-migration group (the 4t row in the 1St

column of Table 4.5). Presumably, the difference is due to the differences in
characteristics between these two groups. For example, a high propensity of labor
migration may indicate that the household has more laborers, and a reduction of one
laborer from a household that has many laborers may not affect as much the supply of
labor for fuelwood collection as that from a household that has few laborers.

In addition to the direct economic contribution of labor migration, following
the deforestation literature we also hypothesized indirect effects of labor migration on

fuelwood consumption. To estimate these effects non-migration income was replaced

with total household income with results reported in the 2nd column of Table 4.5. In

this model, the effect of labor migration is net of the income it contributes to the
household. Labor migration still has significant negative effects on fuelwood

consumption, although the magnitude of effects is smaller than that when non-

migration income was controlled for (lSt column of Table 4.5). This result suggests

90

that labor migration has both a direct economic contribution and an indirect effect on
reducing fuelwood consumption. The indirect effect may occur because migrant
laborers send materials (e. g., food, clothes, and electronic appliances) home, and their
absence may reduce both the need for fuel in the household and the labor available to
gather fuelwood, and may even affect the lifestyles of their household (e.g., using
electric stoves and other appliances which in turn may make electrical use routine for
heating as well).
4.3.5 The robustness of the inference

We base the analysis of the robustness of the inference on the estimate of the
average effect of labor migration, including non-migration income (coefficient = -
1827, standard error = 242). The observed t-ratio of -7.55 translates to a correlation

coefficient of -0.378 and, for a sample size of 344, the threshold for statistical
significance (r#) is a correlation of -0.107. The corresponding impact threshold is -

0.25. That is, to invalidate the inference the magnitude of the impact of an
unmeasured confounding variable must be greater than 0.25. Furthermore, the

magnitude of rvy (the correlation between the unobserved confounding variable and

fuelwood consumption) must be greater than 0.47 and the magnitude of rvm (the
correlation between the unobserved confounding variable and labor migration) must

be greater than 0.54 to invalidate the inference.12 Each component correlation is
large by social science standards (Cohen and Cohen 1983). Moreover, these are zero-

order correlations, assuming that the unmeasured confounder is uncorrelated with the

 

The zero order correlations are not necessarily equal when the impact 18 maxrmrzed wrth covariates
in the model. If the component correlations do not take these exact values then the impact would have
to be greater than .22 to invalidate the inference.

91

measured covariates (Frank 2000). The relevant partial correlations (Cohen and
Cohen 1983) from which the impact of an unobserved confounder would be

constructed would be smaller than the zero-order correlations because of correlations

with existing covariates. 13

Though the magnitude of the impact threshold for an unmeasured variable can
be interpreted in terms of typical patterns of correlation in the social sciences, it is
also helpful to compare the threshold to the impacts of measured covariates. Based
on zero order correlations, the magnitude of the impact of the indicator of township
(Gengda versus Wolong) is the largest of the existing covariates. Its impact is -0.037
and the sign is in the direction that reduces the negative effect of labor migration on
fuelwood consumption. Thus the magnitude of the impact of an unmeasured
confound necessary to invalidate the inference that labor migration affects the amount
of fuelwood consumed in the household (0.25) would have to be more than six times
greater than the magnitude of the strongest impact of the measured covariates, -0.037.
4.4 CONCLUSIONS

We have suggested that an appropriate sociological approach to sustainability
is to consider the strategies individuals and households deploy to generate well-being
from their income and wealth, access to natural capital, human capital in labor and
social capital. This approach is consistent with the existing sustainability literature
that emphasizes problems of capital substitution. But it extends that approach to

include the important sociological insight of the tension between agency, in the form

 

3 . . . . .
Frank (2000) refers to this as absorption of the impact of an unmeasured confound by exrstrng
covariates.

92

of individual and household strategies, and structural constraints, in the form of
limited access to some forms of capital.

In the Wolong Nature Reserve, the most crucial environmental threat is
deforestation and the resultant degradation of panda habitat. On average, local
residents use very substantial amounts of natural capital in the form of fuelwood for
cooking and winter heating. While the possibility of substituting electricity for
fuelwood exists, the costs of electricity and the paucity of local opportunities to
convert human capital, via labor, into income preclude this move away from the use
of natural capital for most households—a structural constraint. However, our
analysis shows that a form of social capital, weak ties, is often used to gain access to
extra-local employment and that the income from this employment then displaces the
use of local natural capital. Of course the electricity generation from hydropower
plants and coal burning may also have negative environmental impacts, which are
beyond the scope of this study.

In addition to providing a “demonstration of concept” for our proposed
sociological approach to sustainability, our results also address two other issues in the
literature. First, we have replicated in rural China a finding developed elsewhere—
that among forms of social capital it is weak ties that matter most in finding
opportunities to find employment (Garip 2008, Pfeffer and Parra 2009). We note that
strong ties may produce weak ties, but in our research and that of others back to
Granovetter (1973), it is weak ties that have the most impact.

Second, we have shown that, at least in the context of the Wolong Nature

Reserve, the effect of labor migration on deforestation comes from the ability to use

93

increased household income to purchase a substitute for local natural capital. It is
well understood that labor migration can have substantial environmental
consequences (Bilsborrow and Ogendo 1992, Bilsborrow 2002, Rudel et al. 2002,
Aide and Grau 2004, Liu and Diamond 2005), an issue first raised by Marx (Foster
1999). But without in—depth understanding of how migration decisions are shaped by
context and why they vary across individuals and households, it is hard to understand
the dynamics and impacts of migration and ultimately the environmental
consequences of migration (Walker 2008). In Wolong context, our results contrast
with some earlier findings on labor migration and deforestation that emphasize the
loss of labor supply as the mechanism by which extra-local employment eases
deforestation (Allen and Barnes 1985, Rudel 1989, Rudel and Roper 1997, Ehrhardt-
Martinez 1998, Tole 1998).

The overall adverse effects of Chinese economic development are well
documented (Liu and Diamond 2005, Liu and Diamond 2008) and by 2015 China is
projected to have, after the US, the second largest ecological footprint of any nation
(Dietz et al. 2007). Policy efforts to ameliorate this impact and move China and other
economies in transition towards a more sustainable path must be designed with
sensitivity to local context to avoid perverse effects (Liu et al. 2007a). The effects of
weak ties in the Wolong Reserve communities suggest a relatively low-cost
mechanism to encourage the substitution of income for use of local natural capital. In
the reserve, it appears that access to extra-local labor markets is the key structural
constraint on household strategies. Creating local labor markets that allow exchange

of labor for income is difficult and the ability to do so without violating the

94

 

sustainability goals of the Nature Reserve may be limited. However, enhancing
social capital by providing better information on and access to extra-local labor
markets is a relatively low cost policy option for government. In the case of Wolong
this could reduce the demand on fuelwood. However, the effects of reducing this
structural constraint on deploying human capital to produce income is context
specific and so might or might not reduce the use of local natural capital in other
contexts. Developing effective policies requires careful analysis of how those
influenced by the policies will respond.

Finally, while we have emphasized the household and individual as units that
deploy capital to enhance their well-being, a sociology of sustainability should not
limit itself to the micro level. Part of the sociological tradition is to consider not only
individuals and households as agents but also communities, social movements, formal
organizations, government and nations. Sociology could contribute fruitfully to our
understanding of sustainability by examining the strategies used by these collective
actors and the constraints they face in deploying the capital resources available to

them.

95

 

Table 4.1. Descriptive statistics of individual level variables.

 

 

Independent Variables Mean

(Standard Deviation)
Male (male = l and female = 0) 0.520 (0.500)
Age (years) 36.034 (11.541)
Age Squared 1431.492 (889.042)
Married (married = 1 and single = 0) 0.791 (0.407)
Education (years) 5.998 (3.490)
Children (number of children with age 5 15 years) 0.867 (0.933)
Extended 0.507 (0.500)

(1 if there is extended member in the household;

0 if no extended member in the household)

Migrant (1 if the individual is a labor migrant; 0.167 (0.373)
0 if the individual is not a labor migrant)

 

(n=912)

96

 

Table 4.2. Descriptive statistics of household level variables.

 

Variables

Mean
(Standard deviation)

 

Household Size (number of people in the household)

Laborers (number of working age people—18~60 years ofage—in the
household)

Senior (1 if there is senior member in the household: 0 if no senior
member in the household)

Non-migration Income (thousands of Yuan)

Total Household Income (thousands of Yuan)

Land (hectares)

Pigs (number of pigs the household breed)

Gengda (Gengda township = l and Wolong township = 0)

Relative (1 if there is relative outside the reserve; 0 if no such relative)
Relative Leader (1 if there is relative outside the reserve holding
leadership position; 0 if no such relative)

Friend (1 if there is friend outside the reserve; 0 if no such friend)
Friend Leader (1 if there is friend outside the reserve holding leadership
position; 0 if no such friend)

Acquaintance (1 if there is acquaintance outside the reserve;

0 if no such acquaintance)

Acquaintance Leader (1 if there is acquaintance outside the reserve
holding leadership position; 0 if no such acquaintance)

Tie (1 if there is any type of social tie outside the reserve; 0 if no tie
outside the reserve)

Tie Leader (1 if there is any type of social tie outside the reserve holding
leadership position; 0 if no such tie)

Migration (1 if there is labor migrant(s) in the household; 0 if no labor
migrant(s) in the household)

F uelwood Consumption (kilograms)

(n = 344)

97

4.663 (1.288)
2.651 (1.061)

0.326 (0.469)
10.253 (7.887)
11.377 (9.376)
0.282 (0.152)
2.881 (2.159)
0.599 (0.491)
0.517 (0.500)
0.215 (0.412)

0.189 (0.392)
0.061 (0.240)

0.244 (0.430)
0.055 (0.229)
0.663 (0.473)
0.302 (0.460)
0.375 (0.485)

6325 (4499)

Table 4.3. Determinants of labor migration models.

 

 

 

Independent Coefficient (Adjusted standard error) [odds ratios]
variables Model 1 Model 2 Model 3 Model 4 Model 5
Male l.029*** 1037*" 1034*“ 1044*" 1035*“

(0.222) [2.798] (0.214) [2.821] (0.217) [2.812] (0.223) [2.841] (0.222) [2.815]
Age 0.316" 0.320" 0.315** 0.325" 0300*

(0.117) [1.372] (0.120) [1.377] (0.121) [1.370] (0.118) [1.384] (0.121) [1.350]
Age Squared -0.005** -0.005** -0.005** -0.005** -0.005**

(0.002) [0.995] (0.002) [0.995] (0.002) [0.995] (0.002) [0.995] (0.002) [0.995]
Married -l.725*** -l.540*** -l.644*** -1.726*** -l.563***

(0.344) [0.178] (0.340) [0.214] (0.335) [0.193] (0.339) [0.178] (0.361) [0.210]
Education 0186*" 0.181*** 0.182*** 0190*" 0186*"

(0.043) [1.204] (0.040) [1.198] (0.041) [1.200] (0.043) [1.209] (0.044) [1.204]
Children 0.072 0.110 0.091 0.066 0.067

(0.182) [1.075] (0.184) [1.116] (0.184) [1.095] (0.182) [1.068] (0.187) [1.069]
Extended 0.314 0.417 0.316 0.287 0.346

(0.312) [1.369] (0.311) [1.517] (0.318) [1.372] (0.315) [1.332] (0.315) [1.413]
Laborers 0.359" 0.336“ 0.325" 0.348“ 0.327*

(0.126) [1.432] (0.121) [1.399] (0.124) [1.384] (0.126) [1.416] (0.129) [1.387]
Land 0267 -0. 108 0.016 -0.300 -0. 160

Non-m igration
Income
Gengda
Relative
Relative
Leader

Friend

Friend Leader
A cquaintance
A cquaintance
Leader

Tie

Tie Leader

Intercept

AIC
2
Pseudo R

(0.870) [0.766]
-0.082***
(0.024) [0.921]
0.782***
(0.238) [2.186]
0.196

(0.217) [1.217]

0.197
(0.281) [1.218]

0930*“
(0.244) [2.535]

-8.003***
(1.959)
589.247
0.319

(0.816) [0.898]
-0.069**
(0.022) [0.933]
0907*"
(0.235) [2.477]
-0015

(0.263) [0.985]
0.560

(0.323) [1.751]

-8.045***
(1.958)
599.763
0.304

(0.863) [1.016]
-0.065**
(0.022) [0.937]
0937*"
(0.238) [2.552]

0.061
(0.333) [1.063]
0.710

(0.513) [2.034]

-7.848***
(1.991)
600.832
0.303

(0.826) [0.741]
-0.079***
(0.024) [0.924]
0793*"
(0.238) [2.210]

0892*"
(0.274) [2.440]
0.307

(0.384) [1.359]

-8.009***
(1.956)
588.033
0.319

(0.859) [0.852]
-0.073***
(0.023) [0.930]
0.830***
(0.246) [2.293]

0.795"
(0.294) [2.214]
0.316

(0.268) [1.372]
-8.126***
(2.014)
587.629

0.319

 

Significance: * p $0.05; ** p $0.01; *** p $0.001 (two-tailed tests). n = 912 ("mi =152)

98

 

Table 4.4. Testing for balance between migration group and non-migration group
with and without propensity weighting.

