ll .

     

W

J

 

1153315 LlRRARY

 

 

'3 Michigan State
1 0 ’._ ’7 UI liverSity

 

This is to certify that the
thesis entitled

WILDLIFE CONSERVATION IN ZAMBIA
IMPACT OF GAME MANAGEMENT AREAS ON
HOUSEHOLD WELFARE

presented by

Ana Fernandez

has been accepted towards fulfillment
of the requirements for the

MS. degree in Agricultural Economics

 

 

 

 

MSU is an Affirmative Action/Equal Opportunity Employer

'vlc-np.-.—.-._.-.— .— —

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 K:/Prolecc&Pres/ClRC/DateDue.indd

 

WILDLIFE CONSERVATION IN ZAMBIA
IMPACT OF GAME MANAGEMENT AREAS ON HOUSEHOLD WELFARE

By

Ana Fernandez

A THESIS

Submitted to
Michigan State University
In partial fulfillment of the requirements
For the degree of

MASTER OF SCIENCE
Agricultural Economics

2010

ABSTRACT

WILDLIFE CONSERVATION IN' ZAMBIA
IMPACT OF GAME MANAGEMENT AREAS ON HOUSEHOLD WELFARE

By

Ana Fernandez

This research investigates the impact of wildlife conservation policies on the welfare of
communities living in Zambian Game Management Areas (GMAs). The study first uses
simple OLS to measure the overall effect that living in a GMA has on household welfare.
Findings indicate that households living in prime GMAs enjoy 17% higher incomes than
those living in non-GMA areas and that these benefits are captured by the wealthier
segments of the community. Secondly, a two-step procedure (the Cragg double hurdle
model) is used to explore the avenues through which this GMA effect is generated by
quantifying the impact that living in a GMA has on self employment, wage employment,
farrn- and nonfarm incomes and crop damage. Households in prime (well stocked) GMAs
are found to be more likely to participate in off-farm wage and self-employment
compared to households living in non-GMAs. With respect to crop damage, results
clearly show that households living in prime GMAs are more likely to suffer crop loss

due to wildlife, an unresolved conflict that needs further exploration.

To my parents, for everything

AKNOWLEDGEMENTS

There are a significant number of institutions and people that have made this thesis
possible. Firstly, I would like to thank the Agriculture Food and Research Economics
Department (AFRE) for giving me the opportunity to participate in the masters program.
I would like to acknowledge the World Bank (WB) and the Zambia Wildlife Authority
(ZAWA) for sharing the Impact of Game Management Areas on Household Welfare
survey (IGMAW) for use in this thesis. I am also grateful to USAID for their financial

support and interest in this study.

I would like to give special thanks to my thesis guidance committee, Dr. David Tschirley,
Dr. Robert B. Richardson and Dr. Steve Haggblade for their continuous encouragement,
advice and patience during the conceptualizing and writing process. I have greatly
enjoyed working with them and highly benefited from their knowledge. I am also
indebted to Dr. Songqing J in and Bill Burke for their assistance with the econometric

section of the study.

From the Zambian side, I am extremely grateful to Dr. Gelson Tembo, from the
University of Zambia (UNZA), whose previous work on Game Management Areas
(GMAs) and local welfare has been a fundamental reference for this thesis. I am also
grateful to Betty Misimuko, from ZAWA for sharing her vast knowledge and experience
on Community Based Natural Resource Management and for facilitating arrangements to

meet with various informants in Chiawa and Lupande GMAs.

iv

Finally, I would like to thank my MSU colleagues and friends who have been extremely
helpful, especially Nicky Mason and Joleen Hadrich who are always willing to help new

students.

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ............................................................................................................ x
1 . INTRODUCTION ....................................................................................................... 1
1.1 Economic Growth and Poverty ................................................................................ 1
1.2 The Importance of the Tourism Sector .................................................................... 5
1.3 Tourism Sector and Poverty Reduction ................................................................. 10
1.4 Problem Statement ................................................................................................. 11
1.5 Objectives of the Study .......................................................................................... 16
1.6 Organization of the Study ...................................................................................... 17
2. GAME MANAGEMENT AREAS ........................................................................... 18
2.1 History and Decentralization of Resource Management ....................................... 18
2.2 Game Management Area Policy ............................................................................ 22
2.3 Wildlife Management and Hunting Policy Outside GMAs ................................... 28
2.4 Community Based Natural Resource Management ............................................... 30
2.4.1 Wildlife Management and Communities ........................................................ 31
2.4.2 Costs and Benefits of CBNRM ...................................................................... 32
3. SURVEY DATA ....................................................................................................... 37
3.1 Coverage and Sample Design .................................... , ............................................ 37
3.2 Coverage ................................................................................................................ 37
3.3 Sampling Process ................................................................................................... 38
3.4 Stratification of GMAs and non-GMAs ................................................................ 38
4. ANALYTICAL APPROACH AND HYPOTHESIS TESTING .............................. 40
4.1 Model 1: Impact of GMA Policies on Household Income .................................... 42
4.1 .1 Model Specification ........................................................................................ 47
4.1 .2 Model Estimation ........................................................................................... 50
4.1.3 Stage 1: Empirical Estimation and Results ..................................................... 53
4.2 Model 2: Avenues through which the GMA Effect is Generated .......................... 60
4.2.1 GMA Effect on Off-Farm Earnings ................................................................ 61
4.2.1.1 Model Specification .................................................................................... 65
4.2.1.2 Model Estimation ........................................................................................ 70
4.2.1.3 Empirical Results of GMA Effect on Wage Employment .......................... 71
4.2.1.4 Empirical Results of GMA Effect on Self Employment ............................ 77
4.2.2 GMA Effect on Crop Agriculture ................................................................... 80
4.2.3 Impact of GMA policies on crop damage ...................................................... 82
4.2.3.1 Model estimation ........................................................................................ 83
4.2.3.2 Empirical estimation and results ................................................................. 86

vi

5. SUMMARY OF RESULTS AND CONCLUSIONS ............................................... 90

5.1 Conclusions ............................................................................................................ 93
6. AREAS FOR FURTHER RESEARCH .................................................................... 95
APPENDICES .................................................................................................................. 98
APPENDIX A: ANALYSIS OF THE MIGRATION VARIABLE ................................. 99
APPENDIX B: OLS RESULTS BY WELFARE LEVEL AND PARK SYSTEM ....... 102
APPENDIX C: CRAGG RESULTS BY WELFARE LEVEL ....................................... 104

APPENDIX D: ENDOGENEITY TEST, CRAGG MODEL ON CROP DAMAGE... 111

BIBLIOGRAPHY ........................................................................................................... 1 l4

vii

LIST OF TABLES

Table 1. Zambian national parks and numbers of visitors (2003-2005) ............................. 9
Table 2. Trophy fees for killed or wounded animals* (USS) ........................................... 26
Table 3. Land use fees by tourist lodge category .............................................................. 27
Table 4. Sample SEAs by National Park and GMA classification ................................... 39
Table 5. Economic activities in GMA and non-GMA areas ............................................. 43
Table 6. Income share over all households by income quartiles ...................................... 44

Table 7. Comparison between variable means for the full sample and the GMA and non-

GMA subsamples. ............................................................................................................. 46
Table 8. OLS regression on the GMA effect on total household income ......................... 57
Table 9. Comparison of GMA effect on household income by welfare level .................. 58
Table 10. GMA effect by park system .............................................................................. 60
Table 11. Mean comparison between households earning wage income and households
not earning wage income .................................................................................................. 63
Table 12. Cragg results for wage employment earnings ................................................... 73
Table 13. GMA effect on wage income by welfare levels ............................................... 76
Table 14. Cragg estimation results: determinants of the probability and total level of self-
employment earnings ........................................................................................................ 78
Table 15. GMA effect on self-employment earnings by welfare levels ........................... 79
Table 16. Cragg estimation results: determinants of the probability and total level of crop
agriculture ......................................................................................................................... 80
Table 17. GMA effect on crop agriculture by welfare level ............................................. 82

Table 18. Comparison between variable means for the full sample and the GMA and non-
GMA subsarnples .............................................................................................................. 84

viii

Table 19. Cragg results: GMA effect on probability and total value of crop loss ............ 87

Table 20. GMA effect on crop damage by welfare level .................................................. 88
Table 21. Summary of Cragg results on sources of income by welfare level. ................. 89
Table A 1. Mean comparison of the migration variable ................................................. 100
Table A 2. Mean comparison of the migration variable by park system ........................ 100
Table B 1. Results of the OLS by welfare level ............................................................. 102
Table B 2. Results of the OLS by park system ............................................................... 103
Table C 1. Results of Cragg model on wage employment by welfare level (lower two
quintiles) ......................................................................................................................... 1 04
Table C 2. Results of Cragg model on wage employment by welfare level (upper two
quintiles) ......................................................................................................................... 1 05
Table C 3. Results of Cragg model on self employment by welfare level (lower two
quintiles) ......................................................................................................................... 1 06
Table C 4. Results of Cragg model on self employment by welfare level (upper two
quintiles) ......................................................................................................................... 107
Table C 5. Results of Cragg model on crop income by welfare level (lower two

quintiles) ......................................................................................................................... 108
Table C 6. Results of Cragg model on crop income by welfare level (upper two
quintiles). ........................................................................................................................ 109
Table C 7. Results of Cragg model on crop damage by welfare level (lower two

quintiles) ......................................................................................................................... 1 10
Table C 8. Results of Cragg model on crop damage by welfare level (upper two

quintiles) ......................................................................................................................... 1 10
Table D 1. Results of OLS on number of scouts ............................................................ 111
Table D 2. Results of the ivtobit on crop damage ........................................................... 112

LIST OF FIGURES

Figure 1. Percentage GDP growth in Zambia (1995-2007) ................................................ 2
Figure 2. GDP distribution by economic sector (in billion Kwacha) ................................. 4
Figure 3. Average annual worldwide tourism growth by income classification ................ 6
Figure 4. International tourism receipts and arrivals in Zambia, 1990-2004 ...................... 7
Figure 5. Zambian National Parks and Game Management Areas .................................... 8
Figure 6. Structure of community representation in a GMA ............................................ 25

1. INTRODUCTION

Most rural households in Zambia are poor. The majority of these households rely on
subsistence agriculture as their main livelihood, which is typically insufficient to ensure
food security. Although most households seek income diversification opportunities as
part of their strategy for risk management and income generation, these are often out of
reach for the poorer households, due to capital or labor constraints as well as a general
lack of off-farm employment opportunities. Providing access to off-farm employment in
rural areas is one of the Zambian Govemment’s key objectives for the 2006-2010 period.
One of the identified sectors for pro-poor growth is the tourism industry, a growing sector
in the country though still underdeveloped. It is the vision of the Government that the
development of the tourism industry can boost rural economies, especially in the areas
surrounding national parks where most of the tourist investments are currently

concentrated.

1.1 Economic Growth and Poverty

During the first decade of independence (1964-1974) Zambia was one of the wealthiest
countries in sub-Saharan Africa. Prosperity was mainly a result of the successful mining
industry which benefited from high international prices. The oil crisis in 1973-74 coupled
with the fall in world copper prices began an extended period of decline in real income.
High oil and energy prices triggered global inflation which made imported capital and
manufactured goods very expensive. The balance of payments deteriorated and the
country became heavily dependent on foreign borrowing to minimize the decline in living

standards (Government of the Republic of Zambia, 2002).

Since the late 19905, however, the economy has experienced a steady recovery (Figure
1). After a steep decline of the economy from 1996 to 1998, partly due to the East Asian
financial crisis (which lowered the demand for copper) and donors’ withdrawal of
support to the balance of payments, GDP began a positive growth trend. Favorable global
economic conditions, the impact of the economic reforms that started in the early 19903,
rapid expansion of the mining industry boosted by an increase in international prices, and
growth of the construction sector through private investments have been some of the key

drivers (Government of the Republic of Zambia, 2006).

Figure 1. Percentage GDP growth in Zambia (1995-2007)

 

 

A

Iv r V T l 1 I r T

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006*2007*

 

2]
/.,\/
‘l

 

 

 

 

 

Source: Elaborated with data from the Central Statistics Office of Zambia (2008). Data for years 2006* and
2007* are estimates.

However, despite the rapid acceleration of the economy Zambia still suffers from
persistent income poverty. The economic benefits have not translated into a significant

reduction of rural poverty, which still ranks among the highest in sub-Saharan Africa.

According to the 2006 Living Conditions Monitoring Survey (LCMS), 80% of the
population in rural Zambia is poor and 67% is extremely poor. Farmers remain highly
vulnerable to erratic weather patterns and generally lack access to markets, credit,
fertilizer and other inputs. These conditions translate into very low productivity, a
phenomenon that has resulted in chronic food insecurity and high prevalence of
malnutrition among children and adults (Government of the Republic of Zambia, 2002).
Mining, wholesale and retail trade and construction, the pillars of economic growth
(Figure 2), are urban-based and capital-intensive, and generate little additional
employment due to weak linkages with the rest of the economy (Government of the
Republic of Zambia, 2006); relatively little investment has taken place in sectors that

generate rural employment and income diversification.

Figure 2. GDP distribution by economic sector (in billion Kwacha)

 

Personalservices I
Community,social and private services
Real estate and business services

Financial and insurance services
I GDP (billion
ZMK)

 

Tra ns portand communication
Restaurants, bars and hotels

Wholesale and retailtrade "

 

Construction

Electricity, gas and water
Manufacturing 1

Mining and quarrying

 

Agriculture, forestryand fishing -

0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Source: Elaborated with data from the Central Statistics Office (CSO) of Zambia using estimates for 2006.

 

 

 

Still, notwithstanding the high incidence of rural poverty, total rural poverty rates have
shown a slight decline over the last decade, diminishing from 92% to 80% from 1993 to
2006. In rural areas, the decline in poverty levels can be partly explained by the increased
supply of food crops such as cassava, sweet potatoes and groundnuts, as well as export
commodities like cotton and tobacco, which have helped boost rural incomes (F ynn and
Haggblade, 2006; Jayne et a1, 2007; Tschirley and Kabwe, 2007). In urban areas, Jayne et
a1 (2007) explain the increasing poverty as a consequence of the declining mining
industry after the fall in copper prices during 1998-2001, and the elimination of consumer

food subsidies in the early 19905.

Acknowledging the need to focus on economic activities that generate broad-based
wealth and job creation, the Government of Zambia plans to promote sectors thought to

be pro-poor and labor intensive. Since poverty is more acute in rural Zambia, agriculture

is one of the main areas of attention for national deveIOpment. The Government seeks to
transform the country’s economic structure by promoting large-scale commercial farms,
more modern technologies and upstream linkages (agro processing) that can boost rural
incomes and translate into higher demand for locally manufactured goods. Apart from
agriculture, the Government envisions the expansion of a diversified export base and a

stronger tourism sector as engines of pro-poor growth.

1.2 The Importance of the Tourism Sector

Tourism is one of the fastest growing economic sectors in the world. Global tourism
revenues grew 11.2% per year between 1950 and 2005 (WTO 2008). International
tourism arrivals grew at an average annual rate of 6.5%, increasing from 25 million to
806 million visitors. The sector has become one of the major businesses in international
commerce, and represents one of the main income sources for many developing
countries. The increasing importance of the sector in developing countries is reflected by
the growth of tourism in low-mid and least developed countries (LDCs) compared to

growth worldwide and in higher income countries (Figure 3).

Figure 3. Average annual worldwide tourism growth by income classification
(1990-2005)

8 .2 8.6
I % Growth
5.8
4.2
2.9

World 50 LDCs Low-Mid Upper-Mid High
Income Income Income

 

 

OHNthIO‘NmtO

 

 

 

Source: World Tourism Organization

In Zambia, the tourism sector has been steadily growing over the past years in terms of
arrivals and tourism receipts (Figure 4), ranking 15‘h out of 73 countries in the list of

emerging tourism destinations during 1995-2004.

Figure 4. International tourism receipts and arrivals in Zambia, 1990-2004

(USS million)

 

600

500

400

300

200

100

 

 

A

/\/

 

 

 

 

 

 

 

2004

+Tourist arrivals (1000)

+Tourism receipts (USS
million)

 

Source: Elaborated with data from the World Tourism Organization

The industry relies on the country’s impressive natural resources; there are 19 national

parks and 35 Game Management Areas (GMA) in the country, representing 30% of the

total territory. Figure 5 shows the current network of protected areas in the country which

are classified as national parks and GMAs. National parks are intended for the protection

and enhancement of wildlife, ecosystems and biodiversity. No human settlements are

permitted and only photographic safaris, also known as non-consumptive wildlife use, are
allowed. GMAs act as buffer zones between the national parks and farming areas. They

are intended to promote sustainable harvest of wildlife through hunting as an alternative

to other economic activities not compatible with wildlife protection. GMAs also offer

wildlife viewing but allow human settlements and licensed hunting (consumptive wildlife

use).