 

 

 

Migration Non-migration

group group

(n=129) (n=215)
Variable Mean (Standard deviation) t—ratio a 2 t-ratio 12

(unweighted) (unweighted) (weighted) (weighted)

Land 0.286 (0.153) 0.279 (0.152) -0.41 -O.39
Non-migration 9.098 (7.796) 10.945 (7.878) 2.1 l 0.37
Income
Laborers 3.093 (1.169) 2.386 (0.894) ~5.9l -l .58
Gengda 0.698 (0.461) 0.540 (0.500) 8.39 10.30
Tie 0.814 (0.391) 0.572 (0.496) 21.10 15.95
Tie Leader 0.426 (0.496) 0.228 (0.420) 15.05 8.29

 

a Positive value indicates households with labor migrant(s) have lower mean than
those without labor mi grant(s).

 

99

Table 4.5. Estimated effect of labor migration on fuelwood consumption (kilograms)

using general linear models (GLM).

 

Models Coefficient (Bootstrap standard error)

 

Covariates including
non-migration income

Covariates including
total household income

 

GLM: unweighted -l647*** -1262**
(467) (461)
GLM: average effect of labor -l827*** -1482***
migration (242) (249)
GLM: effect for migration -1253** —988*
group (424) (409)
GLM: effect for non-migration -2067*** 4668*“
jroup (263) (279)

 

Significance: * p $0.05; ** p $0.01; *** p $0.001 (two-tailed tests). 11 = 344

100

 

Social
capital

 

 

Propensity model

 

 

 

 

 

 

I """" ': """""""""""" I
: Temporary : Environmental impact

I fl

l Labor : (F uelwood consumption)

: migration I

I I

Figure 4.1. The effect of social capital on the environment as mediated by labor

migration.

10]

Propensity Score

High

   

 

 

Overlap between Average
Treatment Group Treatment
and Control Group EffeCt
Low ‘ 1 1
Treatment Group Control Group

Figure 4.2. Overlap in propensity scores between treatment group and control group.

102

Temporary Fuelwood

Labor Migration (m) 3 Consumption (y)

3

   
 

Confounding (v)

Figure 4.3. The impact of a confounding variable on a regression coefficient.

103

 

CHAPTER 5

MODELING THE EFFECTS OF PAYMENTS FOR ECOSYSTEM SERVICES

IN A COUPLED HUMAN-NATURE SYSTEM

In collaboration with

Ashton Shortridge, Andrés Vifia, and Jianguo Liu

104

ABSTRACT

Payments for ecosystem services (PES) have increasingly been implemented
to protect and restore ecosystems worldwide. The efficiency of conservation
investments in PBS may differ under different policy arrangements. In addition, the
effects of PBS programs may be uncertain due to uncertainties in human responses to
policies and complex depending on the dynamic human-natural interactions. To
demonstrate the impacts of human-environment interactions on the effects of PES

programs, we developed a spatially explicit model, human and natural interactions

 

under policies (HANIP). We used HANIP to study the effects of China’s Natural 1
Forest Conservation Program (N F CP) and alternative policy scenarios in a coupled

human-nature system (China’s Wolong Nature Reserve for giant pandas), where

indigenous people’s use of fuelwood affected panda habitat. We estimated the effects

of the current NF CP providing cash payment and an alternative payment scenario

providing electricity payment by comparing habitat dynamics under these policies to

habitat dynamics where no payment is provided. By 2030, there will be 107.68 km2

of panda habitat in the study area if no payment is provided. Under the current NFCP,
about 31.9% of habitat area can be obtained from the conservation payment by 2030.
If the cash payment is replaced with electricity payment, an additional 11.6% of
habitat area can be obtained. In addition, the recovery rate of panda habitat will be
decreasing due to increasing population and households. Conservation effects of the
NFCP may be threatened by the behavior of newly formed households if they are not
included in the payment scheme. Our study demonstrated the advantages of

integrating dynamics in human activities with the natural environment. Our modeling

105

framework may also be applied to understanding the effects of conservation policies
in other coupled human-nature systems.
5.1 INTRODUCTION

Humans have substantial and growing impacts on the Earth’s ecosystems
(Millennium Ecosystem Assessment 2005). For instance, humans have transformed
between one-third and one-half of the land surface (Vitousek et a1. 1997), resulting in
biodiversity loss and ecosystem degradation worldwide (Wackemagel et al. 2002,
Luck et a1. 2004). Human alteration of earth is not limited to human-dominated areas,
but is also common in many protected areas in the world (Liu et al. 2001, Curran et al.
2004). To counter this trend, conservation efforts, including payments for ecosystem
services (PES), have been invested by governments, private sectors, and conservation
non-government organizations (OECD 1997, Ferraro and Kiss 2002). Much of these
efforts have been aimed at reducing human impacts through shaping human activities
(Smith 1995, Zbinden and Lee 2005, Wunder 2008). To improve the efficiency of
conservation investments, PES programs have been implemented to provide
incentives directly to ecosystem services providers (Ferraro and Kiss 2002, Wunder
2007). However, the effects of PES programs depend on the program’s design and
dynamic interactions among components in coupled human and natural systems.

Coupled human and natural systems (CHANS) are integrated human and
natural systems in which human components interact with natural components (Liu et
al. 2007a, Liu et al. 2007b). Although the importance of human-nature interactions
has been recognized for the sustainability of both human and natural systems (Foley

et al. 2005, Millennium Ecosystem Assessment 2005), complex processes in these

106

 

interactions and patterns emerged from such interactions have not been well
understood. Lack of studies of CHANS is mainly because social and ecological
sciences have traditionally been developed separately (Rosa and Dietz 1998).
Understanding of CHANS, however, relies on the integration of both social science
studies and natural science studies.

Traditional methods for human activities use household survey data to study
the impacts of individual characteristics on behavior (Chen et al. 2009a). But little is
known how dynamics of individual characteristics will interact to result in macro-
level ecological dynamics. Moreover, these methods usually cannot address
heterogeneities in human activities across landscapes. For instance, two land parcels
in a landscape with similar biophysical features may have different dynamic
trajectories due to heterogeneities in human activities. Studies of natural systems
often use aggregated data to detect patterns and changes (Liu et al. 2001, Vina et al.
2007). However, it is difficult to understand driving processes of dynamic natural
systems using aggregated ecological data because decision-making of humans is often
at a lower level (e. g., person or household).

Originating in artificial intelligence and paralleling individual-based modeling
in ecology, agent-based modeling (ABM) is a bottom-up method that simulates
actions of individual “agents” (e.g., persons or households) and their interactions with
the environment to produce the aggregated macro-level patterns and processes
(Parker et al. 2003, An et a1. 2005). Agents have autonomous actions and are capable
of interacting with other agents. Because of these features, ABM has been

successfully applied in ecological studies, such as those for land use/cover changes, to

107

understand driving processes of enviromnental changes and explore plausible future
trajectories and policy implications (Deadman et a1. 2004, Manson and Evans 2007,
Matthews et al. 2007). ABM is also an ideal tool for understanding responses of
human activities to institutional transitions and the resulting macro-level
environmental and social changes.

The development of an agent-based model is often on the basis of object-
oriented programming in computer science. In object-oriented programming, each

modular unit is an object that has its own state (represented by its attributes) and

 

behavior (implemented by its methods). Implementations of objects are modules of
the program, and are relatively separated among each other to reduce programming
complexity whereas objects may change state and behavior in response to states and
behavior of other objects. Due to similarity in the paradigm, both object-oriented
programming languages (e.g., Java, C++) and tools (e. g., NetLogo, Swarm) have
been used for developing agent-based models. In studies of human-natural
interactions and responses to policies, ABM is usually parameterized with studies
from traditional approaches that address factors affecting individual decision-making
and forces driving environmental change.

To demonstrate the impacts of human-natural interactions on the effects of
PBS programs, we developed a spatially explicit model, Human and Natural
Interactions under Policies (HANIP). We used HANIP to study the effects of China’s
Natural Forest Conservation Program (NFCP, one of the largest PES programs in the
world) and alternative policy scenarios in Wolong Nature Reserve for providing

habitat to giant pandas (Ailuropoda melanoleuca). HANIP incorporates agent-based

108

modeling techniques for modeling dynamics in population, households, and their
activities and regression-based models for corresponding changes in the landscape.
We chose household as the unit of human activities because household is often the
basic socioeconomic unit for land use decision-making and consumption of resources
(Liu et al. 2003a, Manson and Evans 2007). We focused on fuelwood use and
development of farmland for newly formed households because they are main forces
of dynamics in forests. Changes in the quantity and distribution of forests result in
changes in panda habitat. In HANIP, there are two major types of agents, person and
household. Dynamics in these agents result in dynamic human impacts (fiielwood use
and farming), which produce changes in panda habitat. HANIP was developed using
Java programming language (J DK 1.4.2, Sun Microsystems).
5.2 METHODS
5.2.1 Model summary

The conceptual framework of HANIP is described in Figure 5.1. This
framework was implemented in three submodels. In the demographic submodel,
dynamics in people including population and households were modeled by simulating
individual persons’ life histories. Young adults may form new households after they
get married, and newly formed households were geographically distributed around
their parental households depending on the topographic conditions. Households’
fiielwood use was modeled in the policy submodel (Figure 5.1). Under different
policy scenarios, each household would use different amounts of fuelwood, which
was also determined by household characteristics and topographic conditions. Both

households’ fuelwood use and increased farming of new households affect dynamics

109

 

in forest cover and panda habitat that were modeled in the landscape submodel
(Figure 5.1). Changes in forest cover also depend on topographic conditions. Since
forest cover is one of the most important components of panda habitat that is also
constrained by topographic conditions, dynamics in forest cover may result in
changes in panda habitat.
5. 2.2 Demographic submodel

We obtained characteristics of all households in the reserve and their
corresponding household members, including age, gender, kinship relation, marital
status of household members and amount of cropland of households that are all
available in three possible sources, the 1996 agricultural census (4053 residents in
892 households), the 2000 population census, (43 75 residents in 969 households) and
the 2006 household registration (4505 residents in 1197 households). The geographic
locations and elevations of all households were measured using Global Positioning
System (GPS) receivers.

Population and household dynamics were studied in the reserve (An et a1.
2001, An et al. 2003). Population dynamics was modeled by simulating individual
persons’ life histories in one year increments. Major events of person agents included
married female give birth, students move out of the reserve through going to college,
single persons get married, people move in or out of the reserve through marriage,
grow by one year and die. These events were modeled as stochastic processes based
on person agents’ state that included age, gender, kinship relation, and marital status
(An et al. 2001). The state of household agents included location and elevation of

households, amount of cropland, household size and availability of senior people

110

(>60 years old). When young adults get married, a new household may be formed,
which was set as a stochastic process and depended on the gender of the young adult,
whether the young adult has siblings, the age of the young adult compared to siblings,
and the young adult’s intention of forming a new household (An et al. 2003).

Our demographic submodel adopted previous findings in the reserve (An et al.
2001, An et al. 2003) with the exception that we updated the young adults’
probability of moving out of the reserve through going to college using the 2006
household registration data. On the basis of previous studies in this area, farmlands
were associated with their corresponding households. Since the average farmland area
of households was only 0.28 ha (Chen et al. 2009b), farmland and buildings of each
household were located in the same pixel (90 m by 90 m). Newly formed households
were stochastically located on areas with slopes < 37 degrees and within 800m from
their corresponding parental households; these parameters were based on information
from existing households (An et al. 2005). We assumed that farmland is divided
proportionally to household size when a new household is formed. Major events of
household agents also included change in household size depends on the life history
of each household member and dissolution of a household when there is no member
in the household.
5. 2.3 Policy submodel

We designed three policy scenarios: no payment, cash payment under the
current NF CP, and electricity payment that substitutes cash payment at the same cost.
Under the no payment scenario, households’ fuelwood use was assumed to follow the

fuelwood use pattern prior to the NFCP. Households’ fuelwood use pattern prior to

111

 

the NF CP was modeled on the basis of household size, availability of senior people in

the household, and farmland area (An et al. 2001). On average, each household used

15 m3 of fuelwood per year prior to the NFCP. To understand households’ fuelwood

use patterns under the cash payment and electricity payment, we randomly chose 321
households from a total of 1197 households for in-person interviews (305 valid
interviews, 95% response rate) in the summer of 2006. We chose household heads or
their spouses as interviewees because they are usually the decision-makers of
household affairs. We asked the average amount of daily fuelwood consumption in
the previous year for both the winter season when more fuelwood is consumed and
the summer season when less fuelwood is consumed. Household fuelwood use was
measured as a summation of daily consumption across the year. We also asked about
the amount of fuelwood that each household would demand if the cash payment of
the NF CP was substituted with electricity payment at the price in 2006 (0.18
yuan/kW*h), which leads to an electricity payment of about 14 kW*h per day.

Interviews with 305 households showed an average household fuelwood

consumption of about 9 m3 per year under the current NFCP with cash payment.

Households’ fuelwood use pattern under the cash payment and electricity payment
was modeled on the basis of the household characteristics that are available to all the
households in the reserve (Table 5.1). We corrected for the correlation of households’
responses under the cash payment and the electricity payment using Huber’s variance
correction (Wooldridge 2002). Households’ fuelwood use was significantly positively
correlated to household size because more people in a household usually required

more fuelwood for cooking and heating. Farmland area of households significantly

112

 

positively correlated to households’ fuelwood use because households with more
cropland usually used more crops to feed more pigs, and fuelwood for cooking pig
fodder was an important part of households’ fuelwood use. Elevation of households
also significantly positively correlated to fuelwood use because households living at a
higher elevation usually need more fuelwood for heating in winters than those at
lower elevations due to differences in microclimate.

If the cash payment was substituted with electricity payment, households’

fuelwood use would be reduced by 3.1 m3 per year on average (Table 5.1). Electricity

payment was more efficient than cash payment in reducing fuelwood use because all
the electricity payment would be used to replace fuelwood whereas not all the cash
payment may be used for electricity. Compared to An et al. (2001), availability of
senior people in the households was not significantly correlated to fuelwood use. This
is probably because extra fuelwood use by senior people for heating was reduced
under the NFCP. Although new households were not included in the NFCP, we did
not find significant differences in fuelwood use between new households and other
households partly because all parental households of new households were enrolled in
the NFCP. As the number of new households increases, their firelwood use pattern
can be uncertain if they continue to be excluded from the NFCP.
5. 2.4 Landscape submodel

Forest distributions in Wolong were generated fi'om classification of remotely
sensed imagery (Landsat Thematic Mapper) acquired on June 26, 1994, June 13,
2001 and September 18, 2007. We used an unsupervised classification based on the

ISODATA technique, which is an iterative process for non-hierarchical pixel

113

4.; S m T"

classification (Jensen 1996). A maximum of 1,000 iterations were used for
classification, and produced an output of 100 spectral classes. We then applied a post-
classification sorting method and merged the 100 spectral classes into four
information classes: forest, non-forest, clouds and cloud shadows through a
combination of visual interpretation of these images and information on land cover
obtained from high spatial resolution multispectral imagery (i.e., four IKONOS multi-
spectral scenes (4 x 4 m / pixel) acquired on August 31, October 3, and November 8
and 16 of 2000, respectively and a Quickbird multi-spectral scene (2.4 x 2.4 m / pixel)
acquired on November 23, 2007). A few areas under cloud and cloud shadows were
excluded from further analysis. The accuracy of classifications were assessed using
ground truth points collected during the summers of 1998 (209 points), 2000 (83
points), 2001 (83 points) and 2007 (593 points) that were measured using real-time,
differentially corrected GPS receivers. The overall accuracies of classified forest
distributions were 79.2%, 78.2% and 82.6% for the 1994, 2001 and 2007 imagery,
respectively.

From 1994 to 2001, there was 20.1% forest loss and 12.0% forest recovery,
resulting in 8.6% net deforestation. From 2001 to 2007, there was 12.4% forest loss
and 22.7% forest recovery, resulting in 10.3% net forest regeneration. Forest
dynamics (i.e. forest loss and forest recovery) were analyzed at pixel level, and a 90
m by 90 m pixel size was chosen based on the availability of topographic data and
computational complexity. We randomly selected 4500 pixels, where two-thirds of
the data (3000 pixels) were used for model calibration and one-third of the data (1500

pixels) were used for model validation. Among the 3000 pixels for model calibration,

114

._.-_: . x5:-
5

1982 and 1797 pixels were forest pixels in 1994 and 2001 respectively, pooled to a
total of 3779 forest pixels, and 1018 and 1203 pixels were non-forest pixels in 1994
and 2001 respectively, pooled to a total of 2221 non-forest pixels. We then modeled
forest loss in 1994-2001 or 2001-2007 with 3779 forest pixels and forest recovery in
1994-2001 or 2001-2007 with 2221 non-forest pixels using two logistic regression
models (Table 5.2 and Table 5.3). We corrected for dependencies between pixels that
represented both 1994-2001 and 2001-2007 periods with Huber’s variance correction
(Wooldridge 2002).

We used elevation, slope, aspect [converted into soil moisture classes (Parker
1982)] and distance to forest edge, that were used in previous studies of forest
dynamics (Geoghegan et al. 2001, Nagendra et al. 2003). In addition, we used a
inverse distance weighted fuelwood impact variable that was measured as the
aggregation of the impacts of all eligible households (households within 6-km buffer
from the pixel, Table 5.2 and Table 5.3) on each pixel. The impact of each eligible
household on a pixel was measured as the household’s fuelwood consumption
divided by the distance between the household and the pixel. We also used total
fuelwood use of all households as an explanatory variable. Annual fuelwood use of
households for forest dynamics between 1994 and 2001 were estimated using
household characteristics from 1996 agricultural census data and fuelwood model
developed prior to the NFCP (An et al. 2001), while annual fuelwood use of
households for forest dynamics between 2001 and 2007 were estimated using

household characteristics from 2006 registration data and fuelwood model under the

115

 

NF CP that was developed in the previous section. We did not find multicollinearity
among independent variables.

Although the duration of forest dynamics between 1994 and 2001 was 7 years,
the NFCP enrollment took place in 2000. Therefore we assumed the deforestation
trend before the NFCP was due to fuelwood use in the first 6 years. The duration of
forest dynamics between 2001 and 2007 was also 6 years. Since HANIP was built on
a yearly basis, we approximated the annual dynamics in each pixel by dividing the
estimated probabilities of forest loss and forest recovery from these models by 6 years.
Although topographic variables (elevation, slope, and aspect) of pixels do not change
over time in HANIP, distance to forest edge changes in each year as forest cover
changes. In addition, both fuelwood impact and total fuelwood change in each year
depending on changes in population and households. We estimated probability of
forest loss for forest pixels and probability of forest recovery for non-forest pixels in
each year in HANIP, and determined annual forest dynamics of each pixel with a
Bernoulli trial. The rate parameters of the Bernoulli distributions were the probability
of forest loss or forest recovery.

Pixels may be classified as different levels of panda habitat suitability based
on forest cover and topographic factors. Pixels with forest cover, an elevation
between 1,500 and 3,250 m, and a slope less than 30 degrees were classified as highly
suitable habitat and suitable habitat by Liu et al. (1999), and were combined as habitat
in our model. In HANIP, forest loss in panda habitat due to fuelwood collection and
farming resulted in the loss of panda habitat, while forest recovery in areas with

suitable topographic conditions led to the recovery of panda habitat. Comparisons of

116

 

habitat quantity under different policy scenarios allowed detection ofthe impacts of
conservation investments.
5.2.5 Model validation

Our simulation model was run 30 times under each set of parameters to obtain
results of stochastic processes. Due to the limitation on the data availability, the
demographic model and the landscape model started in different years. For model
,1:

validation, we used 2000 population census data as the starting point for the

demographic model and 2001 land-cover data as the starting point for the landscape

 

model. Our model validation included comparison of simulation results with
empirical data and sensitivity analysis. We evaluated the landscape submodel by
testing regression models for forest loss and forest recovery using a receiver operating
characteristic (ROC) curve (Hanley and Mcneil 1982). The ROC curve is a plot of the
sensitivity values (i.e., true positive fraction) vs. their equivalent l-specificity values
(i.e., false positive fraction) for all possible probability thresholds. The area under the
ROC curve (AUC) is a measure of model accuracy, with AUC values ranging from 0
to 1, where a score of 1 indicates perfect discrimination, a score of 0.5 implies a
prediction that is not better than random, and lower than 0.5 implies a worse than
random prediction. We used the validation data set (i.e. one-third of the 4500
randomly selected pixels) for deriving the AUC value. We also compared the
observed habitat area in 2007 to the mean of predicted habitat areas in 2007 from 30
simulation runs.

For demographic submodel, we compared the observed population size and

number of households in 2006 to the simulation results. Although the validation of

117

the effects of policy scenarios on panda habitat was not feasible, measurement of the
impact of the current cash payment relied on the validation of the landscape submodel,
and the fuelwood use pattern under the circumstance where no payment was provided
and validated in a previous study (An et al. 2001). Finally we conducted sensitivity
analysis to evaluate how sensitive model results were to small changes in several key

model parameters (Haefner 1997). The sensitivity index is defined as Sx =

(AY/Y())/(AX/X0), where X0 is the initial value of a model parameter, AX is a small

change in X, Yo is the initial outcome, and AY is the corresponding change in Y due to

the change in X. Small sensitivity values, suggesting robustness of the outcome to
small changes in parameters, are usually preferred. We also used two-sample t-test to
examine differences in simulation results due to the changes in these parameters.
5.2.6 Simulation experiments

We used 2006 household registration data as the starting point for the
demographic model and 2007 land-cover data as the starting point for the landscape
model and run simulations through 203 0. To demonstrate the conservation effects of
the NF CP and the electricity payment scenario, we also used households’ fuelwood
use pattern prior to the NF CP (An et al. 2001) in the circumstance where no payments
were provided. Since new households are currently not included in the NFCP, their
fuelwood use pattern is uncertain as the number of new households increases in the
fiiture. To explore this uncertainty, we also predicted habitat area under
circumstances where half and all of new households follow the fuelwood use pattern

prior to the NF CP.

118

 

5.3 RESULTS
5.3.1 Model validation

Although accurate prediction of forest dynamics at the pixel level is difficult,
both forest loss and forest recovery models exhibited moderately high accuracy, with

AUC values of 0.775 and 0.773 respectively. Comparisons of model predictions and

observed values showed that predicted mean habitat area in 2007 was 122.77 kmz,

which was close to the observed value (Table 5.4). The difference between the mean

predicted habitat area and the observed habitat area was 0.39 kmz, which was less

than the observed mean yearly change in habitat (2.02 kmz) from 2001 to 2007. The

difference between the predicted mean human population and observed human
population in 2006 was 12, which was also less than the mean yearly population
change (22) from 2000 to 2006. The observed number of households in 2006 was
1197, which were 60 more than the predicted mean households. This difference was
mainly because an unexpectedly large number of new households were formed in

2001, following the implementation of the NFCP, to more effectively capture

conservation subsidies that are distributed on the basis of household (Liu et al. 2007a).