 

Figure 5. Zambian National Parks and Game Management Areas

 

tr

 

ZAMBIA

- National Park

if; Game Management Area

 

 

 

 

Approximately 88% of international visitors that arrive in Zambia are nature tourists who
primarily seek scenery, wildlife viewing and adventure activities such as rafting and
canoeing. The country offers excellent wildlife viewing in an uncrowded landscape,
cultural and adventure opportunities, and the unique Victoria Falls, the largest waterfall
in the world and top tourist attraction in Zambia, situated at the border with Zimbabwe in
Southern Province. The next most popular attraction is South Luangwa national park,
situated in Eastern Province and internationally renowned for its wildlife population and
scenery. Other popular national parks are Mosi-Oa-Tunya (conveniently close to Victoria

Falls), Kafue and Lower Zambezi (Table 1).

Table l. Zambian national parks and numbers of visitors (2003-2005)

 

 

National Parks 2003 2004 2005
Lower Zambezi 4413 6059 6040
Mos i-oa-Tunya 23497 1 7762 19972
South Luangwa 19728 23929 25814
Kafue 3812 3789 6202
Lochnivar 390 415 784
Other Parks * 700 2024 2588
Total 52540 53978 61400

 

" (North Luangwa, Liuwa, Kasanka, Luambe, West Lunga, Sioma Ngwezi, Sumbu, Chete Island)
Source: Hamilton et a1. (2007) from ZAWA Commercial section, 2007

The full potential of the tourism industry in Zambia is still emerging (Hamilton et al.,
2007). The sector is strongly oriented toward nature tourism and is characterized by
small-scale investments; at least 250 nature-tourism accommodation providers are
licensed countrywide. The major tourism hubs have developed around a few key urban
and national park locations, especially around Livingstone where tourists can visit the
Victoria Falls and the nearby Mosi-Oa-Tunya Park (Hamilton et al., 2007). Overall,
tourists spend an average of 6.3 days in the country, which is low compared to its
neighbors Bostwana (8.6 days) and Namibia (12.4 days). One of the major constraints for
tourists willing to visit different parts of the country is the lack of adequate infrastructure
(roads, airports) that can easily facilitate travel between Livingstone in Southern Province

and the other parks or sites of interest in the country.

Other factors contributing to the limited development of the sector are the extreme
seasonality of Zambia’s holiday tourism, which in some areas is limited to five or six
months in the year due to the poor or non-existent all weather road network that restricts

transport during the rainy season (Adrian Coley, pers. comm.) Growth in the sector is

further restricted by limited capital, the lack of properly trained personnel, poorly
developed marketing at the national level and limited policy, legislation and planning for

the sector (Hamilton et al., 2007).

1.3 Tourism Sector and Poverty Reduction

Since the 19803, the concept of ecotourism and nature based tourism has emerged in
Latin America, Africa and Asia as a strategy to conserve biodiversity and reduce poverty
(Honey, 1999). The philosophy is to find models of tourism that minimize impact on
natural resources while generating income for local communities. A more recent concept
is the definition of pro-poor tourism (PPT). This term was first coined by the British
Department for International Development (DF ID) in 1999 and has since received
support from other institutions like the World Tourism Organization. According to the
definition provided by the PPT Partnership, “PPT enhances the linkages between tourism
businesses and poor people, so that tourism’s contribution to poverty reduction is
increased and poor people are able to participate more effectively in product

development” (Ashley, 2004: 1.).

The Government of Zambia has included tourism as one of the key sectors to combat
poverty, giving it a prominent place in its strategy for economic growth and poverty
reduction. The Government envisions the country as a “major tourism destination of
choice with unique features, which contributes to sustainable economic growth and
poverty reduction by 2030” (Government of the Republic of Zambia, 2006:125). The
Government plans to strengthen the sector by tackling some of the shortcomings

mentioned earlier such as investing in infrastructure and tourism facilities, diversifying

10

tourism products beyond its current heavy focus on wildlife, enhancing human resources
through training, restocking of species with declining populations, and encouraging the

participation of communities as business partners.

Community participation in tourism development is one of the major avenues for
promoting pro-poor tourism. During the last two decades, the Government of Zambia has
been implementing co-management agreements for wildlife use (consumptive and non-
consumptive) with communities in GMAs. The Zambia Wildlife Authorityl (ZAWA), the
quasi governmental agency in charge of wildlife protection and management, promotes
the organization of communities in Community Resource Boards (CRB) to become
partners in wildlife protection and in the sharing of benefits from trophy hunting and
photographic safaris. This approach, known as Community Based Natural Resource
Management (CBNRM), has the dual goal of improving the welfare of local communities

and creating incentives for the protection and conservation of natural resources.

1.4 Problem Statement

The co-management of wildlife resources presents opportunities and threats for
communities living in GMAs. Through the CBNRM program, communities receive a
share of the revenues generated from trophy hunting and concession fees paid by hunting
outfitters. The development of the tourism industry also offers opportunities for wage

employment and creation of small businesses, in addition to the benefits from increased

 

I In 1999, the Government of Zambia transformed the former Department of National Parks and Wildlife
Service (N PWS) into an autonomous body, the Zambia Wildlife Authority (ZAWA). ZAWA is governed
by the Zambia Wildlife Act of 1998 and has its own Board of Directors. The Board is responsible for
managing ZAWA under the policy guidance of the Ministry of Tourism, Environment and Natural
Resources.

11

access to infrastructure and services created to cater for tourists. However, the realization
of these opportunities depends on various factors, such as the ability of the tourism
industry to create employment and revenues through hunting licenses, the appropriate
planning of land use and human settlements, the transparency with which the main actors
(ZAWA, area chiefs, community representatives) manage the program, the degree of
devolution of decision making to communities, and the community’s commitment to
protect wildlife. The effectiveness of the program is also threatened by potential
unintended negative effects, namely greater crop destruction as wildlife populations

increase, and the pressure that in-migration puts on land other natural resources.

The livelihoods of communities located near national parks are similar to other rural
Zambian communities, though they face additional hardships. F armers are routinely
affected by crop destruction, especially by elephants which are the most destructive
species. The Tsetse fly, carried by large mammals, affects the health of livestock and
traction animals, limiting their productivity. Bandyopadhay and Tembo (2009) found that
the average annual per capita consumption in areas surrounding national parks is between
ZMK 839,000 and 850,000, approximately 30% lower than the average for a rural

community, estimated at ZMK 1.2 million (Simasiku et al., 2008)2.

In some cases, poaching is a coping strategy for poor households who turn to it for a
source of cash and meat that would be inaccessible otherwise. Illegal use of wildlife

persists where there are limited income alternatives and communities perceive no

 

2 This calculation is based on the results in Bandyopadhay and Tembo (2009) who studied the welfare of
communities living in areas adjacent to National Parks. The results where compared to the Living
Conditions Monitoring Survey results (2006) for rural annual per capita consumption.

12

incentives for wildlife protection. To reduce poaching that had started to threaten wildlife
species, international development agencies including the United Nations (UN), the
United States Agency for International Development (U SAID) and the Department for
International Development (DF ID), in the 19808 began advocating for increased
community participation in wildlife management decisions and increased community
sharing in benefits from wildlife use. In response, the Zambian Government introduced

CBNRM programs in GMAs.

CBNRM programs aim at increasing household welfare and conserving natural resources.
The idea is to provide communities with an incentive in the form of revenues from trophy
hunting licenses that outweigh the benefits they would get from illegal hunting. The
program makes two assumptions: a) that community participation in wildlife
management is a more effective way of conserving wildlife than a centralized regulatory
approach, and b) that sustainable wildlife utilization is a profitable land use option for

local communities as opposed to farming or grazing for example.

In CBNRM programs, communities assume certain rights and responsibilities. They
agree to protect wildlife resources by preventing residents from engaging in all forms of
illegal use of wildlife and to engage in only the land use activities agreed upon with
ZAWA. This ensures that commercial activities dependent on wildlife are not threatened.
In exchange, communities benefit from the wildlife resources found on community land

and from participation in the decisions regarding resource use3.

 

3 The first decentralization policy for co-management of natural resources was named The Administrative
Management Design for Game Management Areas (ADMADE). ADMADE drafted a constitution to set

13

There are clear potential benefits for communities that participate in CBNRM programs.
They are empowered to negotiate agreements with safari and tour operators to engage in
commercial activities and share in the revenues generated from concession fees and
hunting licenses. A percentage of the funds that communities receive are used to hire
local scouts to patrol the areas, which provides income for some households. Another
portion of the funds is used in community projects such as schools and clinics.
Communities can also benefit from the creation of employment by the tourism industry
and (though barriers are likely higher here) business opportunities to provide services to
the sector. Communities near popular tourist attractions benefit from an increase in the
amount and quality of infrastructure and services provided such as electricity, roads, and
communications. These opportunities for income diversification in the non-farm sector
can have a strong positive effect on household welfare (Barrett et al., 2001; De Janvry
and Sadoulet, 2001; Reardon, 1997). However, there could be unintended negative
effects emerging from the CBNRM programs. New job opportunities and an
improvement in social services derived from the CBNRM programs could lead to high
in-migration, putting pressure on resources such as firewood and land that could

eventually dilute the intended positive effects (Ahmed et al., 2001).

Communities may also incur the opportunity cost of not engaging in alternative land uses
such as farming or grazing. Despite the usually low land productivity in GMAs, the
conditions under which access to open areas for grazing or gathering of forest products

generates more benefits for community members than a tourist lodge or a safari hunting

 

the general guidelines for the establishment of agreements between the Zambia Wildlife Authority
(ZAWA) and the community. The rights and responsibilities cited are extracted from the draft constitution.

14

operator needs to be explored empirically. The balance between these effects will
ultimately depend on the potential of tourism to generate employment and hunting
revenues to make an impact on local welfare. Land is administered through customary
law by the area chiefs, who decide to whom they want to lease or allocate land and for
what purposes. An approach that takes into consideration the interests of the communities

will be needed to ensure that they actually attain benefits from the CBNRM programs.

As mentioned earlier, the human-animal conflict is another threat to the welfare and
livelihoods of communities, with elephants being a special problem. Villages in close
proximity to high elephant populations are regularly threatened by crop destruction. The
implementation of CBNRM programs could have an ambiguous effect on the problem.
On one hand, the increased efforts to prevent poaching are likely to have a positive effect
on elephant populations, which can add to the number of conflicts. On the other hand,
larger populations of elephants can attract more tourists, and generate more revenues for
the communities. Elephant hunting was resumed in Zambia in 2004 after the country
gained approval from the Convention on International Trade in Endangered Species of
Wild Fauna and Flora (CITES) to export 20 trophies per year. Hunting is restricted to
two regions, the Luangwa Valley and the Lower Zambezi Valley, and only a few GMAs
within those areas have elephant quotas. There is an ongoing controversy over the
legalization of elephant hunting. Some argue that carefully controlled hunting can help
communities with crop destruction and provide them with revenues from trophy fees.
Others argue that CITES’s ban on ivory trade helped recover elephant population and that

a change in policy would drive numbers down again (Ahmed et al., 2001).

15

1.5 Objectives of the Study

For the consideration of Government’s policies to promote tourism as a pro-poor tool to
combat poverty, it is crucial to understand whether CBNRM programs have the potential
to improve local welfare. Given the alternative uses of land and the negative effects that
CBNRM programs could have in terms of in-migration and resulting pressure on land
and forest resources, as well as on crop destruction, there is a need to understand whether

the program is an effective strategy to enhance community welfare.

To assess the effectiveness of GMA policies, this study presents three analyses. It first
tests econometrically whether households living in GMAs have higher incomes than
those living in non-GMAs. It then endeavours to understand the channels through which
GMA policies affect household welfare (positively or negatively), using two-step
econometric techniques to identify the impact of living in a GMA on i) wage income, ii)
self-employment income, and iii) the value of crop loss from animal damage.

This study complements Bandyopadhay and Tembo (2009) based on the data collected
through the Impact of Game Management Areas on Household Welfare (IGMAW)
survey. Bandyopadhyay and Tembo analyze the impact of GMA policies on household
welfare, as measured by household expenditure. They examine the factors that influence
household participation in natural resource management4 and whether participating
households benefit more than non-participants. They also examine the distribution of the
impact among the poor and non-poor segments of the community. Their main findings

indicate a strong association between living in a GMA and higher levels of consumption

 

4 The participation of a household in resource management refers to the participation in Community
Resource Boards or Village Action Groups, the two organizations that represent community interests in
wildlife management decisions.

16

expenditure, and that the benefits are more likely to be attained by the non-poor segment

of the communities and by those participating in resource management.

This study contributes to the literature by: i) providing a cross-check of the findings on
GMA impact on household welfare using a different indicator, income, for household
welfare and ii) complementing the findings by examining the impact of GMAs on
different sources of income (wage and self-employment earnings, income from crop

harvest, and income losses from crop damage).

1.6 Organization of the Study

The study is organized into six chapters. Chapter two gives an overview of the creation of
protected areas, the evolution of management policies from the colonial times until
present, the origin and evolution of CBNRM programs in Zambia and the effectiveness of
the programs illustrated with various examples from the existing literature. Chapter three
provides a description of the survey data, sampling strategy, and the areas of study used
for the analysis. Chapter four analyzes the impact of GMA policies and regulations on
household income and selected components thereof. Chapter five discusses the results
and implications for wildlife management policy. Finally, Chapter six outlines potential

topics for further research.

17

2. GAME MANAGEMENT AREAS

This chapter describes the history, characteristics and policies of GMAs which are the
focus of this study. It gives an overview of the evolution of wildlife management policies
in Zambia and the decentralization process towards a co-management approach that has
taken place in recent years. It reviews the performance of CBNRM projects and discusses

the achievements and shortcomings of the wildlife management system.

2.1 History and Decentralization of Resource Management

The traditional land tenure system in Zambia is regulated by customary law. Under the
territory where customary law is applicable (94% of the country), land belongs to the
communities for their own use and exclusive individual rights are not recognized.
Traditional authorities are responsible to ensure that all capable members of the
community (criteria for assessing capability are left to the discretion of the chief) are
allocated land (Mudenda, 2006). Individual rights are not recognized in terms of
registered proprietorship but in terms of occupancy rights which can be inherited (van

Loenen, 1999).

Prior to colonization, the responsibility for wildlife protection and use was vested in the
communities. The village chief controlled the allocation of land and access to and use of
forest and wildlife resources. That changed in the 19405 when the colonial Government
passed legislation that established Control Hunting Areas (CHA). Initially, the objective
of CHAS was to secure subsistence (unlicensed) hunting for residents and to keep the

Tsetse fly from spreading to livestock. In 1954, however, all hunting in the better

18

stocked CHAS was restricted to holders of licenses, allowing European hunters access to
areas formerly reserved for local populations (Ooi, 1982). By virtue of this new policy,
ownership of and access to wildlife resources was taken away from local chiefs and the
State took over their management and exploitation. Land remained administered by
customary law, but wildlife and natural resource administration was transferred to the

Government.

Detailed policy for wildlife protected areas first appeared in a ministerial annual report in
1958. This report was followed by a series of guidelines regarding the use and
management of game reserves as well as the formulation of protected area legislation.
The national parks and GMAs were formally established under National Park and
Wildlife Act No. 57 of 1968, which authorized the President to declare any area a
national park. A few years later, the Game Management Area Declaration Order of 1971,
the National Parks Declaration Order of 1972, and Statutory Instrument No. 44 of 1972
established the current network of national parks and Game GMAs (Chundarna et al.,

2004)

Notably, communities were left out of this legislative process. As in most African
countries, post-independence conservation legislation emulated the style of the colonial
government, which had little if any regard for the traditional land uses of local
populations. Communities were seen as a threat to wildlife instead of a pivotal element
for its sustainability (IIED, 1994). The new policies deprived communities of traditional

wildlife food sources, the opportunity to trade wildlife products, and the use of wildlife

l9

for ceremonial and traditional customs (Wright, 1995). All revenues from hunting were
sent back to the Ministry of Finance or kept by the hunting outfitters that were allowed to
operate in the areas (Chundama et al., 2004). As a result, conflicts began to emerge
between villagers and authorities. As it became clear that alienating communities was
counterproductive to conservation efforts, initiatives emerged to include communities in
the benefits generated from wildlife use. In the 19705, the UNESCO Man and the
Biosphere Program promoted the creation of buffer zones between strictly preserved
areas and human settlements. Buffer zones were conceived as lands adjacent to parks
and reserves where human activities were limited to those compatible with maintaining
the ecological security of the protected area, while providing benefits from wildlife
management (e. g., wages, income from hunting licenses or meat) and from development

programs (e. g., schools and clinics) to local communities (Neumann, 1997).