Habitat area was insensitive to small changes in all four selected parameters in

sensitivity analyses (Table 5.5). A 10% increase in juveniles’ college entrance rate

resulted in 0.019 km2 increase in mean habitat (Sx = 0.002), while a 10% increase in

young adult’s intention of forming a new household resulted in 0.063 km2 decrease in

mean habitat (Sx = -0.005). A 50% (400 m) increase in the maximum distance

between a newly formed household and its parental household decreased mean

119

 

habitat by 0.034 km2 (Sx = -0.001). Statistical tests showed that perturbations in these

parameters did not result in significant differences in mean habitat area (Table 5.5). A
10% decrease in the effect of electricity payment (average amount of fuelwood that

can be saved by replacing cash payment with electricity payment) significantly
decreased mean habitat by 0.384 km2 (p = 0.001), even though habitat area was
insensitive to the effect of electricity payment (Sx = 0030). Significant decrease in

mean habitat area due to decrease in the effect of electricity payment was expected

because change in fuelwood was directly related to habitat dynamics in our model.

 

 

5. 3.2 Simulation experiments
Compared to panda habitat and households in 2007 (Figure 5.2), both panda
habitat and households varied under no-payment scenario (Figure 5.3), cash payment

scenario (Figure 5.4) and electricity payment scenario (Figure 5.5). Under the cash
payment of the NFCP, habitat area will be increased from 123.16 km2 in 2007 to
142.07 km2 in 2030, corresponding to an average yearly increase of 0.82 km:2 (Figure

5.6). If cash payment is replaced with electricity payment, the average yearly increase

will be 1.36 kmz, resulting in a total habitat area of 154.49 km2 in 2030. Compared to

the fuelwood use pattern prior to the NFCP under the no-payment scenario, the

effects of cash payment and electricity payment are increasing (Figure 5.6). By 2020,

21.10 km2 (18.3%) and 29.52 km2 (25.5%) of habitat area can be gained, and by 2030,

34.38 km2 (31.9%) and 46.81 km2 (43.5%) of habitat area can be gained through cash

payment and electricity payment, respectively. The increase in habitat area is non-

linear. From 2011 to 2015, the average yearly increase in habitat area is 1.09 km2

120 |

under the cash payment, while the average yearly increase rate reduced to 0.44 km2

from 2025 to 2030. This non-linearity is because of increases in population and
households. In 2015, there will be about 4720 people in about 1340 households in the
reserve, and the population and households will be increased to about 4950 and 1460
by 203 0.

Dynamics in habitat area will also depend on the behavior of newly formed
households (households formed after 2001). The more the new households follow the
fuelwood use pattern prior to the NFCP, the less panda habitat will be gained from the

conservation payment (Figure 5.7). Compared to the fuelwood use pattern under the

no-payment scenario (Figure 5.6), 29.79 km2 (27.7%) and 24.87 km2 (23.1%) of

habitat area can be gained if half and all of new households follow the fuelwood use
pattern prior to the NFCP, which are not much less than if all households follow the
current fiielwood use pattern (Figure 5.7). However, habitat area will start decreasing
in 2028 if all of new households follow the fuelwood use pattern prior to the NFCP.
5.4 DISCUSSION

The trend of dynamics in forest cover in Wolong Nature Reserve has changed
from rapid deforestation between 1994 and 2001 to rapid forest regeneration between
2001 and 2007. Although such an abrupt change could be contributed by multiple
factors, the Natural Forest Conservation Program (N F CP) dramatically reduced local
households’ fuelwood use through increased affordability to electricity use with
conservation payment. While the NFCP has been successful in conserving habitat of
giant pandas and many other wildlife species, its efficiency may be improved under

alternative policy arrangements. In addition, there are uncertainties in the effects of

121

 

such conservation investments due to complex interactions between human and the
environment (Liu et al. 2007a).

We developed a spatially explicit model, HANIP, to study human and natural
interactions under policies. Simulation experiments using HAN [P under different
policy scenarios allowed us to measure the conservation effects of different policies.
By 2030, 31.9% and 43.5% of panda habitat in the study area can be obtained from

cash payment and electricity payment, respectively. Compared to cash payment,

 

conservation payment in the form of electricity is a more direct approach of paying
people to reduce their negative impacts by replacing fuelwood with electricity.
Therefore, electricity payment may improve the efficiency of conservation
investments. As current conservation investments are far below the requirements for
conserving ecosystems globally (James et al. 1999, James et al. 2001), it is important
to improve the efficiency of existing conservation investments. We recognized that
there are potential negative environmental impacts from electricity generation, which
should also be considered for policy implementation.

Through modeling dynamics in population and households, HANIP can also
detect changes in human impacts and policy effects across time. Non-linear increases
in habitat area under the current payment scheme suggested that conservation gains
from PES programs may largely depend on the dynamics in indigenous communities,
such as dynamics in households and population. In addition, there are uncertainties in
conservation gains due to uncertain human responses to policy arrangements. In our
case, the effect of conservation payment may be threatened by the behavior of newly

formed households if they are not included in the program. Uncertainty may also exist

122

 

in the part where not all recovered forest areas, constrained by topographic conditions,
may be used by giant pandas immediately after forest recovery. Future studies may
explore the lagged effects of forest recovery on panda habitat.

Interactions among components in coupled human and natural systems
(CHANS) are complex (Liu et al. 2007a). However, study of these complex
interactions is important for understanding the effects of conservation investments
that may involve complexity (e. g., non-linearity) and uncertainty. By modeling some
of the key dynamic interactions, HANIP can provide important implication to existing
PES programs, such as the NFCP, and conservation investments in the future. The
modeling framework of HANIP may also be used to study conservation policies in

other CHANS.

123

 

 

Table 5.1. Pooled Ordinary Least Squares of fuelwood use pattern.

 

 

Independent variables Parameters Robust SE
Household size 0.441 * 0.241
Farmland (ha) 5.742** 2.522
Elevation (100 m) 0.746**** 0.154
Electricity payment (dummy) -3.097**** 0.267
Constant -8. l 64* * * 3 .074
R-squared 0. 12

 

Dependent variable: fuelwood use (m3; Observations: 610.
Significance: * p $0.1; **p $0.05; *** p $0.01; **** p 30.001.

124

Table 5.2. Pooled logit estimation of forest loss.

 

 

Independent variables Parameters Robust SE
Elevation (100 m) -0.008 0.014
Slope (degree) 0.001 0.006
Aspect (Parker scale) -0.054*** 0.008
Distance to forest edge (m) -0.019*** 0.001

F uelwood impact (m3/m) 0031*“ 0-008
Total fuelwood (1000 m3) 0014*“ 0-002
Constant -1 .279** 0.491

2,2 347.46***

 

Observations: 3779.
Significance: ** p $0.01; *** p $0.001.

 

125

Table 5.3. Pooled logit estimation of forest recovery.

 

Independent variables Parameters Robust SE

 

Elevation -0.008 0.01 1
Slope -0.009 0.006
Aspect 0.064*** 0.010
Distance to forest edge -0.014*** 0.001
Fuelwood impact —0.009 0.008
Total fuelwood -0.016*** 0.002
Constant l.352*** 0.3 84
2,2 263.71 ***

 

Observations: 2221 .
Significance: *** p S 0.001.

126

 

 

Table 5.4. Comparison of model predictions ofpanda habitat, population size and
household number to observed values.

 

 

Factors Observed Observed Model Difference between |Difference] <
value mean yearly mean model mean and observed mean

change observed value yearlychange

Habitat in 123. 16 2.02 122.77 -0.39 Yes

2007 (kmz)

Population 4505 22 4493 -12 Yes

in 2006

Households l 197 38 1 137 -60 No

in 2006

 

127

 

 

Table 5.5. Sensitivity tests for selected model parameters.

 

 

Parameters Default Perturbation Change in t statistic for differences Sensitivity
value habitat area in habitat area (p-value)
(kmz)

College 0.274 +0.0274 0.019 0.169 (0.867) 0.002

entrance rate (10%)

Separate home 0.42 +0.042 -0.063 -0.841 (0.407) -0.005

intention ( 10%)

New household 800 m +400 0034 -0.381 (0.706) -0.001

location (50%)

Electricity 3.] m3 -0.31 -0.384 -3534 (0.001) -0030
Jayment effect ( 10%)

 

128

 

 

Conservation
Policies

 

 

 

 

 

 

People

 

/\

 

 

F uelvvood
Use

 

 

 

 

Topographic
Condition

 

 

 

 

New Household
Formation

 

 

 

 

\ Forest

Figure 5.1. Conceptual framework of the model.

 

 

Cover

 

 

 

 

 

Panda
Habitat

 

 

 

129

7‘7.

 

 

 

 

if HF] Habitat

[T— ] Non-habitat
- Households

 

 

Figure 5.2. Panda habitat and household distribution in 2007.

 

 

130

L3 Habitat

[:J Non-habitat
- Households

 

(a)2020 (b)2030

Figure 5.3. Panda habitat and household distribution in 2020 (a) and 2030 (h) under
no-payment scenario from one run.

131

 

 

[— 3 Habitat

[:3 Non-habitat
- Households

 

(a)2020 ‘ (b)2030

Figure 5.4. Panda habitat and household distribution in 2020 (a) and 2030 (b) under
cash payment scenario from one run.

132

I_] Habitat
I I Non-habitat
- Households

 

(a) 2020 (b) 2030

Figure 5.5. Panda habitat and household distribution in 2020 (a) and 2030 (b) under
electricity payment scenario from one run.

133

 

160 ~ _. ——~ -—-—— e, .. .. . . . 2 --,,,,, ,.

 

 

 

 

 

- - - -No payment ,_.
‘ z
150 - Cash payment ’,/’
/
A i — — Electricity payment], ’z’
E 140
=‘.‘.
“ 1
g 130 . i
E 1
g 120 I
I 1
1
110 » 1
100 i 1 1 1 L I
2005 2010 2015 2020 2025 2030 2035

Year

Figure 5.6. Predicted panda habitat under cash payment, electricity payment, and no
payment scenarios. We did not draw confidence intervals because standard deviations
from 30 runs are small.

134

150 —- ~ ,. . - -W—---— W W ., ,, ,_,-_ L

    
 
 

 

 

 

 

 

 

 

 

 

1
1
_, 1
woe I’LWv” I
as, 130 ~ g —— -' "'
g g _ , L.-. __
a — — All household follow the current fuelwood
E 120 use pattern
5 Half of new households follow the fuelwood
110 use pattern prior to the NFCP
' - - - -All of new households follow the fuelwood
f use pattern priorto the NFCP
100 1 I l - —— 1—‘ _—_—___—l —__ _— _- l _ i
2005 2010 2015 2020 2025 2030 2035
Year

Figure 5.7. Predicted panda habitat under different fuelwood use patterns followed by
new households (households formed after 2001).

135

CHAPTER 6

CONCLUSIONS

136

 

Findings from this dissertation provide important implications to the
management of local environment in Wolong Nature Reserve, China’s current and
future conservation efforts, and global practices of conservation investments through
PES. Using a stated-choice model with a main effects design, we found that social
norms had substantial effects on participants’ intentions of re-enrolling their GTGP
land, suggesting people’s enrollment intentions tend to conform to the majority
(Chapter 2). The extra cost for obtaining an additional unit of land for conservation

would be lower if most people in a community would enroll their land in a PES

 

program. In addition, increased conservation investments can leverage social norms
in communities where most people wound initially not participate. The aggregated
impacts of social norms can be substantial.

This study contributed to the literature by suggesting that social norms should
be integrated with biological values, economic conditions and demographic trends for
efficient conservation investments. In addition, stated-choice model using a main
effects design can relatively easily distinguish the effects of social norms from the
effects of other factors, which was a challenge in the past studies on social norms
(Manski 2000). The effects of social norms can also be substantial in many other
environmental issues, such as common-pool resources management and marketing for
environmentally friendly business. The approach and methods in this study can also
be applied when addressing those environmental issues.

We also found that different sources of income had different effects on
program re-enrollment (Chapter 2). Farming income had a negative effect on program

re-enrollment; however, income from rural-urban labor migrants had a positive effect

137

on program re-enrollment. Although off-farm income may lower farmers’
dependence on the lands that are enrolled in PES programs, off-farm income from
employment within the reserve did not have such an effect. Compared to off-farm
employment outside of the reserve, off-farm employment within the reserve is much
more flexible in terms of labor and time allocation. These results suggested that not
all off-farm income may increase land enrollment in PES programs, and different
types of off-farm employment should be treated differently. The trend of rural-urban

labor migration in transitional economies of many developing countries (United

 

Nations 2004, Korinek et al. 2005) provides a great opportunity for PES programs to
lower costs and sustain the gains from these investments.

Currently conservation investments through PES have been implemented in
many countries. The design and implementation of these PES programs are different.
For instance, both flat payments (all participants paid the same price) and
discriminative payments (participants paid different prices according to opportunity
costs) have been used in PBS programs (Claassen et a1. 2008, Pagiola 2008), resulting
in different levels of efficiency in conservation investments. For the GTGP, there are
only two payment levels nationwide which operate as flat payments within each
region. To demonstrate the potential of improving the efficiency of conservation
investments in the GTGP, we cost-effectively targeted land for maximizing
environmental benefits obtained from the GTGP (Chapter 3). We used environmental
benefits of and cost for lands that were estimated on the basis of land features and

household characteristics. The results suggested that the efficiency of the GTGP can

138

 

be improved up to ten times by switching from the flat payment scheme to cost-
effective targeting in a discriminative payment scheme.

This study made another contribution to the literature by suggesting household
characteristics and regional differences as significant determinants of opportunity
costs of landholders participating in the GTGP (Chapter 3). Therefore, household
characteristics and regional differences should be incorporated with biological values
and physical conditions of lands in the planning of PES programs. Our findings
highlighted the importance of integrating human systems with natural systems for
efficient conservation policy design. In the practice of conservation investments,
however, opportunity costs and many household characteristics of landholders are
often not available to the public. Competitive auctions have successfully been applied
in some PES programs. Cost-effective targeting coupled with competitive auctions
could greatly improve the efficiency of conservation investments through PES and
other conservation policies.

In addition to its direct conservation gains through conversion of agricultural
land to natural vegetation cover, studies on the GTGP suggested that the GTGP has
boosted the trend of rural-urban labor migration by releasing many laborers from
agriculture (Bao et a1. 2005, Liu 2005, Ge et al. 2006, Hu et al. 2006, Uchida et al.
2009). While formal institutions facilitating rural-urban labor migrants have been rare
in rural regions in China and many other countries, social capital is an important
source providing employment information in urban regions (Granovetter 1995, Bian
1 997) and facilitating migration processes (Massey and Espinosa 1997, Korinek et al.

2005). We studied the effects of labor migration on fuelwood consumption from the

139

 

 

perspective of capital substitution (Chapter 4). Social capital, especially weak social
ties, was often used to gain off-farm income through labor migration. Off-farm
income from labor migration then displaced the use of fuelwood (a form of natural
capital) through increased affordability to electricity. Additionally, reduced
population pressure on the local ecosystem and reduced labor supply for fuelwood
collection due to labor migration also resulted in the decreases in fuelwood
consumption.

People in many rural and ecologically significant regions, such as in our study
area, usually have limited access to employment information in urban regions. This
study also contributed to the literature by linking social capital with natural capital
through labor migration. These results suggested policy instruments that aim at
reducing human impacts in ecologically significant regions should provide
employment information and facilitate labor migration (Chapter 4). Numerous off-
farm employment opportunities produced from transitional economies in many
developing countries can be a great opportunity for conservation actions in these
countries.

Even with substantial conservation efforts (e. g., investments), ecosystem
conservation in CHANS presents a formidable challenge. This is at least partly due to
complexity that is inherent to CHANS (Liu et al. 2007a). For instance, people tend to
think linearly, while the effects of conservation investments can be non-linear due to
complex human-natural interactions. In addition, people may respond to conservation

policies differently under different circumstances, resulting in uncertainty in the

140

 

effects of conservation policies. Systems models can be useful tools for understanding
complexity, such as uncertainty and non-linearity, in many CHANS.

The systems model that was developed for this dissertation (Chapter 5)
quantified the effects of the Natural Forest Conservation program (N F CP) and
alternative policy scenarios on panda habitat. Compared to the non-payment scenario,

substantial gains in panda habitat area can be obtained fiom both cash payment under

 

the current NF CP and the electricity payment scenario. Electricity payment, as a more

direct payment approach for reducing human impacts on panda habitat, can improve

 

the efficiency of conservation investments. The conservation gains from conservation 1
payments will decrease as both human population and number of households increase.
Moreover, the effects of conservation payments also depend on the behavior of newly
formed households (households that have been formed after 2000) because these new
households have not been included in the NF CP. Conservation gains from the NFCP
and the electricity payment scheme can be uncertain due to uncertainties in the
behavior of new households.

These outcomes highlighted the importance of integrating dynamic human
society with the changing natural environment in CHANS. By integrating dynamic
human population, households, and activities with land use changes using systems
models, long-term effects of the NF CP can be detected for policy evaluation.
Comparisons among different policy scenarios provide important support to decision-
making of governments and conservation practitioners. Our modeling framework
could also be used in other CHANS especially for evaluating conservation

investments through PES.

141

This dissertation focused on the efficiency and effectiveness of payments for
ecosystem services in China’s Wolong Nature Reserve. Because the objectives in this
research involve both the indigenous community and the natural environment, we
used interdisciplinary methods and tools. Interesting findings have been produced
through this approach. Future research on the efficiency and effectiveness of PES,
such as how pro-environmental social norms can be formed through PES,
understanding the dynamic impacts of social norms and social capital (e. g., weak ties)

using systems models, is needed.

142

LITERATURE CITED

Aide, T. M. and H. R. Grau. 2004. Ecology - Globalization, migration, and Latin
American ecosystems. Science 305:1915-1916.

Airola, J. 2007. The use of remittance income in Mexico. International Migration
Review 41:850-859.

Ajzen, I. 1985. From intentions to actions: A theory of planned behavior. Pages 11-39
in J. Kuhl and J. Beckmann, editors. Action-control: From cognition to
behavior. Springer, Heidelberg.

Ajzen, I. and M. Fishbein. 1980. Understanding Attitudes and Predicting Social
Behavior. Prentice Hall, Englewood Cliffs, NJ.

Alix-Garcia, J., A. De J anvry, and E. Sadoulet. 2008. The role of deforestation risk
and calibrated compensation in designing payments for environmental
services. Environment and Development Economics 13:3 75-3 94.

Allen, J. C. and D. F. Barnes. 1985. The causes of deforestation in developing
countries. Annals of the Association of American Geographers 75:163-184.

An, L., M. Linderman, J. Qi, A. Shortridge, and J. Liu. 2005. Exploring complexity in
a human-environment system: An agent-based spatial model for

multidisciplinary and multiscale integration. Annals of the Association of
American Geographers 95:54-79.

An, L., J. Liu, Z. Ouyang, M. Linderman, S. Zhou, and H. Zhang. 2001. Simulating
demographic and socioeconomic processes on household level and
implications for giant panda habitats. Ecological Modelling 140:31-49.

An, L., F. Lupi, J. Liu, M. A. Linderman, and J. Huang. 2002. Modeling the choice to
switch from fuelwood to electricity Implications for giant panda habitat
conservation. Ecological Economics 42:445-457.

An, L., A. G. Mertig, and J. G. Liu. 2003. Adolescents leaving parental home:
Psychosocial correlates and implications for conservation. Population and
Environment 24:415-444.

Ando, A., J. Camm, S. Polasky, and A. Solow. 1998. Species distributions, land
values, and efficient conservation. Science 279:2126-2128.

Angrist, J. D. and W. N. Evans. 1998. Children and their parents' labor supply:

Evidence from exogenous variation in family size. American Economic
Review 88:450-477.

143

 

Annsworth, P. R., G. C. Daily, P. Kareiva, and J. N. Sanchirico. 2006. Land market
feedbacks can undermine biodiversity conservation. Proceedings of the
National Academy of Sciences of the United States of America 10325403-
5408.

Arrow, K., P. Dasgupta, L. Gouldner, G. C. Daily, P. R. Ehrlich, G. Heal, S. Levin,
K.-G. Miller, S. H. Schneider, D. Starrett, and B. Walker. 2004. Are We
Consuming Too Much? Journal of Economic Perspectives 18:147-172.

Asquith, N. M., M. T. Vargas, and S. Wunder. 2008. Selling two environmental
services: In-kind payments for bird habitat and watershed protection in Los
Negros, Bolivia. Ecological Economics 65:675-684.

Babcock, B. A., P. G. Lakshminarayan, J. J. Wu, and D. Zilberman. 1996. The
economics of a public firnd for environmental amenities: A study of CRP
contracts. American Journal of Agricultural Economics 78:961-971.

Babcock, B. A., P. G. Lakshminarayan, J. J. Wu, and D. Zilberman. 1997. Targeting
tools for the purchase of environmental amenities. Land Economics 73:325-
339.

Balmford, A., K. J. Gaston, S. Blyth, A. James, and V. Kapos. 2003. Global variation
in terrestrial conservation costs, conservation benefits, and unmet
conservation needs. Proceedings of the National Academy of Sciences of the
United States of America 100: 1 046-1050.

Balmford, A., K. J. Gaston, A. S. L. Rodrigues, and A. James. 2000. Integrating costs
of conservation into international priority setting. Conservation Biology
14:597-605.

Bao, J ., D. Tang, and B. Chen. 2005. Socioeconomic effects of Grain to Green
Program in Sichuan Province. Sichuan Forestry Exploration and Designz26-32
(in Chinese).

Becker, C. D. and E. Ostrom. 1995. Human Ecology and Resource Sustainability:
The Importance of Institutional Diversity. Annual Review of Ecology and
Systematics 26: 1 13-133.

Becker, G. S. 1976. The Economic Approach to Human Behavior. University of
Chicago Press, Chicago.

Becker, G. S. 1985. Human-Capital, Effort, and the Sexual Division of Labor. Journal
of Labor Economics 3:S33-858.

144

Bendor, J. and P. Swistak. 2001. The evolution of norms. American Journal of
Sociology 106:1493-1545.

Bernard, H. R., P. Killworth, D. Kronenfeld, and L. Sailer. 1984. The Problem of
Informant Accuracy - the Validity of Retrospective Data. Annual Review of
Anthropology 13:495-517.

Bian, Y. 1997. Bringing strong ties back in: Indirect ties, network bridges, and job
searches in China. American Sociological Review 62:366-3 85.

Bilsborrow, R. E. 2002. Migration, Population Change, and the Rural Environment.

Bilsborrow, R. E. and H. Ogendo. 1992. Population-Driven Changes in Land-Use in
Developing-Countries. Ambio 21:3 7-45.

Bliihdorn, I. 2007. Sustaining the Unsustainable: Symbolic Politics and the Politics
of Simulation. Environmental Politics 16:251-275.

Bltihdom, I. and 1. Welsh. 2007. Eco-politics beyond the paradigm of sustainability:
A conceptual framework and research agenda. Environmental Politics 16:185-
205.

Bourdieu, P. 1986. Forms of Capital. Greenwood Press, Westport, CT.

Braverrnan, H. 1974. Labor and Monopoly Capital. New York, Monthly Review
Press.

Brewer, D. D. 2000. Forgetting in the recall-based elicitation of personal and social
networks. Social Networks 22:29-43.

Brooks, T. M., R. A. Mittermeier, G. A. B. da Fonseca, J. Gerlach, M. Hoffrnann, J. F.
Lamoreux, C. G. Mittermeier, J. D. Pilgrim, and A. S. L. Rodrigues. 2006.
Global biodiversity conservation priorities. Science 313:58-61.

Catton, W. R. J. and R. E. Dunlap. 1978. Environmental Sociology: A New
Paradigm. The American Sociologist 13:41-49.

Chan, K. M. A., M. R. Shaw, D. R. Cameron, E. C. Underwood, and G. C. Daily.
2006. Conservation planning for ecosystem services. Plos Biology 4:2138-
2152.

Chen, X. D., F. Lupi, G. M. He, and J. G. Liu. 2009a. Linking social norms to
efficient conservation investment in payments for ecosystem services.
Proceedings of the National Academy of Sciences of the United States of
America 106:11812-11817.