The Government of the Republic of Zambia (GRZ) followed this philosophy and made a
first attempt to involve communities in wildlife management. The Game Management
Area Declaration Order of 1971 introduced the objective of conserving wildlife and
integrating its management into the rural economy (Lungu, 1990). The idea was to
involve communities in wildlife management decisions and transfer a share of the
hunting revenues to them in an effort to stop or reduce other land-degradation activities
that threatened wildlife habitat. However, the legislation embedded in the Game
Management Area Declaration Order of 1971 did not include specific provisions to
enable communities to meaningfully participate in wildlife management (Lungu, 1990).

Conflicts between communities, government and safari hunters escalated. Locals saw no

20

benefit from conservation, but commercial poaching was becoming a profitable business
as western countries increased their demand for wildlife products (Gibson and Marks,
1995). By the 19805, the Zambian wildlife population was compromised and some
species, like elephant and rhino, were close to extinction. The black rhino population in
the Luangwa Valley shrank from about 8,000 head in the early 19705 to less than 100 by
the mid 19805, and disappeared soon thereafter. Over the same period, elephant

populations fell from 90,000 to 15,000 (Child and Dalal Clayton, 2004).

In 1983, subsidiary legislation was approved to try a new approach aimed at
decentralizing power to the individual GMAs and allowing communities to participate in
the revenues generated from trophy hunting (Lungu, 1990). The new approach was called
Administrative Management Design for Game Management Areas (ADMADE) and fell
under the jurisdiction of the Department for National Parks and Wildlife (DNPW,
predecessor of the Zambia Wildlife Authority). Under the new framework, a pilot
program, the Lupande Development Project (LDP) was tested in Lupande GMA, adjacent
to South Luangwa national park. The program was intended to involve the local
community in wildlife management and its benefits. With the revenues from hunting
licenses and park fees, the project established a revolving fund that was managed by the
GMA. The funds were partly utilized to train and hire village scouts (108 as of 1989),
which were recruited among the young members of the community. It became a very
attractive financial opportunity for otherwise unemployed people and an alternative to

poaching. Other key features of the new project were that communities received 35% of

21

the hunting funds to finance community projects and game culling was implemented to

provide game meat for the community (Lungu, 1990).

At the same time that ADMADE was initiated, another initiative emerged with similar
goals to those of the LDP. With funding from the Norwegian Agency for Development
Cooperation (NORAD) and personal involvement from Zambian President Kenneth
Kaunda, the Ministry of Finance launched the Luangwa Integrated Resource
Development Project (LIRDP). The two programs shared the goal of involving
communities in wildlife management, though their approaches differed. The LIRDP was
the first to introduce grass-roots level institutions (Village Action Groups, or VAGs) to
ensure local participation and administration of funding. In time, the policies of both
programs were harmonized and transferred to Zambia Wildlife Authority (ZAWA)

jurisdiction5 .

Learning from the Lupande experience, CBNRM programs were introduced in the rest of
the GMAs throughout the country. To date, a total of 63 CRBs have been formed

countrywide (ZAWA, 2007).

2.2 Game Management Area Policy
In 1998, the Zambia Wildlife Act introduced ZAWA, the institution that would be in
charge of managing wildlife in protected areas of Zambia. ZAWA is responsible for the

establishment, control and management of GMAs; provides for the sustainable use of

 

5 For a detailed background on the creation and evolution of the LIRDP program see “Lessons from
Luangwa. The Story of the Luangwa Integrated Resource Development Project, Zambia” (Child and Dalal
Clayton,2003).

22

wildlife and the effective management of the wildlife habitat; balances benefits between
local communities and wildlife; and involves local communities in the management of
GMAs (Zambia Wildlife Act, 1998). ZAWA provides wildlife protection, licensed
hunting, regulations for entry in wildlife protected areas, management planning
requirements for national parks and game management areas and the enforcement of
wildlife related activities in compliance with international agreements (Chundama et al.,

2004)

Legal hunting is the principal non-farm economic activity in GMAs (Simasiku et al.,
2008). GMAs with significant wildlife populations are divided into hunting blocks.
Exclusive rights to run trophy hunting operations are leased out to hunting outfitters by

the GRZ’s Tender Board Authority for a period of three years (Lewis and Alpert, 1997).

ZAWA is responsible for establishing the annual hunting quotas, according to animal
stocks. The process is participatory: communities, scouts and hunting outfitters contribute
to the estimation of animal populations, although there has been some criticism of the
accuracy of the numbers, which are said to have been inflated in order to generate more
funds (Adrian Coley, pers. comm..; Simasiku et al., 2008). The issuing of licenses and the
establishment of license fees are also regulated by ZAWA. Hunting game or any
protected species without the appropriate license is an offense and penalized by law with
imprisonment or fines, depending on the species and whether the hunting is done inside a
national park, GMA or open area. For example, the penalty for possessing, buying or

selling any protected animal or trophy of a protected animal is imprisonment for a period

23

of five to 15 years. Unlawful hunting within a national park is penalized with one to 15
years of imprisomnent. Outside a national park, the penalty is reduced to a maximum
imprisonment of 6 months or a fine not exceeding 20,000 penalty units (Zambia Wildlife

Act, 1998)

The responsibility for wildlife management in GMAs is shared between ZAWA and local
communities which are organized in CRBs. The Zambia Wildlife Act stipulates that “a
local community along geographic boundaries contiguous to a chiefdom in a GMA or an
open area or a particular chiefdom with common interest in the wildlife and natural
resources in that area, may apply to the Authority for registration as a community
resources board” (Zambia Wildlife Act, 1998: 17). The CRB represents the highest
management authority at a community level regarding aspects of wildlife management.
Each CRB includes a maximum of 12 members: no more than 10 and not less than seven
representatives democratically elected by the local community, one representative of the
local authority in the area and one representative of a chief in whose area a board is
established to represent that chief. GMAs that are divided into more than one hunting

block will have a CRB representing each hunting block.

Several Village Action Groups (VAG) may be organized under each CRB (Figure 6).
VAGs are constituted by groups of households interested in managing natural resources.
VAGs form VAG committees whose representatives are democratically elected by the

community and are the key structure to represent the interests of the community on the

24

CRBs. One or more VAGs can have representation on a CRB, depending on the number

of villages that share boundaries with the hunting block.

Figure 6. Structure of community representation in a GMA

‘ Hunting Block (1) Hunting Block (2)..

 

      
   

 

      

VAG (C ) . In“,
' ’2...-

Source: Author’s elaboration

 

 

 

 

The functions of a CRB are to promote and develop an integrated approach to the
management of human and natural resources in a GMA or an open area falling within its
jurisdiction. CRB responsibilities are, theoretically, quite broad. They have the power to
negotiate co-management agreements with hunting outfitters and photographic tour
operators, they co-manage the wildlife under their jurisdiction, with quotas specified by
ZAWA, and they appoint village scouts to control poaching. However, CRBs frequently

have difficulty fulfilling all these roles due to inadequate funding and local capacity.

ZAWA collects 100% of safari animal trophy fees and concession fees generated in
GMAs. It distributes 45% of the total revenues obtained from animal fees to CRBs and

5% to chiefs as CRB patrons, keeping 40% for its own administrative expenses and

25

sending 10% to the central Government. Out of the revenues obtained from concession

fees, 15% goes to CRBs, 5% to the chief and 80% to ZAWA.

Table 2 provides an example of current trophy fees for the year 2009 for hunting

concessions in Mulobezi, Mumbwa, Luangwa, Bangweulu and Kafue Flats.

Table 2. Trophy fees for killed or wounded animals* (USS)

 

 

Species U S$ S pec ies US$
Baboon 125 Leopard 6500
Buffalo 2600 Lion 9500
Chobe Bushbuck 850 Oribi 550
Bushpig 550 Puku 850
Black Lechwe 2950 Southern Reedbuck 750
Southern Bush Duiker 550 Southern Roan 5700
Livingstone's Eland 3650 Sable 4850
Sharpe's Grysbok 425 Zambezi Sitatunga 4500
Lichtenstein's Hartebeest l 150 Tsesseby 2750
Hippopotamus 2500 Warthog 550
Hyena 550 Common Waterbuck 1250
Southern lrnpala 300 Blue Wildebeest 1875
Kafue Lechwe 2950 Zebra 1150
Klipspringer l 150 Crocodile 2600
Southern Greater Kudu 2450

 

Source: Swanepoel & Scandrol (2009).
*Trophy fees correspond to 2009 fees for concessions in Mulobezi, Mumbwa,
Luangwa, Bangweulu and Kafue Flats.

The CRBs are mandated to distribute 45% of the revenues received from ZAWA to
resource protection (hiring of scouts and other logistic operations are the main expenses),
35% to the implementation of community projects and 20% to administration costs
(ZAWA, 2008). All projects funded through the CRBs have to be designed to benefit all
members of the community. Examples of community projects include construction of

community schools, health clinics, wells, boreholes, police posts and transport services.

26

In addition to sharing hunting revenues, some communities may sign ad hoc contracts
involving other arrangements. For example, communities in Chiawa and Bangweulu
GMAs have signed agreements with ZAWA and tour operators that operate photographic
safari lodges (game viewing). These agreements provide the community with an
alternative source of income. They receive a 50% share of the annual land use fees paid
by the lodges (Table 3). As in the case of hunting, communities make a commitment to
preserve wildlife and use part of the revenues for poaching patrol (Sydney Tembo, pers.
comm.) There are other conditions for the lodges such as having to hire at least 80% of
the staff locally. In addition, tourist lodges make individual pledges to contribute to the

building of schools, clinics, and other development projects.

Table 3. Land use fees by tourist lodge category

 

 

 

Annual Fees (US$)
Lodge category Fixed (per category) Variable (per # beds)
A 2,000 216
B 1,500 144
C 1,000 72
D (camp sites) n/a n/a

 

Source: Prepared with information from ZAWA’s Tourism and Commercial
Office. (Sidney Tembo, pers. comm.)

A report commissioned by the Ministry of Tourism, Environment and Natural Resources
(Chundama et al., 2004) to examine the institutions and policies governing protected
areas revealed legislative weaknesses in the Zambia Wildlife Act. It highlights the
following weaknesses and areas for improvement: “a) user rights definition, b)
appropriate guidelines for communities; c) review of policy and Act, especially in

relation to compensation for crop damage and loss of human life; d) sharing of costs of

27

management of protected areas is not clearly stated and e) need for transparency and
inclusion in the policy of revenue sharing formula” (Chundama et al., 2004: 110). It also
highlights the lack of definitions for other protected areas in addition to national parks or
GMAs, such as Wildlife and Bird Sanctuaries (Chundama et al., 2004). In addition,
ZAWA’s jurisdiction is limited to the regulation of wildlife use and its protection but not
over the habitat. GMAs are created on trust lands, which are communally owned. They
are administered by traditional authorities (chiefs and headmen) who decide on their
allocation and use. This has created conflicts as chiefs may encourage the use of land for

activities that are not compatible with wildlife needs (Lewis and Alpert, 1997).

2.3 Wildlife Management and Hunting Policy Outside GMAs

ZAWA defines open areas as areas other than national parks and GMAs where wildlife is
found. Although they are not officially designated as protected areas by the legislation,
ZAWA has jurisdiction over the protection and management of wildlife in open areas.
Hunting is allowed provided that the hunter has the appropriate license and permission
from the land owner. Permission to hunt in open areas is generally sought at ZAWA
Headquarters, while in GMAs, the permits are released by the area wardens
(decentralized at GMA level). Animal quotas in open areas are usually lower as
populations are normally sparse but the animal fees are the same (Betty Msimuko, pers.

comm).

For members of the community, hunting in open areas is as restricted as in GMAs.

Hunting of any game or protected animal is restricted to hunters who have paid the

28

corresponding fee, which is unaffordable for the majority of locals. In addition, whereas
in GMAs the license fees are paid to the area warden and shared between ZAWA and the
communities as per the current co-management agreements, in open areas the fees are
paid directly at ZAWA Headquarters, and the firnds remain with the administrative
authority. Although ZAWA supports the establishment of CRBs in open areas,
knowledgeable observers suggest that there is little interest from hunting outfitters or
game viewing lodges to establish business in those areas, since wildlife populations are
often too low to sustain tourism, limiting the capacity of communities to tap into the
revenues from legal hunting that could be generated. Even if there were a hunting
operator, the revenue sharing adopted in GMAs would not be automatically applied in
open areas. Communities would have to negotiate the terms of revenue sharing on a case-
by-case basis with hunting outfitters, photographic safari operators or private game

ranching businesses (Betty Msimuko, pers. comm).

Given these restrictions, it is clear that communities in open areas face additional
obstacles to benefitting from wildlife in a legal and sustained manner. Recognizing the
situation, ZAWA promotes community game ranching for local economic development
in order to make wildlife conservation profitable for the communities in open areas.
Game ranches are “owned by individuals and or consortiums (including business, trusts,
associations, non-govemmental organizations, community based organizations, and the
state), to sustain or enhance wildlife production for financial, economic, or social
purposes” (Davies et al., 1997 quoted in Chundama et al. 2004: 35). The idea is to

promote the establishment of joint ventures by private investors, chiefdoms and CRBs as

29

well as ZAWA. ZAWA has even indicated its willingness to restock such areas, for a fee,

to help develop these initiatives (Chundama et al., 2004).

2.4 Community Based Natural Resource Management

The new decentralized approaches to wildlife management and conservation were partly
a consequence of an international recognition that communities need to play a key role if
conservation objectives are to be met. Globally, donors (including USAID, the
Norwegian Agency for Development Co-operation and the Danish International
Development Agency, among others) have promoted initiatives that involve communities
in the decision making process, giving rise to various CBNRM initiatives in Africa and
Asia. Also, developing countries had found the centralized approach to be highly
burdensome, given their usually limited budgets, and had generally failed to control

poaching effectively (Leader-Williams and Albon, 1988).

The premise of the CBNRM programs is that “people who live close to a resource and
whose livelihoods directly depend upon it have more interest in sustainable use and
management than state authorities or distant corporations” (Li, 2002: 265). The CBNRM
model is based on the assumption that local communities will be interested and willing to
adopt and implement wildlife conservation programs as long as they are entitled to any
resultant ownership of resources and benefits (Songorwa, 1999). The proposition behind
CBNRM schemes is that they can serve two different goals at the same time:

conservation objectives (preservation of national parks, wildlife) and improved welfare of

communities.

30

In Zambia, building on the experience of the pilot project in Lupande GMA, the National
Parks and Wildlife Act (1991) made provisions for formal community participation and
benefit. Public debate, with the involvement of Parliament, began to focus on the
participation of local communities in wildlife conservation (Child, 2004; Chundama et
al., 2004). Wildlife management policies gradually evolved to incorporate mechanisms
for decentralization so the ADMADE approach could be replicated in GMAs throughout

the country.

2.4.1 Wildlife Management and Communities

CBNRM programs assume that the community is a homogeneous group of individuals
with similar interests, asset endowments, incomes and power relationships that facilitate
collective action. In fact, homogeneity of individual characteristics and interests is not the
reality in rural villages as pointed out by Agrawal and Gibson (1999). The incentive to
preserve wildlife is indeed strong, though only for some. Poaching is a coping
mechanism in GMAs where there is chronic food insecurity; it provides income and is
one of the only sources of protein for many households. For others such as farmers and

livestock owners, wildlife is mainly a threat to their livelihood (Songorwa, 1999).

The lack of attention to individual differences in communities has led to problems of
inequitable access to resources and distribution of benefits, and has subsequently
jeopardized the commitment of locals to preserve the resource base in the long run. In the
Lupande GMA, restrictions on female participation in community gatherings and

discussions, as well as the support for boys’ education over girls’ resulted in women

31

being further marginalized within the community, as their male counterparts became
better educated and had greater access to outside opportunities (IIED, 1994). Wainwright
and Wehrmeyer (1998) suggest that the lack of support for women in the Lupande GMA
“may impede development goals as it appears to focus narrowly on half of the human
population”. Key informants interviewed in Lupande and Chiawa GMAs in 2008 pointed
out that lodges rarely employ women (Adrian Coley, pers. comm), due in part to the
remoteness of the lodges, which are often located far away from the villages and with
limited accessibility. This makes it difficult for women to take care of the household and
the children while spending long periods of time at the lodges. This is another example of
how some groups within the communities might not benefit equally due to social

imbalances.