145

 

Chen, X. D., F. Lupi, G. M. He, Z. Y. Ouyang, and J. G. Liu. 2009b. Factors affecting
land reconversion plans following a payment for ecosystem service program.
Biological Conservation 142:1740-1747.

Cialdini, R. B. and N. J. Goldstein. 2004. Social influence: Compliance and
conformity. Annual Review of Psychology 55:591-621.

Cincotta, R. P., J. Wisnewski, and R. Engelman. 2000. Human population in the
biodiversity hotspots. Nature 404:990-992.

Claassen, R., A. Cattaneo, and R. Johansson. 2008. Cost-effective design of agri-
environmental payment programs: US experience in theory and practice.
Ecological Economics 65:737-752.

Clark, W. C. and N. M. Dickson. 2003. Sustainability science: The emerging research
program. Proceedings of the National Academy of Sciences, USA 100:8059-
8061.

Cohen, J. and P. Cohen. 1983. Applied Multiple Regression/Correlation Analysis for
the Behavioral Sciences L. Erlbaum Associates, Hillsdale, NJ.

Coleman, J. S. 1990. Foundations of social theory Harvard University Press,
Cambridge, Mass.

Cooper, J. C. and C. T. Osborn. 1998. The effect of rental rates on the extension of
conservation reserve program contracts. American Journal of Agricultural
Economics 80: 1 84-194.

Copas, J. B. and H. G. Li. 1997. Inference for non-random samples. Journal of the
Royal Statistical Society Series B-Methodological 59:55-77.

Costanza, R., R. dArge, R. deGroot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S.
Naeem, R. V. Oneill, J. Paruelo, R. G. Raskin, P. Sutton, and M. vandenBelt.
1997. The value of the world's ecosystem services and natural capital. Nature
387 :253-260.

Curran, L. M., S. N. Trigg, A. K. McDonald, D. Astiani, Y. M. Hardiono, P. Siregar,
I. Caniago, and E. Kasischke. 2004. Lowland forest loss in protected areas of
Indonesian Borneo. Science 303: 1000-1003.

Daily, G. C., editor. 1997. Nature's services: Societal dependence on natural
ecosystems. Island Press, Washington, DC, USA.

Daily, G. C., T. Soderqvist, S. Aniyar, K. Arrow, P. Dasgupta, P. R. Ehrlich, C. Folke,
A. Jansson, B. O. Jansson, N. Kautsky, S. Levin, J. Lubchenco, K. G. Maler,

146

 

 

 

D. Simpson, D. Starrett, D. Tilman, and B. Walker. 2000. Ecology - The value
of nature and the nature of value. Science 289:395-396.

Dasgupta, P. in press. The place of nature in economic development.in D. Rodrik and
M. Rosenzweig, editors. Handbook of Development Economics, Vol. 5. North
Holland, Amsterdam.

De Jong, G. F. 2000. Expectations, gender, and norms in migration decision-making.
Population Studies-a Journal of Demography 54:307-319.

Deadman, P., D. Robinson, E. Moran, and E. Brondizio. 2004. Colonist household
decisionmaking and land-use change in the Amazon Rainforest: an agent-
based simulation. Environment and Planning B-Planning & Design 31:693-
709.

Dietz, R. D. 2002. The estimation of neighborhood effects in the social sciences: An
interdisciplinary approach. Social Science Research 31 :539-575.

Dietz, T. and A. D. Henry. 2008. Context and the commons. Proceedings of the
National Academy of Sciences 105:13189-13190.

Dietz, T., E. Ostrom, and P. C. Stern. 2003. The struggle to govern the commons.
Science 302:1907-1912.

Dietz, T., E. A. Rosa, and R. York. 2007. Driving the Human Ecological Footprint.
Frontiers in Ecology and Environment 5:13-18.

Dietz, T., E. A. Rosa, and R. York. 2008. Environmentally Efficient Well-Being:
Rethinking Sustainability as the Relationship between Human Well-being and
Environmental Impacts. Human Ecology Review 16:113-122.

Dunlap, R. E. 1994. Struggling with Human Exemptionalism: The Rise, Decline and
Revitalization of Environmental Sociology. The American Sociologist 25:5-
30.

Dunn, C. P., F. Steams, G. R. Guntenspergen, and D. M. Sharpe. 1993. Ecological
Benefits of the Conservation Reserve Program. Conservation Biology 7: 1 32-
139.

Economy, E. C. 2004. The River Runs Black: The Environmental Challenge to
China's Future. Cornell University Press, Ithaca, New York.

Efron, B. and R. J. Tibshirani. 1993. An Introduction to the Bootstrap. Chapman &
Hall, New York.

147

 

 

Ehrhardt-Martinez, K. 1998. Social determinants of deforestation in developing
countries: A cross-national study. Social Forces 77:567-586.

Ehrlich, P. R. and S. A. Levin. 2005. The evolution of norms. Plos Biology 32943-948.

Elster, J. 1989. SOCIAL NORMS AND ECONOMIC-THEORY. Journal of
Economic Perspectives 3:99-1 17.

Engelbrecht, H.-J. 2009. Natural capital, subjective well-being, and the new welfare
economics of sustainability: Some evidence from cross-country regressions.
Ecological Economics In press.

F an, C. C. 2003. Rural-urban migration and gender division of labor in transitional
China. International Journal of Urban and Regional Research 27:24-+.

F eamside, P. M. 2005. Deforestation in Brazilian Amazonia: History, rates, and
consequences. Conservation Biology 19:680-688.

Feld, S. L. and W. C. Carter. 2002. Detecting measurement bias in respondent reports
of personal networks. Social Networks 24:365-383.

Ferraro, P. J. 2001. Global habitat protection: Limitations of development
interventions and a role for conservation performance payments. Conservation
Biology 15:990-1000.

Ferraro, P. J. 2003. Assigning priority to environmental policy interventions in a
heterogeneous world. Journal of Policy Analysis and Management 22:27-43.

Ferraro, P. J. 2004. Targeting conservation investments in heterogeneous landscapes:
A distance-function approach and application to watershed management.
American Journal of Agricultural Economics 86:905-918.

Ferraro, P. J. 2008. Asymmetric information and contract design for payments for
environmental services. Ecological Economics 65:810-821.

Ferraro, P. J. and A. Kiss. 2002. Direct payments to conserve biodiversity. Science
298:1718-1719.

Ferraro, P. J. and S. K. Pattanayak. 2006. Money for nothing? A call for empirical
evaluation of biodiversity conservation investments. Plos Biology 4:482-488.

Ferraro, P. J. and R. D. Simpson. 2002. The cost-effectiveness of conservation
payments. Land Economics 78:339-353.

Fishbein, M. and I. Ajzen. 1975. Belief, Attitude, Intention and Behavior: An
Introduction to Theory and Research. Addison-Wesley, Reading, MA.

148

 

 

 

Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S.
Chapin, M. T. Coe, G. C. Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway,
E. A. Howard, C. J. Kucharik, C. Monfreda, J. A. Patz, I. C. Prentice, N.
Ramankutty, and P. K. Snyder. 2005. Global consequences of land use.
Science 309:570-574.

Foster, J. B. 1999. Marx's Theory of Metabolic Rift: Classical Foundations for
Enviromnental Sociology. American Journal of Sociology 105:366-405.

Frank, D. J., A. Hironaka, and E. Schofer. 2000. The nation-state and the natural
environment over the twentieth century. American Sociological Review
65:96-116.

Frank, K. A. 2000. Impact of a confounding variable on a regression coefficient.
Sociological Methods & Research 29:147-194.

Garip, F. 2008. Social Capital and Migration: How Do Similar Resources Lead to
Divergent Outcomes? Demography 45:591-617.

Ge, W., L. Li, and Y. Li. 2006. Analysis on the sustainability about Grain to Green
Program — A field survey of Grain to Green Program in Wuqi and Zhidan
counties of Shaanxi. Forestry Economicsz33-49 (in Chinese).

Geoghegan, J., S. C. Villar, P. Klepeis, P. M. Mendoza, Y. Ogneva-Himmelberger, R.
R. Chowdhury, B. L. Turner, and C. Vance. 2001. Modeling tropical
deforestation in the southern Yucatan peninsular region: comparing survey
and satellite data. Agriculture Ecosystems & Environment 85:25-46.

Goldstein, A., M. White, and S. Goldstein. 1997. Migration, fertility, and state policy
in Hubei Province, China. Demography 34:481-491.

Granovetter, M. 1973. Strength Of Weak Ties. American Journal of Sociology
78:1360-1380.

Granovetter, M. 1985. Economic-Action and Social-Structure: The Problem of
Embeddedness. American Journal of Sociology 91:481-510.

Granovetter, M. 1995. Getting a Job: A Study of Contacts and Careers. 2nd edition.
University of Chicago Press, Chicago, IL.

Green, R. E., S. J. Cornell, J. P. W. Scharlemann, and A. Balmford. 2005. Farming
and the fate of wild nature. Science 307 1550-555.

Greene, W. H. 2003. Econometric analysis. 5th edition. Prentice Hall, Upper Saddle
River, NJ.

149

 

 

 

Haefner, J. W. 1997. Modeling biological systems: principles and applications.
Chapman & Hall, New York, USA.

Hanley, J. A. and B. J. Mcneil. 1982. The Meaning and Use of the Area under a
Receiver Operating Characteristic (Roc) Curve. Radiology 143:29-36.
Hardin, G. 1968. Tragedy of Commons. Science 16221243-1248.

He, G. M. 2008. Balancing Human Energy Needs and Conservation of Panda Habitat.
Dissertation. Michigan State University.

Heckman, J. J. 2005. The scientific model of causality. Pages 1-97 Sociological
Methodology 2005, Vol 35.

Hedayat, A. S., N. J. A. Sloane, and J. Stufken. 1999. Orthogonal Arrays: Theory and
Applications Springer, New York.

Henry, A. D. 2009. The Challenge of Learning for Sustainability: A Prolegomenon to
Theory. Human Ecology Review 16:131-140.

Hirano, K. and G. W. Imbens. 2002. Estimation of Causal Effects using Propensity
Score Weighting: An Application to Data on Right Heart Catheterization.
Health Services & Outcomes Research Methodology 2:259-278.

Hirano, K., G. W. Imbens, and G. Ridder. 2003. Efficient estimation of average
treatment effects using the estimated propensity score. Econometrica 71:1161-
1189.

Holland, P. W. 1989. Choosing among Alternative Nonexperimental Methods for
Estimating the Impact of Social Programs - the Case of Manpower Training -
Comment. Journal Of The American Statistical Association 84:875-877.

Hu, C., B. Fu, and L. Chen. 2006. Impacts of "Grain for Green Project" on agriculture
and rural economics development in the loess hilly and gully area — A case
study in Ansai county Journal of Arid Land Resources and Environment
20:67-72 (in Chinese).

Hu, H. and G. Liu. 2006. Carbon sequestration of China's Natural Forest
Conservation Program. Acta Ecologica Sinica 26:291-296 (in Chinese).

Hubacek, K. and J. van den Bergh. 2006. Changing concepts of 'land' in economic
theory: From single to multi-disciplinary approaches. Ecological Economics
5615-27.

Hugo, G. 1998. The Demographic Underpinnings of Current and Future International
Migration in Asia. Asian and Pacific Migration Journal 721-25.

150

 

International Union for the Conservation of Nature. 1980. World Conservation
Strategy. International Union for the Conservation of Nature, Gland,
Switzerland.

Jack, B. K., C. Kousky, and K. R. E. Sims. 2008. Designing payments for ecosystem
services: Lessons from previous experience with incentive-based mechanisms.
Proceedings of the National Academy of Sciences of the United States of
America 105:9465-9470.

James, A., K. J. Gaston, and A. Balmford. 2001. Can we afford to conserve
biodiversity? Bioscience 51:43-52.

James, A. N., K. J. Gaston, and A. Balmford. 1999. Balancing the Earth's accounts.
Nature 401:323-324.

Jaynes, E. T. 1957. Information Theory and Statistical Mechanics. The Physical
Review 106:620-630.

Jensen, J. R. 1996. Introductory digital image processing, a remote sensing
perspective. Second edition. Prentice Hall, New Jersey.

Johnson, D. H. and M. D. Schwartz. 1993. The Conservation Reserve Program and
Grassland Birds. Conservation Biology 7:934-93 7.

Johnson, P. N., S. K. Misra, and R. T. Ervin. 1997. A Qualitative Choice Analysis of
Factors Influencing Post-CRP Land Use Decisions. Journal of Agricultural
and Applied Economics 29:163-173.

Kates, R. W., W. C. Clark, R. Corell, J. M. Hall, C. Jaeger, I. Lowe, J. J. McCarthy,
H.-J. Schellnhuber, B. Bolin, N. M. Dickson, S. Faucheux, G. C. Gallopin, A.
Grubler, B. Huntley, J. J ager, N. S. Jodha, R. E. Kasperson, A. Mabogunje, P.
Matson, H. Mooney, B. 1. Moore, T. O'Riordan, and U. Svedin. 2001.
Sustainability Science. Science 292:641-642.

Khanna, M., W. H. Yang, R. Famsworth, and H. Onal. 2003. Cost-effective targeting
of land retirement to improve water quality with endogenous sediment
deposition coefficients. American Journal of Agricultural Economics 85:53 8-
553.

Kirwan, B., R. N. Lubowski, and M. J. Roberts. 2005. How cost-effective are land
retirement auctions? Estimating the difference between payments and
willingness to accept in the conservation reserve program. American Journal
of Agricultural Economics 87: 1239-1247.

151

 

 

Koc, I. and I. Onan. 2004. International mi grants' remittances and welfare status of
the left-behind families in Turkey. International Migration Review 38:78-112.

Korinek, K., B. Entwisle, and A. Jampaklay. 2005. Through thick and thin: Layers of
social ties and urban settlement among Thai migrants. American Sociological
Review 70:779-800.

Langpap, C. 2004. Conservation incentives programs for endangered species: An
analysis of landowner participation. Land Economics 80:375-388.

Latacz-Lohmann, U. and C. Van der Hamsvoort. 1997. Auctioning conservation
contracts: A theoretical analysis and an application. American Journal of
Agricultural Economics 79:407-418.

Levin, S. A. 2006. Learning to live in a global commons: socioeconomic challenges
for a sustainable environment. Ecological Research 21:328-333.

Li, D., F. Bo, and J. Tao. 2006. Achievements in and strategies for the Grain to Green
Program in Hunan Province. Hunan Forestry Science & Technology 33:1-5
(in Chinese).

Li, H. Z. and S. Zahniser. 2002. The determinants of temporary rural-to-urban
migration in China. Urban Studies 39:2219-2235.

Liang, W., C. Bai, B. Sun, D. Hao, and J. Qi. 2006. Soil moisture and physical
properties of regions under Grain to Green Program in Gullied Rolling Loess
Area. Soil and Water Conservation in Chinazl7-18 (in Chinese).

Liang, Z. 2001. The age of migration in China. Population and Development Review
27:499-+.

Lin, N. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge
University Press, New York.

Lin, N., W. M. Ensel, and J. C. Vaughn. 1981. Social Resources And Strength Of
Ties - Structural Factors In Occupational-Status Attainment. American
Sociological Review 46:393-403.

Lindbeck, A. 1997. Incentives and social norms in household behavior. American
Economic Review 87:370-377.

Lindbeck, A., S. Nyberg, and J. W. Weibull. 1999. Social norms and economic
incentives in the welfare state. Quarterly Journal of Economics 114:1-35.

152

Linderman, M. A., L. An, S. Bearer, G. M. He, Z. Y. Ouyang, and J. G. Liu. 2005.
Modeling the spatio-temporal dynamics and interactions of households,
landscapes, and giant panda habitat. Ecological Modelling 183:47-65.

Liu, F., C. Huang, T. He, X. Qian, Y. Liu, and H. Luo. 2002. Roles of Grain to Green
Program in reducing loss of phosphorus from yellow soil in hilly areas.
Journal of Soil Water Conservation 16:20-23 (in Chinese).

Liu, J. and J. Diamond. 2008. Revolutionizing China's Environmental Protection.
Science 319237-38.

Liu, J ., Z. Ouyang, W. Taylor, R. Groop, Y. Tan, and H. Zhang. 1999. A framework
for evaluating effects of human factors on wildlife habitats: The case on the
giant pandas. Conservation Biology 13:1360-1370.

Liu, J. G., G. C. Daily, P. R. Ehrlich, and G. W. Luck. 2003a. Effects of household
dynamics on resource consumption and biodiversity. Nature 421:530-533.

Liu, J. G. and J. Diamond. 2005. China's environment in a globalizing world. Nature
435:1179-1186.

Liu, J. G., T. Dietz, S. R. Carpenter, M. Alberti, C. Folke, E. Moran, A. N. Pell, P.
Deadman, T. Kratz, J. Lubchenco, E. Ostrom, Z. Ouyang, W. Provencher, C.
L. Redman, S. H. Schneider, and W. W. Taylor. 2007a. Complexity of
coupled human and natural systems. Science 317: 1 513-15 1 6.

Liu, J. G., T. Dietz, S. R. Carpenter, C. Folke, M. Alberti, C. L. Redman, S. H.
Schneider, E. Ostrom, A. N. Pell, J. Lubchenco, W. W. Taylor, Z. Y. Ouyang,
P. Deadman, T. Kratz, and W. Provencher. 2007b. Coupled human and natural
systems. Ambio 36:639-649.

Liu, J. G., S. X. Li, Z. Y. Ouyang, C. Tam, and X. D. Chen. 2008. Ecological and
socioeconomic effects of China's policies for ecosystem services. Proceedings
of the National Academy of Sciences of the United States of America
105:9477-9482.

Liu, J. G., M. Linderman, Z. Y. Ouyang, L. An, J. Yang, and H. M. Zhang. 2001.
Ecological degradation in protected areas: The case of Wolong Nature
Reserve for giant pandas. Science 292:98-101.

Liu, J. G., Z. Y. Ouyang, S. L. Pimm, P. H. Raven, X. K. Wang, H. Miao, and N. Y.
Han. 2003b. Protecting China's biodiversity. Science 300: 1240-1241.

Liu, K. 2005. Analysis on the Prospect after Grain for Green Subsidy Policy. Green
Chinaz30-31 (in Chinese).

153

Long, H. L., G. K. Heilig, J. Wang, X. B. Li, M. Luo, X. Q. Wu, and M. Zhang. 2006.
Land use and soil erosion in the upper reaches of the Yangtze River: Some
socio-economic considerations on China's Grain-for-Green Programme. Land
Degradation & Development 17:589-603.

Loucks, C. J ., Z. Lu, E. Dinerstein, H. Wang, D. M. Olson, C. Q. Zhu, and D. J.
Wang. 2001. Giant pandas in a changing landscape. Science 294: 1465-1465.

Louviere, J. J ., D. A. Hensher, and J. D. Swait. 2000. Stated choice methods : analysis
and applications Cambridge University Press, Cambridge, UK.

Lu, D. and Q. Weng. 2007. A survey of image classification methods and techniques
for improving classification performance. International Journal of Remote
Sensing 28:823-870.

Luck, G. W., T. H. Ricketts, G. C. Daily, and M. Imhoff. 2004. Alleviating spatial
conflict between people and biodiversity. Proceedings of the National
Academy of Sciences of the United States of America 101:182-186.

Ma, Y. and S. Fan. 2005. Ecological-economic effects of Grain to Green Program in
desertification areas. Journal of Natural Resources 20:590-596 (in Chinese).

Madden, T. J ., P. S. Ellen, and I. Ajzen. 1992. A Comparison of the Theory of
Planned Behavior and the Theory of Reasoned Action. Personality and Social
Psychology Bulletin 18:3-9.

Manski, C. F . 2000. Economic analysis of social interactions. Journal of Economic
Perspectives 14:115-136.

Manson, S. M. and T. Evans. 2007. Agent-based modeling of deforestation in
southern Yucatan, Mexico, and reforestation in the Midwest United States.
Proceedings of the National Academy of Sciences of the United States of
America 104:20678-20683.

Marsden, P. V. 1990. Network Data and Measurement. Annual Review of Sociology
16:435-463.

Massey, D. S. 1990. Social-Structure, Household Strategies, And The Cumulative
Causation Of Migration. Population Index 5623-26.

Massey, D. S. and F. G. Espana. 1987. The Social-Process of International Migration.
Science 237 :733-73 8.

Massey, D. S. and K. E. Espinosa. 1997. What's driving Mexico-US migration? A

theoretical, empirical, and policy analysis. American Journal of Sociology
102:939-999.

154

Matthews, R. B., N. G. Gilbert, A. Roach, J. G. Polhill, and N. M. Gotts. 2007.
Agent-based land-use models: a review of applications. Landscape Ecology
22:1447-1459.

McMaster, D. G. and S. K. Davis. 2001. An evaluation of Canada's Permanent Cover
Program: Habitat for grassland birds? Journal of Field Ornithology 72:195-
210.

Millennium Ecosystem Assessment. 2005. Ecosystems and human well-being. Island
Press, Washington, DC

Mittermeier, R. A., N. Myers, J. B. Thomsen, G. A. B. da Fonseca, and S. Olivieri.
1998. Biodiversity hotspots and major tropical wilderness areas: Approaches
to setting conservation priorities. Conservation Biology 12:516-520.

Morgan, S. L. and D. J. Harding. 2006. Matching estimators of causal effects -
Prospects and pitfalls in theory and practice. Sociological Methods &
Research 35:3-60.

Morton, D. C., R. S. DeFries, Y. E. Shimabukuro, L. 0. Anderson, E. Arai, F. D.
Espirito-Santo, R. Freitas, and J. Morisette. 2006. Cropland expansion
changes deforestation dynamics in the southern Brazilian Amazon.
Proceedings of the National Academy of Sciences of the United States of
America 103:14637-14641.

Munoz-Pina, C., A. Guevara, J. M. Torres, and J. Brana. 2008. Paying for the
hydrological services of Mexico's forests: Analysis, negotiations and results.
Ecological Economics 65:725-736.

Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, and J. Kent.
2000. Biodiversity hotspots for conservation priorities. Nature 403:853-858.

Nagendra, H., J. Southworth, and C. Tucker. 2003. Accessibility as a determinant of
landscape transformation in western Honduras: linking pattern and process.
Landscape Ecology 18:141-158.

Nagubadi, V., K. T. McNamara, W. L. Hoover, and W. L. Mills. 1996. Program
Participation Behavior of Nonindustrial Forest Landowners: A Probit Analysis.
Journal of Agricultural and Applied Economics 28:323-336.

Naidoo, R. and W. L. Adamowicz. 2005. Economic benefits of biodiversity exceed
costs of conservation at an African rainforest reserve. Proceedings of the
National Academy of Sciences of the United States of America 102:16712-
16716.

155

Naidoo, R., A. Balmford, P. J. Ferraro, S. Polasky, T. H. Ricketts, and M. Rouget.
2006. Integrating economic costs into conservation planning. Trends in
Ecology & Evolution 21:681-687.

Neyer, F. J. 1997. Free recall or recognition in collecting egocentered networks: The
role of survey techniques. Journal of Social and Personal Relationships
14:305-316.

Norton, B. G. 2005. Sustainability: A Philosophy of Adaptive Ecosystem
Management. Oxford University Press, New York.

Nyborg, K. and M. Rege. 2003. On social norms: the evolution of considerate
smoking behavior. Journal of Economic Behavior & Organization 52:323-340.

O'Connor, C., M. Marvier, and P. Kareiva. 2003. Biological vs. social, economic and
political priority-setting in conservation. Ecology Letters 6:706-711.

Odling-Smee, L. 2005. Dollars and sense. Nature 437:614-616.

OECD. 1997. The Environmental Effects of Agricultural Land Diversion Schemes.
Organization for Economic Cooperation and Development (OECD), Paris.

Osborn, T., S. Brownback, and M. Schamberger. 1993. The Conservation Reserve
Program - Status, Future, and Policy Options. Journal of Soil and Water
Conservation 48:271-278.