2.4.2 Costs and Benefits of CBNRM

For a CBNRM project to be successful, it has to “improve both the well-being among
local people and maintain (if not increase) biodiversity” (Wainwright and Wehrmeyer,
1998: 934). After two decades of CBNRM implementation, a body of literature has
emerged that examines the success of the co-management of natural resources (Gibson
and Marks, 1995; Marks 2001; Barret and Arcese, 1995; Mutandwa and Gadzirayi, 2007;
Child and Dalal-Clayton, 2004; Bwalya, 2003). Most of the case studies used primary
and secondary data. For example, focus groups and meetings with traditional leaders
were used to characterize the views of the CBNRM programs from local communities,
such as whether they thought it had a positive impact on their welfare. Other studies

relied on quantitative data such as number of arrests, total revenues received by the

32

CRBs, the level of employment generated through hiring of scouts, guides or cooks for
the safari lodges or the levels of meat consumed as tools to measure achievements. They
also describe the costs associated with the programs such as crop damage, loss of access
to meat, loss of wildlife use for traditional ceremonies or the social implications of

preventing hunting.

In Zambia, the majority of studies have been conducted in the Lupande GMA, where the
first CBNRM programs were implemented6. Based on a survey conducted in 1996,
Wainwright and Wehrmeyer (1998) identified some positive impacts of the LIRDP
program on attitude towards wildlife. Most respondents claimed that poaching had
decreased, and by a factor of more than two to one, local people claimed that wildlife was
more important at the time of the survey than before the project. However, poor
management, lack of accountability, and insufficient involvement of the community in
decision making were reported as some of the negative aspects. Approximately 57% of
respondents indicated that the implementation of the program had not increased their
living standards, 15% were neutral, 16% responded that the program had increased their
living standards “somewhat” and 12% responded that their living standards had increased

“very much”.

Dalal-Clayton and Child (2003) also examined the impact of the LIRDP program in
Lupande. Their study distinguishes between two stages of implementation. The first

stage, up to 1996, was marked by a concentration of decision making power among

 

6 Namely the LDP and the LIRDP.

33

district authorities and chiefs, who were trusted to act in the interest of community
members. Chiefs approved projects without adequate consultation with residents and
often appropriated meat intended for the community. As a consequence, community
members mistrusted the system and felt disenfranchised by the program. A change of the
LIRDP policy in 1996 shifted power from chiefs to grass-roots level institutions; VAGs
were set up to democratically decide on the use of the revenues. Accountability systems
were also put in place so the use of funds was transparent. The study makes reference to
studies conducted in 1998 that showed a change in communities’ perception about
wildlife, with a better understanding of the economic potential of wildlife and a greater

sense of ownership.

Bwalya (2003) analyzed the performance of the ADMADE program in the Blue Lagoon
GMA, which surrounds the Blue Lagoon National Park. He cites an increase in the
number of patrols and arrests as one of the achievements that reduced poaching and
contributed to increasing populations of large mammals. However, he describes the
shortcomings of the project in terms of weak community participation and involvement in
decision making, lack of transparency and accountability regarding the management of
funds, and poor information sharing on community entitlements. Although the
conservation revolving fund was established, local communities were not properly
informed about their entitlements. Furthermore, they felt that the revenues received from
ZAWA did not left much remaining for community development projects. As a result,
despite local interest in the programs, communities perceived a high opportunity cost of

wildlife management against other competing land uses such as livestock grazing,

34

thereby raising the incentive to poach. Wildlife farming was perceived to be
economically unprofitable, making the competition for pasture and water between

wildlife and cattle a source of dispute.

The findings in Bwalya (2003) coincide with the views of people interviewed in Chiawa
and Lupande GMAs during visits in July 2008. Chairpersons of CRBs and villagers
complained of inadequate funding received by the CRBs and lack of transparency by
ZAWA. They had no access to reports indicating the total revenues from hunting, so they
had to trust ZAWA’s management for equitable distribution. They also complained about
delayed payments, impeding the proper planning and prioritizing of projects. As a
consequence, locals often remain indifferent to the program. Although overall they
seemed to link the wellbeing of animals with potential benefits for themselves, they also
perceived costs, mainly due to conflict with elephants, which were unanimously blamed
for causing the worst crop damage. Elephant attacks destroy farms and houses and in
some cases claim the lives of villagers. Scouts are reluctant to kill the animals; rather they
try to scare them away by beating drums or shooting in the air which is not an effective
deterrent in the long-terrn. There is no official compensation mechanism for losses or
damages due to wildlife. Only some GMAs that receive significant funding such as
Lupande, can sometimes release money to rebuild a house or to buy food for affected

households.

In summary, the effectiveness of CBNRM programs is unclear. While there seems to be

some progress in terms of increasing wildlife populations, especially for large mammals,

35

and a change in attitudes towards wildlife, it remains unclear if communities actually
benefit, if at all, from the current arrangements given the losses they experience. The
studies cited above provide insight into community views and data on employment or
hunting revenues, but generally lack an analytical approach that accounts for the gains
and losses that communities might experience. For example, the revenues from hunting
and from employment in tourism can be significant, but these could be outweighed by

crop losses and the opportunity cost of alternative land uses.

The work conducted by Bandyopadhay and Tembo (2009), mentioned in Chapter one
provides the major contribution to the analytical work on the performance of GMA
policies in Zambia. The “Impact of Game Management Areas on Household Welfare”
(IGMAW) survey was commissioned by the Government of Zambia to study the impact
of GMAs on the welfare of the households living in them. The results of this study, which
has been a key source for the elaboration of this paper, will be extensively referenced in

the following chapters.

36

3. SURVEY DATA

The research uses data collected by the Impact of Game Management Areas on
Household Welfare (IGMAW) survey. The author conducted additional data cleaning

with the collaboration of Dr. Gelson Tembo, from the University of Zambia.

3.1 Coverage and Sample Design

The IGMAW survey was jointly commissioned by the Natural Resources Consultative
Forum (NRCF), the World Bank (WB) and ZAWA as part of an effort to inform policy
on the effectiveness of GMA interventions carried out by government, private sector and
the respective communities. The objective of the survey was to determine the impact of
game management areas on the living conditions of the households residing in the

GMAs.

3.2 Coverage

The survey covered areas adjacent to almost all the National Parks; only those in the
north and north-western part of the country were omitted for logistical reasons. The
remaining parks were grouped in park systems based on geographical location, as
follows: Bangweulu (including Isangano, Lavushi and Kasanka NPs), Kafue (including
Kafue, Blue Lagoon and Lochinvar NPs), Lower Zambezi (Lower Zambezi NP) and
South Luangwa NP. Each of the Park systems was considered a reporting domain in the

sampling process.

37

3.3 Sampling Process

Sampling was done in two stages. In the first stage the list of Standard Enumeration
Areas (SEAS) within GMAs was obtained by overlaying GMA digital maps from ZAWA
with maps of SEAS from the Central Statistical Office (CSO). All SEAS outside GMAs
but bordering a national park were also listed to serve as control areas. A sample of 139
SEAs was drawn from the two lists using probability proportional to size (PPS), and

drawing upon the 2000 census of population and housing.

At the second stage, all households in each SEA were listed, and sample households were
selected for interviewing using a systematic sampling scheme. The total number of
households interviewed was 2,769 out of a target of 2,800, reaching a 99% success in
response. Approximately half of the respondents reside in GMAS (58%) and the other
half in non-GMA or control areas (42%). Data were collected at the household and
community levels using household and community questionnaires, respectively. For the
community questionnaire, key informants were interviewed including the headmen,

chairpersons of CRBs, school headmasters, chairpersons of VAGs, and others.

3.4 Stratification of GMAs and non-GMAs

Following ZAWA classification of GMAS based on wildlife stocks and tourism activity,
the GMAS were classified into prime, secondary, specialized and understocked (see table
4 for distribution of SEAS among park systems and GMA vs. non-GMA areas). Prime
areas are those in which trophy species such as lion, leopard, roan, and sable antelope are
abundant and can sustain classic and mini safari hunting. Classic safari hunting permits

licensed hunters to hunt trophy species while mini safaris permit hunting of seven minor

38

 

species but exclude lion, leopard, and roan and sable antelopes. GMAs classified as
secondary are areas in which species are less abundant but can still sustain mini safari
hunting. Specialized GMAS are found in wetland areas and are characterized by the
presence of only a few species such as lechwe, Sitatunga and tsessebe. In understocked
areas, wildlife populations are Sparse and hunting quotas are limited (Simasiku etal.,
2008). ZAWA also classifies GMAs as depleted if wildlife is very fragmented and is
therefore not suitable for safari hunting. Because only one SEA belonged to this category,

it was dropped from the study.

Table 4. Sample SEAS by National Park and GMA classification

 

Park systems
Sample stratification Bangweulu Kafue Lower Zambezi South Luangwa Total

Primary 4 19 23
Secondary 2 l6 l8
Specialized 6 5 l l
Understocked 20 6 2 28
Non-GMA 10 21 17 1 l 59
Total 36 38 33 32 139

 

Source: Compiled from IGMAW survey.

39

4. ANALYTICAL APPROACH AND HYPOTHESIS TESTING

The objectives of this research are to analyze the impact of GMA policies on local
welfare, to identify the avenues through which this effect might be felt and finally, to

estimate the impact that each of these effects may have on household welfare.

Chapter one described the potential benefits and costs that households living in GMAs
may experience as a result of decentralized wildlife management policies. Greater access
to off-farm income opportunities through the tourism sector is the most direct link to
higher levels of income. However there could be other avenues for households to benefit
from (or be disadvantaged by) GMA policies. The following list enumerates the main

avenues through which GMAs may positively and negatively affect household incomes:

i) Hunting and photo-safari revenues are invested to fund community projects
such as the construction of schools, clinics, wells and health centers;

ii) Employment Opportunities arise as community members are hired as scouts to
patrol the areas;

iii) Communities are given free access to game meat from trophy hunting;

iv) Opportunities for off-farm employment through the tourism sector generate
greater levels of off-farm income;

v) Self-employment or business opportunities are generated through increased
economic activity in the area;

vi) Social capital rises as communities recognize the need to organize in

democratically elected CRBs/VAGs to administer local resources. These

40

organizations can be the base for other initiatives that communities may
explore with the private or public sector; for example the establishment of

game ranches or other business ventures.

On the other hand, there are potential costs associated with living in a GMA:

i) Communities give up the opportunity to utilize land in other ways that might
generate greater benefits than game management does (for example through
accessibility of land for livestock grazing or for cultivation);

ii) Households might be negatively affected by crop loss in well stocked GMAs;

iii) The ever present possibility of animal attack instills fear in the villagers and
may prevent some of the innovations for making money; night life, for

example, is greatly minimized.

The potential for the above mentioned factors to influence the income of households
living in GMAS will in turn depend on certain conditions, for example, on the effective
implementation of CBNRM policies. If hunting revenues are properly managed and
efficiently distributed to CRBs, they can be used to help better plan for the recruitment of
scouts and the implementation of development projects. Also, the status of wildlife stocks
will affect the ability of the GMA to generate funding and to sustain the CBNRM
programs. GMAs with higher numbers and variety of Species will have a greater chance
to generate funding from hunting revenues and attract a larger number of tourists,

increasing the opportunities for off-farm employment. All these factors that influence

41

household income and the conditions described above constitute what this research calls

the “GMA effect”.

The research is conducted in two stages. Stage one (Model 1) will estimate the GMA
effect on the level of household income. The specific question to be answered is: do
households living in GMAS realize higher average incomes from the current wildlife

management policies?

The first stage determines the existence and magnitude of a GMA effect. The second
stage (Model 2) represents the major contribution of this paper, which is to reveal
information about the avenues through which the GMA effect, if found, influences
household income. A positive GMA effect on household income could be generated
through increased non-farm employment opportunities in the tourism sector or through
the direct effect of CBNRM programs (development projects, hiring of scouts, and free
access to game meat). A negative GMA effect could be the result of increased crop
damage on the farms, which could outweigh the potential gains from the GMA policies.
To better understand the paths through which the GMA effect is felt, stage two will
analyze the GMA effect on i) sources of income; including crop agriculture and non-farm

earnings from wage and self employment, and ii) crop damage.
4.1 Model 1: Impact of GMA Policies on Household Income

As advanced in Chapter one, the livelihoods of households living in areas adjacent to

national parks are similar to the average rural Zambian. According to data from the

42

IGMAW survey, subsistence agriculture stands out as the main livelihood in both GMA

and non-GMA areas (see Table 5).

Table 5. Economic activities in GMA and non-GMA areas

 

 

 

GMAs Non GMAS
% of Mean Income % of Mean Income
Economic activities households earnings share over horseholds earnings Share over
engaged In among those all engaged 111 among those all
this activity engaged households this activity engaged household
Crop agriculture (total) 87% 1,010,918 38% 90% 1,338,792 41%
-Crop agriculture (sold) 43% 263,515 8% 49% 583,477 15%
-Crop agriculture (retained) 86% 747,404 30% 88% 755,314 26%
Livestock agriculture 67% 172,632 6% 73% 332,362 9%
Forest products 13% 231,491 1% 14% 431,947 2%
Wage employment 1 1% 5,373,789 9% 6% 9,791,228 4%
Selfemployment 50% 2,918,647 34% 46% 4,940,745 31%
Remittances 15% 1,618,911 5% 15% 1,736,153 6%
Other activities 19% 1,333,838 6% 19% 2,600,525 8%

 

Source: Calculated from IGMAW survey

The great majority of households in both areas participate in crop agriculture. This

activity has the largest income share over all households, with 41% in non-GMAS and

38% in GMAS. In both GMA and non-GMA areas, livestock agriculture is the second

activity in terms of number of households engaged, followed by self-employment, other

activities (including income from pensions, gifts, rents, game meat and sales of honey),

remittances, forest products and wage employment. Table 5 Shows that the relative

importance of self employment and wage employment is greater within GMAs in terms

of number of households engaged and income share over all households. This could

reflect the existence of additional off-farm opportunities in GMAs as a result of the

CBNRM programs or the result of land use plans that allocate arable land to the tourism

sector.

43

 

Table 6 shows a breakdown of income share of activities among households by income
quartiles. The share of income from agricultural (crop and livestock) activities rapidly
decreases as income increases. In GMAs, for the poorest 25% of hhs (1St quartile) crop
agriculture represents 56% of total income among all households, while for the least poor
25% (4th quartile) this share drops to 19%. In non-GMAS the trend is similar, where the
share of crop agriculture declines from 62% for the lSt quartile to 17% for the 4‘h quartile.
Conversely, the income share of self and wage employment increases as income rises.
For example, wage employment accounts for 0.4% of income over all households for the

1St quartile, rising to 13% for the 4th quartile in GMAS and from 0.4% to 22% in non-

GMAs.

Table 6. Income share over all households by income quartiles

 

 

 

GMAs Non GMAs

Economic activities lst qrtile 2nd 3rd 4th qrtile lst 2nd 3rd 4th
(a) qrtile qrtile (b) qrtile qrtile qrtile qrtile

Crop agriculture (total) 56% 48% 41% 19% 62% 46% 27% 17%
- Crop agriculture (sold) 14% 17% 19% 8% 10% 12% 8% 4%
- Crop agriculture (retained) 42% 32% 21% 11% 52% 34% 19% 13%
Livestock agriculture 15% 9% 7% . 3% 9% 8% 4% 2%
Forest products 2% 2% 2% 1% 2% 2% 1% 0%
Wage employment 0.4% 1% 4% 13% 0.4% 4% 11% 22%
Self employment 18% 26% 30% 49% 19% 31% 44% 43%
Remittances 4% 8% 9% 4% 5% 4% 4% 8%
Other activities 6% 6% 8% l 1% 2% 5% 10% 7%

 

Source: Calculated from IGMAW survey
(0) Lowest income
(b) Highest income

The descriptive statistics of the main variables selected for the analysis are presented in
Table 7. The total number of observations in the sample is 2,717 after excluding the zero

observations for total income. The subsamples for non-GMA and GMA are

44

approximately divided in half (58% of the surveyed households reside in GMAs and 42%
in non-GMAs). The t test for differences in means shows that the majority of variables
are significantly different across GMA and non-GMA areas. Households living in GMAs
are typically smaller in size, have younger heads of household, are more likely to be
headed by a female, have lower levels of education, are farther away from all-weather
roads, have fewer consumption and production assets and live in more densely populated
areas. The overall average household income is 4.2 million ZMK (801,544 ZMK per
capita). In GMAS the average income per capita is 706,844 ZMK while in non-GMAS the

average is 920,599 ZMK.

45

Table 7. Comparison between variable means for the full sample and the GMA and
non-GMA subsamples.