Ostrom, E. 2000. Collective action and the evolution of social norms. Journal of
Economic Perspectives 14: 137-158.

Ostrom, E., T. Dietz, N. Dolsak, P. C. Stem, S. Stonich, and E. Weber, editors. 2002.
The Drama of the Commons. National Academy Press, Washington, DC.

Pagiola, S. 2008. Payments for environmental services in Costa Rica. Ecological
Economics 65:712-724.

Palloni, A., D. S. Massey, M. Ceballos, K. Espinosa, and M. Spittel. 2001. Social
capital and international migration: A test using information on family

networks. American Journal of Sociology 106:1262-1298.

Pan, W. and K. A. Frank. 2003. A probability index of the robustness of a causal
inference. Journal of Educational and Behavioral Statistics 28:3 1 5-33 7.

Parker, A. J. 1982. The topographic relative moisture index: An approach to soil-
moisture assessment in mountain terrain. Physical Geography 3: 160-168.

156

Parker, D. C., S. M. Manson, M. A. Janssen, M. J. Hoffmann, and P. Deadman. 2003.
Multi-agent systems for the simulation of land-use and land-cover change: A
review. Annals of the Association of American Geographers 93:314-337.

Parks, P. J. and J. P. Schorr. 1997. Sustaining open space benefits in the northeast: An
evaluation of the conservation reserve program. Journal of Environmental
Economics and Management 32:85-94.

Pfeffer, M. J. and P. A. Parra. 2009. Strong Ties, Weak Ties, and Human Capital:
Latino Immigrant Employment Outside the Enclave. Rural Sociology 74:241-
269.

Phillips, S. J ., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling
of species geographic distributions. Ecological Modelling 190:231-259.

Polasky, S., E. Nelson, E. Lonsdorf, P. Fackler, and A. Starfield. 2005. Conserving
species in a working landscape: Land use with biological and economic
objectives. Ecological Applications 15: 1387-1401.

Portes, A. 1998. Social Capital: Its origins and applications in modern sociology.
Annual Review of Sociology 24:1-24.

Presser, S., M. P. Couper, J. T. Lessler, E. Martin, J. Martin, J. M. Rothgeb, and E.
Singer. 2004. Methods for testing and evaluating survey questions. Public
Opinion Quarterly 68: 109-130.

Pretty, J. 2003. Social capital and the collective management of resources. Science
302:1912-1914.

Robins, J. M. and A. Rotnitzky. 1995. Semiparametric Efficiency In Multivariate
Regression-Models With Missing Data. Journal Of The American Statistical
Association 90: 122-129.

Robins, J. M., A. Rotnitzky, and D. Scharfstein. 2000. Sensitivity Analysis for
Selection Bias and Unmeasured Confounding in Missing Data and Causal
Inference Models. Pages 1-94 in E. Halloran and D. Berry, editors. Statistical
Models in Epidemiology: The Environment and Clinical Trials. Springer-
Verlag, New York.

Rockwood, L. L., R. E. Stewart, and T. Dietz, editors. 2008. Foundations of
Environmental Sustainability: The Coevolution of Science and Policy.

Oxford University Press, Oxford.

Rosa, E. A. and T. Dietz. 1998. Climate change and society - Speculation,
construction and scientific investigation. International Sociology 13:421-455.

157

Rosenbaum, P. R. and D. B. Rubin. 1983. The Central Role of the Propensity Score in
Observational Studies for Causal Effects. Biometrika 70:41-55.

Rudel, T. and J. Roper. 1997. The paths to rain forest destruction: Crossnational
patterns of tropical deforestation, 1975-90. World Development 25:53-65.

Rudel, T. K. 1989. Population, development, and tropical deforestation - A cross-
national study. Rural Sociology 54:327-338.

Rudel, T. K., D. Bates, and R. Machinguiashi. 2002. A tropical forest transition?
Agricultural change, out-migration, and secondary forests in the Ecuadorian
Amazon. Annals of the Association of American Geographers 92:87-102.

Schaller, G. B., J. Hu, W. Pan, and J. Zhu. 1985. The Giant pandas of Wolong. The
University of Chicago Press, Chicago.

Scharfstein, D. O. and R. A. Irizarry. 2003. Generalized additive selection models for
the analysis of studies with potentially nonignorable missing outcome data.
Biometrics 59:601-613.

Schultz, P. W. and S. Oskamp. 1996. Effort as a moderator 0f the attitude-behavior
relationship: General environmental concern and recycling. Social Psychology
Quarterly 59:375-383.

Sethi, R. and E. Somanathan. 1996. The evolution of social norms in common
property resource use. American Economic Review 86:766-788.

Shadish, W. R., T. D. Cook, and D. T. Campbell. 2002. Experimental and Quasi-
Experimental Designs for Generalized Causal Inference. Houghton Mif’flin,
Boston.

Shelton, B. A. and D. John. 1996. The division of household labor. Annual Review of
Sociology 22:299-322.

Sierra, R. and E. Russman. 2006. On the efficiency of environmental service
payments: A forest conservation assessment in the Osa Peninsula, Costa Rica.
Ecological Economics 59:1 31-141 .

Siikamaki, J. and D. F. Layton. 2007. Potential cost-effectiveness of incentive
payment programs for the protection of non-industrial private forests. Land
Economics 83:539-560.

Skaggs, R. K., R. E. Kirksey, and W. M. Harper. 1994. Determinants and

Implications of Post-Crp Land-Use Decisions. Journal of Agricultural and
Resource Economics 19:299-312.

158

Smith, R. B. W. 1995. The Conservation Reserve Program as a Least-Cost Land

Retirement Mechanism. American Journal of Agricultural Economics 77:93-
105.

Smith, R. J ., R. D. J. Muir, M. J. Walpole, A. Balmford, and N. Leader-Williams.
2003. Governance and the loss of biodiversity. Nature 426:67-70.

State Forestry Administration of China. 2005-2007. China Forestry Development
Report. China Forestry Publishing House, Beijing (in Chinese).

Tao, Y., Y. Liu, N. Song, Z. Ma, and C. Luo. 2006. The comparative analysis: The
effect of conversion of cropland to forest and grassland on the income of
farmers. Journal of Arid Land Resources and Environment 20:36-42 (in
Chinese).

Terry, D. J. and M. A. Hogg. 1996. Group norms and the attitude-behavior
relationship: A role for group identification. Personality and Social
Psychology Bulletin 22:776-793.

Tole, L. 1998. Sources of deforestation in tropical developing countries.
Environmental Management 22: 19-33.

US. Department of Agriculture. 2001. Revised Universal Soil Loss Equation Version
2 (RUSLE2) Handbook find/ftp-
fc.sc.egov.usda.gov/MD/webtdocuments/technical/RUSLE2/RUSL-E2_Handb

ookpdf

US. National Research Council. 1999. Our Common Journey: A Transition Toward
Sustainability. National Academy Press, Washington, DC.

Uchida, E., S. Rozelle, and J. T. Xu. 2009. Conservation Payments, Liquidity
Constraints, and Off-F arm Labor: Impact of the Grain-for-Green Program on
Rural Households in China. American Journal of Agricultural Economics
91:70-86.

Uchida, E., J. T. Xu, and S. Rozelle. 2005. Grain for green: Cost-effectiveness and
sustainability of China's conservation set-aside program. Land Economics
81:247-264.

United Nations. 2004. World Urbanization Prospects: The 2003 Revision. The United
Nations, New York.

Vifia, A., S. Bearer, X. D. Chen, G. M. He, M. Linderman, L. An, H. M. Zhang, Z. Y.
Ouyang, and J. G. Liu. 2007. Temporal changes in giant panda habitat
connectivity across boundaries of Wolong Nature Reserve, China. Ecological
Applications 17:1019-1030.

159

Vincent, J. R. 2007. Spatial dynamics, social norms, and the opportunity of the
commons. Ecological Research 22:3-7.

Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997. Human
domination of Earth's ecosystems. Science 277:494-499.

Wackemagel, M., N. B. Schulz, D. Deumling, A. C. Linares, M. Jenkins, V. Kapos, C.
Monfreda, J. Loh, N. Myers, R. Norgaard, and J. Randers. 2002. Tracking the
ecological overshoot of the human economy. Proceedings of the National
Academy of Sciences of the United States of America 99:9266-9271.

Walker, R. 2008. Forest transition: Without complexity, without scale. Professional
Geographer 60: 136-140.

Wang, Z., X. Wang, Y. Shi, L. Pan, X. Yu, and Z. Tang. 2007. Effects of Grain to
Green Program on soil and water conservation in Zigui county of the Three
Gorges Reservoir Region. Science of Soil and Water Conservation 5:68-72 (in
Chinese).

Wiley, E. O., K. M. Mchset, A. T. Peterson, R. CR, and A. M. Stewart. 2003.
Niche modeling and geographic range predictions in the marine environment
using a machine-learning algorithm. Oceanography 16:120-127.

Wilson, K. A., M. F. McBride, M. Bode, and H. P. Possingham. 2006. Prioritizing
global conservation efforts. Nature 440:337-340.

Wilson, T. D. 1998. Weak ties, strong ties: Network principles in Mexican migration.
Human Organization 57:394-403.

Winship, C. and S. L. Morgan. 1999. The estimation of causal effects from
observational data. Annual Review of Sociology 25:659-706.

Wolong Nature Reserve. 2005. History of the Development of Wolong Nature
Reserve (in Chinese). Sichuan Science Publisher, Chengdu.

Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data. MIT
Press, Cambridge, MA.

Wooldridge, J. M. 2003. Introductory econometrics : a modern approach. South-
Westem College Pub., Cincinnati, Ohio.

World Bank. 2001. China: Air, land, and Water: Enviromnental Priorities for a New
Millennium, Washington, DC: World Bank.

160

World Commission on Environment and Development. 1987. Our Common Future.
Oxford University Press, Oxford, England.

Wright, E. O. 1979. Class Structure and Income Determination. Academic Press,
New York.

Wunder, S. 2007. The efficiency of payments for environmental services in tropical
conservation. Conservation Biology 21:48-58.

Wunder, S. 2008. Payments for environmental services and the poor: concepts and
preliminary evidence. Environment and Development Economics 13:279-297.

Wunder, S. and M. Alban. 2008. Decentralized payments for environmental services:
The cases of Pimampiro and PROFAF OR in Ecuador. Ecological Economics
65:685-698.

Xu, J. and Y. Cao. 2002. On sustainability of converting farmland to
forests/ grasslands. International Economic Review Z2z56-60 (in Chinese).

Xu, J., R. Yin, Z. Li, and C. Liu. 2006. China's ecological rehabilitation:
Unprecedented efforts, dramatic impacts, and requisite policies. Ecological
Economics 57:595-607.

Yakubovich, V. 2005. Weak ties, information, and influence: How workers find jobs
in a local Russian labor market. American Sociological Review 70:408-421.

Yang, X. and F. Guo. 1999. Gender differences in determinants of temporary labor
migration in China: A multilevel analysis. lntemational Migration Review
33:929-953.

Yang, X. S. 2000. Determinants of migration intentions in Hubei province, China:
individual versus family migration. Environment and Planning A 32:769-787.

Zbinden, S. and D. R. Lee. 2005. Paying for environmental services: An analysis of
participation in Costa Rica's PSA program. World Development 33:255-272.

Zhang, X. and G. Li. 2003. Does guanxi matter to nonfarm employment? Journal of
Comparative Economics 31 :3 1 5-33 1 .

Zhang, Z. 2006. Review of the 10th F ive-Year Plan and perspective on the 11th Five-
Year Plan in relation to Natural Forest Conservation Program. Forestry
Economicsz49-52 (in Chinese).

Zuo, T. 2002a. Implementation of the NFPP.in J. Xu., E. Katsigris., and T. A. White,
editors. Implementing the Natural Forest Protection Program and the Sloping

161

Land Conversion Program: Lessons and Policy Recommendations. China
Forestry Publishing House, Beijing, China.

Zuo, T. 2002b. Implementation of the SLCP.in J. Xu., E. Katsigris., and T. A. White,
editors. Implementing the Natural Forest Protection Program and the Sloping
Land Conversion Program: Lessons and Policy Recommendations. China
Forestry Publishing House, Beijing, China.

162