 

 

 

Subsanrples

Variable description Full sample GMAs Non-GMAs Sig.“
Number ofhouseholds 2,717 1,574 1,143

Total household income 4,235,762 3,591,253 5,123,301
Household size 5.28 5.08 5.57 ***
Total household income per capita 801 ,544 706,844 920,599

Age ofhousehold head in years 42.46 41.00 44.48 ***
Sex ofhousehold head (=1 ifmale) 0.74 0.73 0.76 **
Maximum education of hh member (in years) 6.78 6.42 7.27 ***
Number ofchildren (<15 years) 2.55 2.46 2.67 ***
Number offemale adults (15-60 years) 1.10 1.08 1.12
Number ofmale adults (15-60 years) 1.03 1.00 1.07 **
Distance to nearest all-weather road (in km) 5.09 6.08 3.80 ***
Cropped area (in hectares) 0.92 0.93 0.92
Value of consumption durable assets (in ZMK) 401,588 285,362 561,641 **
Value ofproduction durable assets (in ZMK) 618,036 256,729 1,115,584 ***
Popuhtion density (in sq km) 35.2 41.4 26.9 ***
Types of ex'sting Infrastructure (#) 3.62 3.64 3.59
Tourist lodge in the community (=1) 0.07 0.10 0.02 ***
Household lives in prime GMA (=1) 0.17 0.30 - n/a”
Household lives in secondary or special'med GMA (=1) 0.20 0.35 - n/a

 

Source: Calculated from IGMAW survey data

a . . . . . . .
Signlficance based on variance t test between means: *10% Significance, **5%srgnlficance,
M"‘10%significance

b n/a: not applicable

These results are similar to those obtained by Bandyopadhay and Tembo (2009), though
the absolute figures vary due to additional data cleaning conducted for this study. They
found that per capita expenditure levels are, on average, higher in non-GMAS (853,750
ZMK) compared to GMAs (839,359 ZMK), although the difference was not statistically
significant. This is also the case for per capita income; the average is found to be higher
outside GMAs than within GMAS (896,070 ZMK vs. 829,622 ZMK) but the difference is
not significant. The rest of the common variables used in both analyses (age, sex,

maximum education, number of children, female adults, male adults, distance to all-

46

weather road and value of consumption durable assets) follow the same pattern. The
mean comparison results suggest that households living in GMAS face more difficult
conditions (remoteness, lower education levels, lower incomes) than those living outside
GMAS. However, the comparison between means does not control for other factors that
affect household income. To isolate the GMA effect, other variables need to be kept
constant. The following sections present results of various econometric models that allow

uS to control for most of these other variables.

4.1.1 Model Specification

In this study, household welfare is measured by total income. All welfare indicators have
advantages and disadvantages. Barret et a1. (2001) advocate for the use of multiple
welfare indicators to cross check on inference. Although measuring household income
can be less straightforward than measuring expenditure, and is sensitive to
underreporting, it allows comparison with the results obtained by Bandyopadhay and
Tembo (2009), which was based on total household expenditure as a measure of welfare.
In addition, the income variable can be broken down into different sources, which will
allow exploring the channels through which the GMA effect is felt. Finally, the mean
income and expenditure figures calculated from the survey are very close, suggesting that

under-enumeration of income was not a serious problem in the survey.

Household income in rural areas comes from many sources. An income indicator is

created to capture the total value of household income, including farm income (total

value of sold and retained harvest, value of livestock sold and consumed and value of

47

livestock products sold), value of sales from honey, value of forest products consumed
and sold, income from hiring of equipment, income from game meat sold and off-farm

income (from wage employment and self employment).

To estimate the GMA effect, all other factors that affect household income need to be
held constant. Typically, the determinants of household income include human capital,
physical assets, locational characteristics, and other social and institutional assets (De

Janvry and Sadoulet, 2001). The relationship can be generally represented as:

Y =f(HC, PC, LC, SA) (1)

Where Y is the level of household income;

HC is a vector of human capital and socio-demographic variables;

PC is a vector of physical capital variables;

LC is a vector of locational variables; and

SA is a vector of social and institutional asset variables.
Human capital variables such as gender composition, household size, age of household
head and educational levels have an impact on the income strategies and earnings of
households. For example, households that have older members may tend to participate
less in the agricultural wage labor market and focus more on livestock activities,
particularly if the household head is male (De Janvry and Sadoulet, 2001). A married
household head may have higher earnings from farming than a female headed household
because of the availability of additional household labor, and may receive remittances
from children living outside the area. Education levels of adult members have an impact

on the allocation of labor. Education raises the marginal productivity off the farm as well

as the reservation wage, increasing the likelihood of entering the higher end of the labor

48

market. Households with better educated adults are expected to be wealthier as they are
able to attain the most remunerative income opportunities (De J anvry and Sadoulet,

2001).

In a rural economy where agriculture predominates, land access becomes a key asset for
household welfare. Jayne et al. (2003) find a positive association between household per
capita land holdings and per capita income in Ethiopia, Kenya, Rwanda, Mozambique
and Zambia. For small farms, a very small increment in land holdings is associated with a
large relative rise in income. The study also shows that the bottom 25% of rural
agricultural households have access to only 0.1 hectares per capita or less in each of the
examined countries. However, although this result might point towards a policy to
facilitate access to land, dynamic studies recognize the importance of off-farm income for

household growth, as will be discussed in the next section.

Locational characteristics determine household income opportunities and constraints in a
variety of ways. Access to markets, proximity to towns and higher population densities
generally enhance the capability of households to diversify their activities and participate
in the rural nonfarm sector. Income levels will be higher in areas with favorable agro
climate due to higher yields and the availability of jobs generated through production
linkage activities (Reardon, 1997). De J anvry and Sadoulet (2001) found that the number
of commercial centers within an hour traveling distance enhances nonagricultural wage
income in rural Mexico. Shorter distances to schools and clinics are generally positively

correlated with higher household welfare. Population density itself may have an

49

additional effect on income opportunities since it is typically positively correlated with
infrastructure variables, hence having the additional effect of facilitating exchange for

any given level of infrastructure.

Among social and institutional assets, De Janvry and Sadoulet (2001) found a positive
correlation between access to technical assistance and formal credit and household
income. Other variables that are likely to affect household income are related to
community structure and cohesion, for example communities that are able to organize
themselves and work together would be more likely to participate in CBNRM programs

which require a high level of community organization.

4.1.2 Model Estimation

The first consideration for selecting the model estimation is the distribution of the
dependent variable. In this case the dependent variable, household income, is continuous
with a small number of zero observations reported. Since the survey included an
exhaustive list of incomelsources, including farm and off-farm, gifis in-kind or in cash,
remittances and pensions, it is likely that the reported zero observations correspond to
missing or incorrectly recorded data. Since missing observations represented only 1.9%
of the total sample, they were dropped for the purposes of this study. Eliminating the
zero values makes the income variable positive for all observations and allows the use of

Ordinary Least Square (OLS) estimation.

Although OLS is a simple and trusted estimation technique, one obstacle for its proper

implementation is the potential presence of endogeneity, which creates bias and

50

inconsistency in the parameters. For the case of household incomes in this study,
endogeneity is suspected to be present, because of sample selection bias derived from the
household’s decision to migrate to or from GMAS. Households migrating into GMAs
may be attracted by employment opportunities or existing amenities derived from the
investment in community projects. Those households emigrating from GMAS (perhaps
those most oriented towards agriculture) might do so as a consequence of human-wildlife
conflicts. Bandyopadhay and Tembo (2009) acknowledge the problem of endogeneity
and note that the absence of quantitative historical data for the creation of GMAS could
indicate that there are unobserved factors that cause selection bias. To control for the bias
they use Maddala’s (1983) treatment effects model. This model considers the effect of an
endogenously chosen binary treatment (in this case, living in a GMA) on another
endogenous continuous variable, conditional on two sets of independent variables. The
model uses either a two-step consistent estimator or full maximum likelihood to obtain

the estimates (Cong and Drukker, 2001).

Effective control of selection bias in treatment models requires a set of instrumental
variables (IV) that are correlated with the probability of living in a GMA (the endogenous
treatment variable) and uncorrelated with the level of household income (the dependent
variable of the continuous equation) as explained in Wooldridge (2002: 463). Yet in this
specific application, all of the variables thought to affect the probability of living in a
GMA (for example household participation on a VAG committee or the number of
community projects funded by the CRB) are also likely to influence the level of income.

Moreover, a comparison of results obtained from treatment effects models using various

51

combinations of these IVs shows that results are highly sensitive to which IVs are chosen

(see Appendix A).

Appendix A investigates further the extent to which household migration to or from
GMAS should be a concern for this analysis. It compares migration percentages in GMAS
and non-GMAS and conducts a series of tests to determine i) whether the total percentage
of migration is large enough to suggest a serious potential sample selection bias and ii)
whether there are structural differences between the subsample of households that have
not migrated vs. those that have migrated. The decision is to estimate the model using
OLS given the relatively low levels of migration and the results of the Chow (1960) test

for structural differences (see Appendix A).

The basic OLS estimation for the determinants of income takes the form7:

[MI], 2 a+fl1HC+ flzPC‘l” fl3LC+ fl4S/I +fli (2)

Where i takes the values from 1 to 2,717
lnY,‘ is the logarithm of income for each household. The logarithm is used as
the right hand side variables are expected to affect the dependent variable
in percentage terms
HC is the vector of variables associated with human capital and associated
demographic factors
PC is the vector of variables associated with physical capital

LC is the vector of variables associated with location and community
characteristics

SA is the vector of variables associated with social and institutional assets

at is the intercept term

,8; is the vector of parameters associated with the human capital vector, HC

,6; is the vector of parameters associated with the physical capital vector, PC

 

7 Stata’s hetresr suggested strong heteroskedasticity with respect to the set of right hand side variables. All
inference in this model is therefore made on the basis of standard errors robust to this problem (produced
through the hc3 option under the regression command).

52

,83 is the vector of parameters associated with the location vector, LC

,64 is the vector of parameters associated with the social and institutional
vector, SA
#i is the error term

4.1.3 Stage 1: Empirical Estimation and Results

The selection of variables included in the model is guided by a review of literature on the
determinants of rural household income (Barret et al., 2001; De Janvry and Sadoulet,
2001; Yunez-Naude and Taylor, 2001; Reardon, 1997), the factors hypothesized to have

a positive or negative impact on household income, and data availability.

Human capital and socio-demographic variables include household characteristics such
as the age in years of the household head; the sex of the household head, represented by a
dummy variable that takes the value 1 for male headed households; the level of
education, measured by the number of years of schooling of the highest educated
household member; and the gender distribution of the adult members of the household

(number of females and males between 15 and 60 years).

Total area cropped in hectares is used as a proxy for land holdings since the latter is not
available from the survey data. Other assets are represented by a productive assets
variable, generated by adding the ZMK value of assets such as tractors, ploughs, wheel
barrows, fishing nets and traction animals (e. g., oxen, donkeys). The consumer durables
variable represents items such as radios, refrigerators, cell phones, bicycles, and sewing

machines.

53

The vector of locational variables describes community characteristics in terms of
location and availability of amenities which are hypothesized to have an effect on
opportunities for employment and access to markets. The selected variables include the
distance in kilometers to the nearest all weather road; an infrastructure index created by
counting the existing types of infrastructure funded by Government, non-govemmental
organizations (NGOS) or private companies, including schools, clinics, wells, and dip
tanks; a dummy variable to indicate the existence of a tourist lodge in the community;
and the population density of the SEA. The density variable is included in order to
capture (a) any remaining unobserved aspects of infrastructure and (b) opportunities for

exchange, holding infrastructure constant.

Finally, to analyze the effect of the GMA on household income, the model includes
dummy variables to indicate whether the household lives in a GMA. Since we
hypothesize that the stock and variety of wildlife has an impact on the potential to
generate hunting revenues and the level of crop destruction, the GMA variable is
classified into GMAl for prime areas and GMA2 for secondary and specialized areas.
The former takes the value of 1 if the household lives in a prime GMA (well stocked with
a high variety of species; otherwise, the value is zero) and the latter takes the value of 1 if
the household lives in a secondary or specialized area (lower stocks and variety than in

prime areass; otherwise, the value is zero).

 

8 See Chapter 3 for a description of ZAWA’s GMA classification.

54

The results of the OLS regression are presented in Table 8. The coefficients obtained
from the OLS estimation have the a priori expected signs and for the most part are
significant at a 1%, 5% or 10% level. The age of the household head is negatively and
Significantly associated with household income. Male-headed households Show a positive
but insignificant association. The level of education (maximum education level of any
household member) is, as expected, significantly associated with higher levels of income.
An additional year of education of the highest educated household member increases total
household income by 4.3%. The number of adults (men and women) is significant and
positive, which is an expected result since income is aggregated at household level.
Distance to the nearest all-weather road has a negative effect on income by 5%. This
result is consistent with the hypothesis that remoteness has a negative effect on household
welfare by limiting opportunities for off-farm employment, raising the cost of transport,
limiting access to markets, and increasing transaction costs (e.g., access to information,

search costs).

Other factors positively and significantly affecting household income are the existing
level of infrastructure, population density, and the presence of a tourist lodge in the area.
Infrastructure and population density have been found to be positively associated with
wage earnings by Reardon (1997) and Haggblade et al. (1989). Infrastructure levels may
be associated with a reduction in transport costs, increased access to markets, greater
provision of services (banks, extension services) and facilities (clinics, schools, wells)

and greater access to employment opportunities. Population density is generally

55

positively associated with income too; for any given level of infrastructure, population

density generates greater opportunities for exchange.

56

Table 8. OLS regression on the GMA effect on total household income

 

Variable description

Coefficient
(standard error) Significance

 

 

Intercept 13.101 ***
(0.12)
Age of household head in years -0.003 *
(0.00)
Sex of household head (=1 if male) 0.069
(0.06)
Maximum education of hh member (in years) 0.043 ***
(0.01)
Number of children (<15 years) 0.019
(0.01)
Number of female adults (15-60 years) 0.113 ***
(0.03)
Number of male adults (15-60 years) 0.070 **
(0.03)
Distance to nearest all-weather road (in km) -0.005 ***
(0.00)
Cropped area (in hectares) 0.039 *
(0.02)
Value of consumption durable assets (in ZMK) 0.020 ***
(0.00)
Value of production durable assets (in ZMK) 0.010 ***
(0.00)
Population density (in sq km) 0.001 ***
(0.00)
Types of existing Infrastructure (#) 0.032 ***
(0.01)
Tourist lodge in the community (=1) 0.186 *
(0.10)
Household lives in primary GMA (=1) 0.170 **
(0.08)
Household lives in secondary or specialized GMA (=1) 0.022
(0.07)
Dependant variable: Total household income (in logarithm)
Number of Observations 2264
R-squared 0.21

 

Source: Calculated from survey data.

*10% significance level, “5% significance level, ***1% significance level

57

Finally, results Show that living in a prime GMA increases households incomes by 17%
over what they would be without the GMA designation. For households living in
secondary or specialized GMAs, the result is positive though not significant and
relatively low in absolute terms. This finding confirms the results in Bandyopadhay and
Tembo (2009) who found a positive association between living in a GMA and household
welfare (measured by consumption expenditure). However, by breaking GMAs down by
classification, we show that the GMA effect is dependent on the level and variety of
wildlife population. The results suggest that any loss from crop damage is more than

offset by the positive impacts.

To investigate how the positive GMA effect is distributed across households, we
stratified households by the value of their durable asset holdings (the variable consumer
durable assets is stratified by quintiles), then repeated the OLS analysis while interacting
the GMA variables with the dummies for lowest two quintiles, lowest three quintiles and
upper two quintiles (Appendix B). The results indicate that the GMA‘effect is more likely

to be attained. by wealthier households (see Table 9).

Table 9. Comparison of GMA effect on household income by welfare level

 

 

Consumer durable assets GMA 1 GMA 2
Lowest 2 quintiles 0.033 -0.059
Lowest 3 quintiles 0.031 0.04
Upper 2 quintiles 0.046“ -0.008

 

"5% Significance level

The lowest 2 quintiles refer to the poorest 40% of the population, who according to the

results are not significantly impacted by living in a GMA; the same result applies if the

58

analysis is expanded to include the poorest 60% segment of the population. Only when
the upper two quintiles are considered do the results become positive and significant,
indicating that the gains derived from living in a GMA are likely to be attained by the
non-poor segment of the population; this too is consistent with Bandyopadhay and
Tembo (2009). It is also worth noting that the impact is insignificant for all segments
living in secondary or Specialized GMAs. That wealthier households capture the positive
impact of the GMA effect is not surprising. They are in a better position (in terms of
access to financial, human and political capital) to take advantage of the opportunities
offered in the non-farm sector as entrepreneurs and as wage employees (Haggblade et al.,

2007)

An additional regression by park systems was conducted (for full results see Appendix
B), interacting the park dummy variables with a single GMA variable.9. The results of the
coefficients are presented in Table 10. The highest GMA effect is found in Luangwa
where households have, on average, 18% higher incomes in GMAs than in non-GMAS.
In Kafue, households in GMAS have 14% higher incomes and in Lower Zambezi, the
GMA effect is positive though not significant. The results are probably a consequence of
the distribution of prime GMAs across the parks: Luangwa includes 19 prime GMAs and

Kafue 4, while Lower Zambezi has no prime GMAs.

 

9 The two GMA variables were collapsed into one because using park dummy variables introduces its

own control for wildlife population and variety that was previously captured with the two GMA dummies

59

Table 10. GMA effect by park system

 

 

Var'nble description Coefficient Significance
Households living in GMA within Luwangwa Park system 0.177 **
Households living in GMA within Kafue Park system 0.135 *
Households living in GMA within Zambezi Park system 0.046

 

Source: Calculated from IGMAW survey data.
Control park system: Bangweulu.
Full results shown in Appendix B.

In summary, the results suggest that, ceteris paribus, there is a positive association
between prime GMAs and household income. The benefits derived from living in a GMA
(mainly through tourism and CBNRM programs) seem to outweigh the potential costs
associated with living in GMAs (mainly the possible opportunity cost of land use and the
increased probability of crop damage). The fact that the effect is only significant in
GMAS classified as prime, indicates that the state of wildlife population is a key factor
for the potential of tourism and CBNRM programs to generate employment and hunting
revenues. The results also reveal that benefits are more likely to be attained by those
groups in the upper quintiles of the welfare scale, suggesting an uneven distribution of
the GMA effect among community members, a common finding in the literature of

welfare and non-farm rural income (Haggblade et al., 2007).

4.2 Model 2: Avenues through which the GMA Effect is Generated

The results of the first analysis identify the effect of GMAS on total income but provide
no insight into the specific avenues through which the GMA effect is generated. To shed
light on this, Model 2 analyzes the GMA effect on a) the different sources of income;
namely wage employment earnings, self employment earnings and crop agriculture and

b) crop damage.

60

Participation in non-farm activities is typically associated with relatively high levels of
welfare. Therefore, this study investigates the GMA effect on the probability of
participating in off-farm activities and on the level of earnings conditional on
participation. The analysis also looks at the GMA effect on agriculture to evaluate
whether the land use planning and crop damage associated with GMA policies impact the
probability of farming and the value of the total crop. Finally, living in GMAs could
negatively impact income via higher crop damage by wildlife. Model 2 analyzes each of

these effects separately. The specific questions of interest are:

i) Does living in a GMA increase the likelihood that a household will participate in
non-farm employment activities? If so, how does living in the GMA affect the
total value of non-farm earnings?10

ii) Does living in a GMA increase the likelihood that a household will participate in
crop agriculture? And if so, how does living in the GMA affect total crop value?

iii) Does living in a GMA increase the likelihood that a household is affected by crop
damage? And if so, how does living in the GMA affect the total value of crop

1055?

4.2.1 GMA Effect on Off-F arm Earnings

Empirical studies show that non-farm activities are typically positively associated with
income and wealth in Africa (Barret et al., 2001; Reardon, 1997). “Push factors” capture

the fact that some households diversify not so much due higher income earnings

 

10 For the purpose of this paper, non-farm earnings are defined as wage earnings received from the farm or
non-farm sector and self-employment earnings from non-farm businesses.

61

possibilities but rather to manage riskand a cope with adverse Shocks. The dependency
on rain-fed agriculture and the absence of credit markets are two examples of the risks
that farmers face. Participating in the wage labor market (typically on neighboring farms)
or setting up their own micro-enterprises are some strategies to assure food security in the
event of a drought or a sudden reduction in commodity prices for example. Non-farm
employment is also sought by rural households in the presence of land constraints derived
from population growth and decreased productivity (Reardon, 1997; Barret et al., 2001).
“Pull factors” capture household response to greater income earnings opportunities off
the farm. For example, households that reach certain levels of labor productivity on the
farm can release labor to the non-farm labor market. Others may diversify to take
advantage of production-linkage activities, for example milling and hog production
(Barret et al., 2001). Education is also positively associated with off-farm income, as
better educated households have more opportunities for wage employment and typically

access higher remunerated jobs.

Of the 2,717 households included in this analysis, 242 (9% of the total sample) reported
wage earnings and 1,322 (almost half) reported earnings from non-farm businesses; 79
households (3%) reported having earnings from both sources. Table 11 shows that
households employed in the wage labor market have higher average incomes (4.6 times
higher) compared to households not receiving wage or self employment earnings. They
are also more likely to be younger and male-headed, better educated and larger in size.
The average distance to a main road is 2.9 km for wage-earning households compared to

4.8 km for non-wage and non self-employed households. Tschirley and Benfica (2001)

62

find the same pattern in rural Mozambique when comparing household and community
characteristics of wage-earning and non-wage households. The table also shows that
households participating in the wage labor market are more likely to live in more densely
populated areas with greater access to infrastructure. Wage earning households are also
more likely to live in places where there is a tourist lodge and in GMAs as compared to

the group of households not participating in wage or self-employment earning activities.

Table 11. Mean comparison between households earning wage income and
households not earning wage income

 

Mean values lbr subsamples
Households Households Households not
eaming earning self- earning wage

wage employment or self-

 

 

 

Var'nble description income mcome employment

Total income per capita 2,225,952 1,070,635 487,190
Age of household head in years 39.13 40.20 45.43
Sex of household head (=1 if male) 0.83 0.78 0.69
Maximum education of hh member (in years) 9.28 6.77 6.46
Number of children (<1 5 years) 2.63 2.69 2.42
Number of female adults (15-60 years) 1.29 1.17 1.01
Number of male adults (15-60 years) 1.20 1.08 0.96
Distance to nearest all-weather road (in km) 2.91 5.70 4.77
Cropped area (in hectares) 0.74 0.84 1.03
Value of consumption durable assets (in ZMK) 1,026,838 464,244 278,084
Value of production durable assets (in ZMK) 761,126 718,125 491,208
Population density (in sq km) 89.70 39.38 26.42
Infrastructure (#) 3.78 3.80 3.45
Tourist lodge in the community (=1) 0.17 0.08 0.05
Household lives in prime GMA (=1) 0.33 0.20 0.12
Household lives in secondary or specialized GMA (=1) 0.35 0.20 0.19
Number of Observations 242 1,322 1,232

 

Source: Calculated from IGMAW survey
Significance based on variance t test between means among indicated households compared to those with
neither wage nor self-employment income: *10% significance, ** 5% significance, *** 1% significance

63

A5 with wage-earning households, households that operate their own businesses (i.e.,
self-employed) have higher mean total income compared to the third group (households
notearning off-farm income). They are also better educated, more likely to be headed by
a male, have younger household heads and are larger in Size than households not
receiving wage or self-employment earnings. However, self-employed households are
more isolated than wage-earning households or those with no wage or self-employment
income (the average distance to a main road is higher for self-employed households).
Like wage-earning households, self-employed households tend to live in more densely
populated areas with higher access to infrastructure and are also more likely to live in

GMAs.

These results suggest that households participating in the wage labor market and those
who are self-employed are generally better off than non wage-earning and non self-
employed households. However, as indicated by Tschirley and Benfica (2001), using
cross-sectional data does not allow for the observation of trends over a period of time, so
the differences could be due to specific events that occurred during the year the data were
collected rather than long-term factors. In any case, literature on rural welfare (Barret el
al., 2001; De Janvry and Sadoulet, 2001; Reardon, 1997) and off-farm income presents
compelling evidence for the positive relationship between income diversification and
higher total income, which points to income diversification as a way out of poverty.
However, other studies have shown that wealthier households have greater opportunities
to enter the non-farm labor market (Barret et al., 2001), and that once they have accessed

the high end of the wage labor market, they tend to stay there and build up advantages

64

over time (Tschirley and Benfica, 2001). Less privileged community members are likely
to face entry barriers to participate in the non-farm employment sector, further

intensifying welfare disparities within the population.

4. 2. 1.1 Model Specification

The number of households reporting wage earnings and self-employment earnings is 9%
and 49% respectively. This model falls under the category of comer solution models
(Gujarati, 2003; Wooldridge, 2002), in which the dependent variable takes a zero value

for a non-trivial part of the population and the values greater than zero are continuous.

The presence of a large number of zeros for the dependent variable leads to biased and

inconsistent parameter estimates from OLS. Excluding the zero observations would not

guarantee that E(u,°)=0 which is a violation of the Gauss-Markov Theorem (Wooldridge,

2002)

The Tobit model is often used in comer solution samples. It is commonly expressed in

the following form:

Yi =fl1+ flgXi‘i'ul' IfRHS>O (3)
= 0 otherwise

Where RHS represents right-hand side explanatory variables and Y ,- is the observed value

of the dependent variable.

65

The technique for Tobit models involves maximum likelihood (ML) estimation. The

method defines a latent variable Yit as:

Y I": X 136 + 51° where 8i e Normal (0, 0'2) (4)

The latent variable satisfies the classical liner model assumptions; it has a standard

normal and homoskedastic distribution with a conditional mean (Gujarati, 2003).

The relationship between the observed Y i and the latent variable Yit is expressed as

follows:

Yi= Y,‘ I ifo>0
0 ifo<0

or alternatively, Y i =max (0, X ifl + 8,” ) (5)
Because Y i * is normally distributed, Y ,- has a continuous distribution over strictly positive

values. The density of Y given X is the same as the density of Y 1" given X for positive

values. It follows that:

P (Y,-’ s 01 X,) = 1 - o (XI-,6 /0') and (6)
P 01‘ >01 X.) =(1/o) - o [(Yi-Xifl/o] (7)

Combining both probabilities we obtain the log likelihood function:

”(13:07: I (Y.- s 0) log [1- e (Xifl/0)]+ 1(Yi>0)108{(1/0') — (0 [(Yi — new]; (8)

66

Where [i is the log likelihood function

CI) is the standard normal cumulative distribution function (cdf)
(o is the standard normal density
0' is the standard deviation

Finally, the estimates of ,6 and care obtained by maximizing the log likelihood.

One shortcoming of the Tobit model is that it calculates the determinants of the
probability of an outcome being positive and the magnitude of the effect of these
determinants on the dependent variable simultaneously. The model estimates only one set
of coefficients, which are assumed to be equal for both equations (the probability and the
level of output). However, it is not always reasonable to assume that the explanatory
variables should be the same for both estimations and affect the probability of y > 0 and
the total value of y in the same way. Lin and Schmidt (1984) illustrate this providing an
example of fire 1055 using a random sample of buildings. If the dependent variable is the
“value of loss due to fire” then the Tobit model would assume, for example, that the age
of the houses would have the same effect on the probability of a fire as it would have on
the effect of total value lost in the fire. However, it is possible that while newer houses
may have a lower probability of fire (due to improved wiring, updated appliances, and
modern materials), once involved in a fire, the losses from newer houses could be

considerably higher.

For this analysis, it is unclear how some of the variables will affect the probability and
the level of Y respectively. For example, the age of the household head could be

negatively associated with the probability of accessing a remunerated job, but once it is

67

accessed, the individual could have greater earnings due to accumulated experience. A
model that estimates the probability and amount equations separately is preferred since it
will reveal more information about the effect of the independent variables on access to

different income sources.

The Cragg Tobit alternative model (Cragg, 1971) presents a variation of the Tobit model
that allows for the two estimations to be determined in two different stages or hurdles. In
this study, the “double hurdle” model takes into account that individuals have to pass two
hurdles to participate in off-farm work. First they have to desire to participate in off-farm
work and second they have to be able to overcome potential barriers to join the labor

market such as transaction costs, gender barriers, etc. (Woldehanna et al., 2000).

Non-farm employment can be treated as a latent variable, which may be observed when a
household decides to participate in non-farm work and is able to participate in the labor
market. Let D,- be the household’s decision to work outside their own farm. As
Woldehanna et al. (200022) point out, “in an agricultural household model an individual
is willing to participate in non-farm work when his reservation wage (wri), is less than
the non-farm wage (Wmi) net of commuting and expected transaction costs”
(Woldehanna et al., 2000z2). Di or the participation decision can be expressed as a latent

variable:

Di *= Xi 'y +211: where #1" ~ Normal (0, 0'2) (9)

68

Di: 1 ifof>0 or(wri<wmy (10)
0 if Dl'*<0 or (wriZwmI)

The next step is to specify the relationship between the participation decision and the
level of the outcome. The Cragg double-hurdle model is specified by modifying the
standard Tobit model where is a latent variable as defined in equation (4). The
relationship between the latent variable Y; t, the observed variable Y; and the

participation decision can be expressed as follows:

Y): Y,” if Yi‘>0 andDi=1 (11)
owl-=0 (Yr-30)

In this specification, the model allows separate sets of factors to influence the decision to
participate in non-farm employment and the level of non-farm earnings. X ,- ' and X i are
the vectors of explanatory variables that affect the two-stage processes. The error terms in
both equations are assumed to be normally and independently distributed, which implies

that any unobservable variables do not affect both stages.

In the independent double hurdle model, the probability that non-farm income (Yi) is
zero is the product of the probability that the latent variable representing non-farm
income (Y ,- I) is negative and the probability that (1’; ~‘) is positive, introducing the
possibility that the household would chose to participate in the non-farm labor market but
there are barriers that prevents it from doing so (Woldehanna et al., 2000). Consequently,

the log likelihood function for the independent double hurdle takes the form:

69

[FIG/1'5 0) logo-ab (X.- '7) Mrfl/o)]+1( Y.- >0) Iogrra/a) (M Y,- -Xifl)/0] M.- '7) 02)

(a) (b)

 

The first term (a) corresponds to the contribution of all the observations with an observed
zero. It indicates that the zero observations are obtained not only from the households
that decided not to participate in the non-farm labor market but also from the households

that decided to participate but could not do so due to entry barriers.

As with the Tobit model, the parameter estimation of y, ,6 and ais done through
maximum likelihood. The parameter results can be obtained using the command Craggit

built for Stata9 (Burke, 2009).

4. 2. 1. 2 Model Estimation

Given the advantages of Cragg over the Tobit model, the former is used to estimate the
probability of households receiving non-farm income and the determinants of the total

non-farm household earnings.

The independent variables selected for the analysis are the same as were used in the
model for total farm income as these are also hypothesized to influence the probability of
participating in non-farm employment and the total value of non-farm earnings. The

Cragg model for the following analysis can therefore be expressed as:

P(Di =1| Xi) =7Xi +r1i (Tier 1) (13)
Ln Y1: = a + ,BX,‘ +51 (Tier 2) (14)

70

Where D; is the participation decision variable which takes the value 1 if the
household decides to enter the non-farm employment sector

Y; is the total value of earnings from non-farm employment

Xi is the vector of household and community characteristics, assumed to
be the same for equations (1) and (2)

yis the vector of coefficients associated with the X i in the probability
equation

,6 is the vector of coefficients associated with X i in continuous equation
a is the intercept in the continuous equation

4. 2. 1.3 Empirical Results of GMA Eflect on Wage Employment

As mentioned in the previous section, one of the advantages of the Cragg model is that it
allows one to obtain different coefficients for the two stages, making it possible to
analyze the impact of the explanatory variables on both the household’s probability of
entering the wage labor market and on the total value of the wage earnings. The results of

the Cragg estimation are presented in Table 12.

The first two tiers Show the results of the probit and continuous regressions respectively.
For easier interpretation, the coefficient for the first tier is displayed as the marginal
effect in the third column (probit mfx) and the second tier is displayed as average partial
effects (APE) in the last two columns (for calculation of APES, see Burke, forthcoming).
The third column (probit mfx) represents the marginal effects of the independent
variables on the probability of participating in the wage labor market. The fourth column
(CAPE) represents the average partial effectll on the conditional expected value of wage

earnings E(Y|X i: Y > 0). This coefficient measures the effect of the independent variables

 

H The terms “marginal effect” and “partial effect” are used to be consistent with the literature on these

techniques but refer to the same thing.

71

on the level of earnings only for the households who participate in wage earnings. The
fifth column (UAPE) represents the average partial effect of the unconditional expected
value E(Yle°). This coefficient measures the effect of the independent variables on the
level of wage earnings, unconditional on household participation in wage income. The
UAPES are therefore dependent on both stages of the estimation, the probit and the
truncated regression, and their interpretation must consider both results. From a policy
perspective, the UAPEs represent the overall expected impact of a GMA on the variable
of interest, and are therefore useful as a summary indicator of the benefit of the GMA

approach.

72

Table 12. Cragg results for wage employment earnings

 

Marginal Effects

 

 

Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept -3.084 *** 15.533 *** n/a n/a . n/a
Age of household head in years -0.004 -0.017 -0.0004 -0.017 -0.009
Sex of household head (=1 if male) 0.259 ** -3.172 ** 0.023 ** -3.109 "' 0.234
Max. education hh member (years) 0.156 *** 0.169 0.015 *** 0.166 0.331 ***
Number of children (<1 5 years) -0.008 -0.306 -0.001 -0.300 -0.043
Number of female adults (15-60 years) 0.001 0.166 0.0001 0.162 0.175
Number of male adults (15-60 years) 0.015 2.502 *** 0.001 2.452 ** 0.259 *
Distance to nearest road(km) -0.008 ** -0.016 -0.001 ** -0.016 -0.018
Cropped area (in hectares) -0.046 -0.086 -0.005 -0.085 -0. 101
Value of consumption assets (ZMK) 0.017 *** 0.021 0.002 "* 0.021 0.037 "*
Value of production assets (ZMK) -0.010 “'* -0.002 -0.001 *** -0.002 0.035 **"‘
Population density (in sq km) 0.002 *** 0.000 0.003 "* 0.0002 0.004 ***
Infrastructure (#) 0.028 "‘ -0.207 0.003 * -0.203 0.038
Tourist lodge in the community (=1) 0.476 *** 0.718 0.066 *** 0.703 1.031 ***
Prime GMA (=1) 0.577 *** -0.090 0.078 *" -0.088 1.160 *"
Secondary or specialized GMA (=1) 0.567 *** 1.046 0.074 *** 1.025 1.245 "*

 

Number of observations: 2264

Log likelihood: -527.9

Source: Calculated from survey data.

*10% significance level, "5% significance level, *** 1% significance level
n/a: not applicable

The results of the probit (third column) have in general the a priori expected signs. Male
headed and better educated households have a higher probability of entering the wage
labor market (De J anvry and Sadoulet, 2001; Reardon, 1997). Distance to the nearest all-
weather road also has the expected sign; an additional 10 km from a main road decreases
the probability of being employed by 1% (a rather low figure, though it should be noted
that only 9% of the sampled population have such employment). Population density and
infrastructure are also found to be positively associated with the probability of wage
earnings; more densely populated areas with better infrastructure are likely provide
opportunities for wage employment. The existence of a tourist lodge has a comparatively

large impact on the probability of wage earnings: households living in communities

73

where there is a tourist lodge are 6.6% more likely to earn a wage compared to

households living in areas in which there is no tourist accommodation.

The coefficient for the variable representing consumer durable assets has a positive and
significant effect on access to wage labor. This can be an indication that, as highlighted
by the empirical literature on non-farm income, wealth and participation in wage
employment are typically positively associated. Conversely, agricultural productive
assets and cropped area decrease the probability of participating in wage employment.
The negative association may reflect the cumulative effect over time of decisions to focus
on agriculture. This result could also point to the fact that households that do not access
enough land to attain food security may search for off-farm opportunities to complement
their incomes (Beyene, 2008). The negative or statistically insignificant relationship
between land holdings and participation in wage employment is consistent with the

findings of other studies (Tschirley and Benfica, 2001; Yunez-Naude and Taylor, 2001).

The main variable of interest, the GMA effect, shows positive and Significant results.
Households living in prime GMAs are 7.8% more likely to be employed than households
living in non-GMAS. The results are similar for households living in secondary or
specialized GMAS, in which they are 7.4% more likely to enter the wage labor market.
This result can be interpreted as being a consequence of greater employment
opportunities in GMAs through the tourism sector and through the employment of scouts;
however, it could also suggest that households living in GMAs have lower reservation

wages due to lower agricultural productivity and higher chances of crop destruction (push

74

 

factors), and therefore are more prone to search for off-farm wage employment
opportunities, though not necessarily outside the agricultural sector. The second stage

provides useful insights for a more careful interpretation of results.

Significance testing levels for CAPEs and UAPES was done through bootstrapping. In the
case of CAPES, only the sex of the household head and the number of male adults in the
households are significant. The CAPE for the variable representing sex of household
head indicates that wage earnings are three times lower for those households that
participate in wage employment, which contradicts the a priori expectation. However, the
number of male adults is positively associated with level of wage earnings for those

households engaged in the labor market.

The interpretation of UAPES takes into account the entire population and the results of
both tiers. For example, the UAPE indicates that the prime GMA effect on the level of
wage earnings is positive and statistically significant (116%) despite a negative CAPE (-
8.8%). The positive UAPE reflects the effect of the higher probability (7.8%) of a
household having wage earnings when living in a prime GMA, which outweighs the
lower wage incomes in prime GMAs for those engaged in the labor market. In summary,
the UAPE indicates that, in the absence of any ex ante knowledge of a household’s
employment status, households in prime GMAs can be expected to earn 116% more on
average from wage labor than households outside GMAs. Note also that this effect
obtains even while controlling for the presence of a tourist lodge. Due probably in part to

this control for a lodge, results are similar in secondary or Specialized GMAs: households

75

living in such GMAS earn, on average, 124% more than households living outside

GMAs.

One explanation for the negative CAPE for prime GMAs is the effect of in-migration on
local wages. Prime GMAS are likely to attract job seekers in search of job opportunities
given by the tourism sector which could eventually saturate the labor market and drive

wages down.

An analysis of the GMA effect on wage income by welfare segments in presented in
Table 13. The Cragg estimation was conducted by welfare subsamples (lowest two
wealth quintiles and highest two wealth quintiles) using the consumer durable assets as a
welfare indicator. The results clearly Show a positive and significant GMA effect on the
probability and level (UAPE) of wage employment for all households, poor and non-
poor. The effect is slightly higher for the highest two quintiles though this difference

might not be statistically significant.

Table 13. GMA effect on wage income by welfare levels

 

 

 

GMA type/wealth quintile Tier 1 Tier 2 Probit mix CAPE UAPE
Prime GMAs
Lowest 2 wealth quintiles 0618*” -0.38 0.057*** 0.34 0.784”
Top 2 wealth quintiles 0762*“ -l.406 0.144*** - 1.400 1.726***
Secondary GMAs
Lowest 2 wealth quintiles 0655*“ 2.055 0.057*** 1.838 0.900M
Top 2 wealth quintiles 0.424“ -0.290 0.071M -0.288 1.031“

 

Source: Calculated from IGMAW survey.
Note: The significance level of CAPE could not be calculated.
See full results in Appendix C.

76

4.2.1.4 Empirical Results of GMA Effect on Self Employment

The next analysis examines the GMA effect on self-employment in a similar way. The
results of the Cragg estimation are presented in Table 14.The negative Sign of the probit
marginal effects for age suggests that older headed households are less likely to be self-
employed. Interestingly though, the CAPE shows that those that are self-employed tend
to have higher returns to labor than younger households, possibly due to the effect of
their accumulated experience. The UAPE has a negative sign, indicating that the negative
effect on the probability of being self-employed outweighs the positive effect on the level
of earnings. Education has the opposite impact here compared to its impact on wage
earnings; households with more educated members are less likely to be self-employed, as
perhaps they tend to be hired in the wage labor market. Education also has a negative
impact on the level of earnings as indicated by the sign and significance of the UAPE and
the CAPE. The number of children and adults are positively associated with self-
employment; the extra labor force can free time for the household head to attend to their
own business, and may also push the head into starting such businesses. These variables
also have a positive effect on the level of earnings (both CAPE and UAPE are positive
and Significant). Hectares of cropped area are negatively associated with the probability
of self-employment, arguably for the same reasons as for wage employment. The
coefficient of the value of consumer durable assets is positive, which follows the notion

that better off households are more likely to participate in off-farm activities.

77

Table 14. Cragg estimation results: determinants of the probability and total level of

self-employment earnings

 

Variable description

Tlerl Sig.

 

 

Intercept

Age of household head in years

Sex of household head (=1 if male)
Max. education hh member (years)
Number of children (<1 5 years)
Number of female adults (15-60 years)
Number of male adults (15-60 years)
Distance to nearest road (in km)
Cropped area (in hectares)

Value of consumption assets (ZMK)
Value of production assets (ZMK)
Population density (in sq km)
Infrastructure (#)

Tourist lodge in the community (=1)
Prime GMA (=1)

Secondary or specialized GMAEI)
Number of observations: 2264

Log likelihood: -5144.9
Source: Calculated from survey data.

0.031
-0.010 ***
0.137 *
-0.032 **
0.029 *
0.101 *
0.066 *

0.003
-0.106 **"'
0.006 *

0.00001
0.0003

0.045 ***
0.057
0.174 *
-0.132 *

11.380 ***
0.056 ***
0.843

-0.l83 *
0.089
1.781 "'*"‘
1.241 ***

-0.020
0.019
0.016

—0.002
0.007 **

-0.072

-0.494
0.084
0.330

Marginal Effects
Tler2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
n/a n/a n/a
-0.004 *** 0.052 *** -0.037 **
0.055 * 0.787 1.267 *
-0.013 ** -0.l71 * -0.287 **"‘
0.012 "‘ 0.083 0.231 *
0.040 * 1.663 ** 1.456 "*
0.026 * 1.158 ** 0.984 ***
0.001 -0.019 * 0.110
-0.042 *" 0.018 -0.675 *"
0.002 * 0.015 0.043 *
0.000 -0.002 0.035
0.000 0.0070 0.005
0.018 *** -0.067 0.261 ***
0.023 -0.461 0.146
0.069 * 0.078 1.164 *
-0.053 * 0.308 -0.707

*** 1% significance level, "5% significance level, *10% significance level

n/a: not applicable

Higher levels of infrastructure impacts the probability that a household is self-employed,

though as in the case of wage earnings, it has a negative (though insignificant) effect on

the level of earnings for the sub-sample of self-employed households (CAPE). This might

be the effect of excess labor supply in areas where services and infrastructure attract job

seekers.

Finally, prime GMAs have a significant and positive impact on the probability of being

self-employed, though both the magnitude and the significance are lower than for wage

earnings. As advanced in the introduction of Chapter 4, one of the benefits of living in

GMAs is the generation of opportunities for micro-enterprises as a result of the tourism

78

sector; for example, the increasing population and economic activity in Lupande GMA
creates opportunities for small business (selling mobile phone “talk time”, owning
restaurants, providing accommodation for local workers). In secondary and specialized
GMAS, the GMA effect on the probability of being self employed is negative and
significant, which does not have an intuitive explanation. Additional data is needed to

formulate a sound hypothesis for this result.

An analysis of the GMA effect on self-employment earnings by welfare level of the
household is presented in Table 15. Poorer households living in prime GMAS seem to
enjoy higher chances to earn income from small businesses and obtain higher levels of
income than those in non-GMAs. However, for those that live in secondary or specialized
GMAs the effect is the opposite. For the two upper quintiles the GMA effect is

insignificant in probability and level.

Table 15. GMA effect on self-employment earnings by welfare levels

 

GMA type/wealth quintile Tier 1 Tier 2 Probit mfx CAPE UAPE
Prime GMAs
lowest 2 wealth quintiles 0.360” -1.167 0.143** - 1.107 1.680*

 

 

Top 2 wealth quintiles -0.151 1.435 -0.060 1.310 -0.352
Secondary GMAs

Lowest 2 wealth quintiles -0.223* -1.446 -0.087* -1.371 -1.953*

Top 2 wealth quintiles -0.106 0.728 -0.042 0.665 -0.379

 

Source: Calculated from IGMAW survey data.
Note: The significance level of CAPE could not be calculated.
See full results in Appendix C.

79

4.2.2 GMA Effect on Crop Agriculture

We analyze the GMA effect on crop agriculture with the same Cragg model and same
explanatory variables as were used for off-farm earnings. The dependent variables for
this model are the probability of reporting income from crop agriculture and the total

value of crop agriculture. The results are presented in Table 16.

Table 16. Cragg estimation results: determinants of the probability and total level of
crop agriculture

 

 

 

Marginal Effects
Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept 0.753 *** 20.101 *"'* n/a n/a n/a
Age of household head in years 0.001 -0.025 0.0001 -0.02 -0.02
Sex of household head (=1 if male) -0.028 -1.483 * -0.001 -1.172 -1.15
Maximum education hh member (years) -0.030 ** 0.041 -0.002 * 0.033 -0.07
Number of children (<1 5 years) 0.033 0.434 * 0.002 0.343 0.421
Number of female adults (15-60 years) -0.070 0.359 -0.004 0.284 0.022
Number of male adults (15-60 years) 0.015 0.126 0.001 0.099 0.138
Distance to nearest road (in km) -0.005 * 0.005 -0.0002 * 0.004 -0.01
Cropped area (in hectares) 1.440 *** 3.163 *** 0.076 "'** 2.5 7.054
Value of consumption assets (ZMK) -0.001 0.079 * -0.00003 0.062 0.054
Value of production assets (in ZMK) 0.007 ** 0.105 *** 0.0004 "‘ 0.083 0.073
Population density (in sq km) -0.001 *** -0.024 *** -0.0001 *** -0.019 -0.02
Infrastructure (#) 0.009 -0.351 * 0.0005 -0.278 -0.22
Tourist lodge in the community (=1) -0.357 * 0.974 -0.026 * 0.77 -0.49
Prime GMA (=1) 0.016 -0.382 *** 0.001 -0.302 -0.22
Secondary or specialized GMA (=1) -0.328 *** -8.115 *** -0.021 *** -6.415 -6.88

 

Number of observations: 2264

Log likelihood: -8,517.6

Source: Calculated from survey data.

*10% significance level, "5% significance level, *** 1% significance level

n/a: not applicable

Note: the bootstrapping of the CAPE and UAPE could not be calculated using the available programming

 

The results of the probit show that the probability of crop agriculture is negatively
associated with the level of education, reflecting the possibility that better educated

households have employment opportunities outside the farming sector. Distance to roads

80

has a negative impact on the probability of farming, possibly a reflection of the
disincentive that distance to markets can have on the decision to engage in farming
activities. Population density and the existence of a tourist lodge in the community are
negatively associated with the probability of farming, both variables being linked to the
existence of off-farm employment opportunities. Finally, the GMA effect on the
probability of having crop agriculture is positive though insignificant for prime GMAS,

and negative and significant in secondary or specialized GMAS.

With regard to the GMA effect on the value of crops, though the significance levels for
the UAPE and CAPE could not be computed using the available bootstrapping program,
it is notable that all the coefficients are negative, both for prime (GMAl) and secondary
or specialized GMAS (GMA2). This could be the result of a number of influences,
including the effect that wildlife damaged crops can have on the value of crops, the effect
of encroachment due to in-migration (which may put pressure on access to land) or the

availability of off-farm employment (which may reduce the time’dedicated to farming).

An analysis of the GMA effect on crop agriculture by wealth level of the household is
presented in Table 17. It should be noted again, the significance levels for the CAPES
could not be computed using the available bootstrapping program. Still, the results
suggest that GMA2 could have a negative effect on the probability of engaging in
cropping agriculture and the income of those who do engage, both for the lowest two

quintiles and the upper two quintiles. This can perhaps be explained by the fact that

81

secondary or specialized GMAS may generate lower revenues from hunting than prime

GMAS to be invested in hiring of scouts that can help protect the crops.

This result might also explain why all households living in secondary or specialized
GMAS, and those from the poorest segment of prime GMAS, do not earn higher overall
incomes by being in the GMA (see Model 1). For households living in secondary or
specialized GMAs, the gains from access to wage earnings appear to be offset by
negative effects on business income and (though these are not statistically significant)
crop income. For the poorest households in prime GMAs, their higher earnings off the

farm appear to be more than offset by lower earnings in crop income.

Table 17. GMA effect on crop agriculture by welfare level

 

 

 

GMA type/wealth quintile Tier 1 Tier 2 Probit mfit CAPE UAPE
Prime GMAs
Lowest 2 wealth quintiles -0.001 -1.027 -0.0001 -0.781 -0.702***
Top 2 wealth quintiles -0.060 -0.661 -0.0004 -0.557 -0.673
Secondary GMAS
Lowest 2 wealth quintiles -0.399** -5.661* -0.047** -4.304 -5.189
Top 2 wealth quintiles -0.392* -9.113*** -0.003 -7.683 -8.125

 

Source: Calculated from IGMAW survey.
Note: The significance level of CAPE could not be calculated.
See full results in Appendix C.

4.2.3 Impact of GMA policies on crop damage

Human-wildlife conflict represents one of the biggest challenges for communities living
in GMAs. As mentioned in Chapters 1 and 2, farmers are routinely affected by crop
destruction, mainly by elephants which have proven to be extremely difficult to control.

Despite efforts from NGOS and ZAWA to help communities with electric fences and

82

other strategies to keep elephants away from crops (e. g., chili fences, beating of drums),
the problem continues to be serious in areas where significant elephant populations are
found. In addition, the combined efforts from ZAWA, hunting and photographic safari
outfitters and CRBs to control poaching in popular tourist areas may exacerbate the
problem for households whose prime economic activity is farming. The hypothesis is that
households living in well stocked GMAs are more likely to experience crop damage and
report higher values of crop loss than households living in lower stocked GMAs and in g.

non-GMAs.

 

4. 2. 3. 1 Model estimation

Since only 14% of respondents reported crop loss, the analysis uses a Cragg model to
examine the GMA effecton both the probability of sustaining crop damage and the value

of crop loss. The model is specified analogously to the case for non-farm employment:

P(CDi =11Xi) =rXi + #i (Tier 1) (15)
Ln Yi: a+flXj +67 (Tier 2) (16)
Where CD is the crop damage variable which takes the value 1 if the household

reported crop loss

Yi is the total value of crop loss

Xi is the vector of household and community characteristics, assumed to
be the same for equations (1) and (2)

y is the vector of coefficients associated with Xi in the probability
equation

)6 is the vector of coefficients associated with Xi in the continuous
equation

or is the intercept in the continuous equation

83

The selected variables for the analysis are shown in Table 18. The model includes four
variables not included in previous models: the percentage of households that reported
crop damage, the value of crop damage (these first two are the dependent variables), the
number of scouts hired in the community, and the total value of harvest, which is
inclusive of the reported value of harvest and the reported value of crop lost due to

wildlife, to control for the effect that the total value of harvest will have on crop damage.

Table'18. Comparison between variable means for the full sample and the GMA and
non-GMA subsamples

 

 

 

Subsamples

Variable description Full sample GMAs Non-GMAs Sig."
Number of households 2,717 1,574 1,143
Number of households that reported crop damage 369 251 110

Value of crop damage 28,423 30,079 26,140

Age of household head in years 42.46 41.00 44.48 ***
Sex of household head (=1 if male) 0.74 0.73 0.76 **
Household size 5.28 5.08 5.57 ***
Distance to nearest all-weather road (in km) 5.09 6.08 3.80 ***
Cropped area (in hectares) 0.92 0.93 0.92

Value of consumption durable assets (ZMK) 401,588 285,362 561,641 **
Value of production durable assets (ZMK) 618,036 256,729 1,115,584 ***
P0pulation density (in sq km) 35.2 41.4 26.9 ***
Infrastructure (#) 3.62 3.64 3.59
Number of scouts hired in the community (#) 1.00 1.56 0.24

Total value of harvest (ZMK) 120,884 628,249 681,214

Prime GMA (=1) 0.17 0.30 - n/a”
Secondary or specialized GMA (=1) 0.20 0.35 - n/a

 

Source: Calculated from IGMAW survey data
*10% significance level, "5% significance level, *** 1% significance level

The means for the GMA and non-GMA subsamples indicate that on average, more
households report crop damage in GMAS compared to non-GMAS (16% vs. 9.6%) as

expected. However, while the average value of crop loss is also higher in GMAs, this

84

difference is not statistically Significant. The average number of scouts hired in the

community is also higher within GMAs (1.6) compared to non-GMAS (0.2).

The effect of the number of scouts on the probability and value of crop damage is
difficult to predict a priori. On one hand, more scouts could more effectively control
poaching, thereby exacerbating the human-animal conflict by maintaining higher animal
populations. On the other, scouts can help communities scare away offending elephants,
or in rare occasions even kill them with the permission of ZAWA. The second additional
variableincluded in this analysis is the total value of harvested plus lost crop, which
controls for the impact that higher value crops have in the potential total value of crop

loss.

It is suspected that the variable number of scouts may be endogenous to the crop damage
regression, since higher stocks of wildlife (and thus higher crop damage) are likely to
lead to the hiring of more scouts, using revenues from hunting through the CBNRM
programs, especially in GMAs where wildlife is abundant and revenues are sufficient to
support adequate staff. This possibility was tested using stata’s ivtobit command and the
Hausman test for endogenous regressors (Hausman, 1978). Instruments for number of
scouts were a dummy variable indicating whether the VAG had received funds from
ZAWA, and three park system dummies. All four instruments were statistically
significant above 0.01, and the Hausman test generated a Wald p-value of 0.33, which

leads to failure to reject the null hypothesis of exogeneity. Therefore, we conclude that

85

the number of scouts is not an endogenous regressor in the two-stage model of crop

damage associated with wildlife conflicts (full results in Appendix D).

4. 2. 3.2 Empirical estimation and results

Household size has a negative impact on the probability of crop 1055, indicating that
access to extra manpower can help contain wildlife and protect the fields. The total value
of loss can also be reduced, as indicated by the signs of the CAPE and UAPE. The
distance to all weather roads is positively associated with the probability of crop damage,
suggesting that, logically, more remote places are likely to have greater wildlife
populations. Cropped area and total value of the harvest are control variables to account
for the effect that larger areas under cultivation and higher value crops (or higher yields
per unit area) will have in the probability and total value of crop 1055. As expected, both
are positively associated with the probability of crop damage, though interestingly, for
those households that suffered crop damage, the cultivated area is negatively associated

with the total value of loss (CAPE).

86

Table 19. Cragg results: GMA effect on probability and total value of crop loss

 

Marginal Effects

 

 

Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept -2.324 **"‘ 5.587 n/a n/a n/a
Age of household head in years -0.002 0.013 -0.0002 0.002 -0.001
Sex of household head (=1 if male) -0.041 0.249 -0.006 0.032 -0.046
Household size (#) -0.039 * -0.020 -0.006 * -0.003 -0.076 *
Distance to nearest road (km) 0.006 * -0.014 0.001 * -0.002 0.009 *
Cropped area (in hectares) 0.068 * -0.003 * 0.010 * -0.051 * 0.077
Value of consumption assets (ZMK) -0.002 -0.015 -0.0003 0.000 -0.005
Value of production assets (ZMK) -0.004 * -0.394 -0.001 * -0.002 -0.009 *
Population density (in sq km) 0.000 -0.001 -0.00002 0.000 0.000
Types of existing Infrastructure (#) -0.006 -0.014 -0.001 -0.002 -0.013
Number of scouts (#) 0.025 0.047 0.004 0.006 0.053 *
Total value of harvest (in ZMK) 0.041 **"' 0.189 *** 0.006 *** 0.024 *** 0.102 ***
Prime GMA (=1) 0.780 *** 0.080 0.161 *** 0.010 1.486 ***
Secondary or specialized GMA (=1) 0.643 *** 0.172 0.122 *** 0.022 1.238 ***

 

Number of observations: 2266

Log likelihood: -1513.7
Source: Calculated from IGMAW survey data
*10% significance level, M5% significance level, ***1% significance level

The number of scouts hired in the community has a significant and positive effect on both
the probability and the value of crop damage. This finding could suggest effective anti-
poaching patrol which may help to increase wildlife population; but it could also reflect
the likely endogeneity of this variable, since higher stocks of wildlife are likely to lead to
the hiring of more scouts, using revenues from hunting through the CBNRM programs,
especially in GMAs where wildlife is abundant and revenues are sufficient to support

adequate staff.

Finally, the GMA effect on the probability of crop loss is, as expected, positive and

significant, more so in prime GMAs than in secondary or specialized. The results clearly

confirm the hypothesis that households are more likely to be affected by crop loss in

87

 

better stocked GMAs. As mentioned before, the human-animal conflict represents one the

biggest threats for the success of CBNRM programs.

An analysis of the GMA effect on crop damage by wealth level of the household is
presented in Table 20. Households seem to be equally affected by crop damage regardless
the welfare level. There are slight differences in the probability and level of crop damage

but in general the effect is significant and positive for all households.

Table 20. GMA effect on crop damage by welfare level

 

 

 

GMA type/wealth quintile Tier 1 Tier 2 Probit mix CAPE UAPE
Prime GMAs
Lowest 2 wealth quintiles 0.922*** -1.246 0.185*** -1.185 1.388***
Top 2 wealth quintiles 0.635*** 1.080 0.132*** 0.951 1.473***
Secondary GMAs
Lowest 2 wealth quintiles 0.544*** -0.864 0.087*** -0.821 0.804M
Top 2 wealth quintiles 0.712*** 2.072* 0.156*"‘* ' 1.826 1.794***

 

Source: Calculated from IGMAW survey data.
Note: Significance test for CAPE could not be calculated.
See full results in Appendix C.

The table below presents a summary of the results of the Cragg model by welfare levels
(based on the UAPES). The GMA effect on total household income is positive for the
non-poor households. When disaggregating by source of income however we see that
poorer households living in prime GMAS do enjoy higher levels of wage and self
employment income compared to those living outside GMAs. This effect might be offset
by the negative impact on crop harvest, which neutralizes the overall effect on total
income. The GMA effect on crop loss is clearly significant (and positive) across all types

of GMAS and all levels of welfare.

88

Table 21. Summary of Cragg results on sources of income by welfare level.

 

 

 

 

Type of income Prime GMA Secondary/Specialized GMA
Overall Non-poor Poor Overall Non-poor Poor
Total + sig + sig insig imig insig insig
Wage + sig + sig + sig + sig + sig + sig
Self employment + sig insig + sig imig insig - sig
Crop Inrvest - ? insig - sig - ? insig insig
Crop 1055 + sig + sig + Sig + sig + sig + sig

 

 

 

 

Source: Calculated from IGMAW survey data.

89

5. SUMMARY OF RESULTS AND CONCLUSIONS

The main goals of this study were to estimate the effect of living in a GMA on household
income and to examine the avenues through which any effect might be generated. These
questions are evaluated using econometric techniques that seek to isolate the variables of
interest. Three research questions are considered. The first question uses OLS to explore
the relationship between GMAS and total household income. The second research
question uses a Cragg double-hurdle model to evaluate the effect of GMAs on household
participation in and earnings from cropping agriculture, wage employment, and self
employment). Finally, the third model, which also uses the Cragg method, seeks to

identify the extent to which households living in GMAs are more prone to crop damage.

The results from the first model suggest that prime GMAs positively affect household
income, while secondary and specialized GMAs have no effect. This result reinforces the
findings in Bandyopadhay and Tembo (2009) who also found a positive association
between GMAs and welfare, measured by expenditure. Another result in common with
their study is that only households in the upper two quintiles are found to benefit from
living in GMAS. However, there is a major difference with respect to the magnitude of
the GMA effect. While they find an overall GMA effect (without distinguishing between
stocking levels of GMAs) on household expenditures of 66%, the prime GMA effect on
household income found in Model 1 in this study is 17%, and insignificant for secondary
or specialized GMAS. Another difference emerges when considering the effect of park
systems. Bandyopadhay and Tembo (2009) find a very large GMA effect in Bangweulu

(73%) and Luangwa (74%) while the effect for Kafue and Lower Zambezi is negative

90

and insignificant. In contrast, this study finds a positive but lower GMA effect in
Luangwa (18%) and a positive GMA effect in Kafue (13.5%). The latter results seem
sensible, considering the distribution of prime GMAs by park (higher impacts are found

where more prime GMAs were sampled).

The results in Model 1 also provide new findings that go beyond those in Bandyopadhay
and Tembo (2009). Model 1 supports the hypothesis that the level and variety of wildlife
positively influences the income levels of households. This is reflected by the sign and
magnitude of the coefficients for the variables representing prime and secondary or
specialized GMAs. Specifically, households living in prime GMAS enjoy 17% higher
income compared to non-GMA areas. The effect for households living in secondary or
specialized GMAs is positive, though insignificant, suggesting that areas with lower
wildlife p0pulations have less capacity to generate income. Additionally, the first model
highlights the uneven distribution of the potential benefits to living in a prime GMA,
which is a common finding in the literature of poverty reduction and rural non-farm

income.

Model 2, which investigates the GMA effect on non-farm income and on crop damage,
represents an additional new contribution of this study. These results suggest that GMAS
have a positive effect on the probability and the value of non-farm earnings. Households
living in prime GMAS have a 7.8% higher chance of being employed in the wage labor
market than households living in non-GMAs. Similarly, households living in secondary

or specialized GMAs have a 7.4% higher chance of participating in the wage labor

91

market. The overall effect (UAPEs) of both types of GMAS on the expected wage
earnings is positive, though the effect of prime GMAS on the level of wage earnings for
the subsample of households that reported a wage income (CAPE) is negative. This
seemingly contradictory result suggests that although GMAs provide more opportunities
for households to engage in wage labor, they may also attract job seekers which in turn
drive wages down. Another interesting finding is the positive effect of the presence of a
tourist lodge in the area on the probability of participating in non-farm income and on the
level of wage earnings, confirming the importance of the tourism sector for the

generation of non-farm employment opportunities.

Results are similar regarding the GMA effect on income from self-employment.
Households living in prime GMAS are 6.9% more likely to be self-employed than those
living outside GMA areas. The UAPE indicates that households in prime GMAS are
expected to earn 116% more than those outside GMAs. The results for secondary and
specialized GMAs however do not show an effect on self-employment. The GMA effect
on probability of self-employment is negative and significant but the CAPE and UAPE

are insignificant.

The results from the GMA effect on crop damage clearly support the hypothesis that
households living in areas with higher wildlife populations suffer more intensely from
crop destruction. This reaffirms the statements from people interviewed in two GMAs
near the Lower Zambezi and South Luangwa national parks who named the human-

elephant conflict the most challenging problem facing households living in a GMA.

92

5.1 Conclusions

Results in this thesis confirm the potential of GMA policies to generate increased income
for households living within their boundaries, and makes a case for the continuation of
these programs as a strategy to combat poverty. Additionally, by approaching this
question from a number of different angles, this study highlights several crucial details

that are relevant for future policy and programmatic design:

First, the GMA effect on total household income is only found in prime GMAs. The level
and diversity of wildlife stocks are clearly linked to the potential of these areas to
generate benefits for the community. While secondary or specialized GMAs have a
positive effect on the probability and level of wage earnings, the effect is insignificant for
the probability and level of self-employment. Investing in wildlife restocking and
protection in less animal populated areas could maximize the benefits from CBNRM

programs.

Second, the results of the analysis of the GMA effect on wage earnings suggest that in-
migration to prime GMAs may put negative pressure on wages. This may mean that even
if households enjoy higher chances of being employed, their wages may be lower than
households living in non-GMAS. The expansion of the tourism industry to other areas
outside the main tourist circuits could further broaden the impact of GMA policies on
welfare and avoid the problems associated with high levels of in-migration influx. This
should be accompanied by marketing campaigns to attract tourists to other areas, and by

enhancing existing infrastructure (roads, communications, electricity, etc.).

93

Finally, notwithstanding the positive link between well stocked GMAS and income, the
results also indicate a higher prevalence of crop damage in prime GMAS as opposed to
non-GMAs. This threat is further reinforced by the results of the GMA effect on crop
agriculture, which indicate a negative association between GMAS, both prime and
secondary or specialized, and the probability of participation in and earnings from crop
agriculture. While this effect might not only be the result of crop damage by wildlife,
crop damage could be one of the contributors. Hence, if the attacks proliferate unabated,
there could be a point at which the GMA effect changes sign. Policies should not only
focus on wildlife conservation but also on mechanisms to protect or compensate farmers
for the losses. This is of particular importance if the goal is to change community
attitudes towards wildlife. If households perceive wildlife as a destructive element for
their livelihoods rather than a source of income, conservation efforts will continue to face
resistance, which will ultimately threaten the sustainability of wildlife populations and

the program as a whole.

94

6. AREAS FOR FURTHER RESEARCH

Evidence that GMA policies positively affect household income is a promising outcome
which encourages the continuation of CBNRM programs. Additional research based on
the IGMAW survey, complemented with other information, could help refine the current

GMA policies and focus on elements that work best in the interest of the communities.

One area of further research should differentiate between consumptive and non-
consumptive tourism in estimating the impact of tourism on total household income.
While on one hand hunting safaris generate large revenues from relatively few hunting
licenses and contribute to the community by sharing game meat, non-consumptive
tourism can attract a wider range and greater number of tourists who pay park entry fees
and book guided game viewing trips. Knowing the impact of each type of investment on
the welfare of the communities could help maximize the returns to land. For this,
additional information beyond the data provided by the IGMAW survey is needed,
principally, the number of tourist and safari lodges by GMA and the capacity of each
tourist lodge and safari camp. Other useful information would be the revenues obtained
by hunting licenses for each GMA and the share of these revenues delivered to CRBS.
Also knowing the number of locally hired employees and the direct pledges that lodges
and safari camp sites make to the community would be beneficial. ZAWA is clearly the

institution best placed to generate this type of information.

One issue that clearly threatens the success of the GMA policies towards wealth creation

is cr0p losses associated with wildlife protection. Overall, the findings show that

95

households living in GMAs obtain higher incomes compared to those in non-GMA
designated areas, despite these losses. However, there might be a threshold beyond which
crop losses could reverse the positive GMA effect, making agriculturally dependent
communities even poorer. Finding this threshold would be key to determine the risks
associated with the CBNRM programs. Successful GMA policies that increase wildlife
populations to a point where they are incompatible with community livelihoods could
eventually cause more harm than benefit. Further research could consider a model that
tests different scenarios, for example, analyzing the outcome of the GMA effect in case a
significantly larger number of farmers are affected by crop loss, or by increasing the
average value of crop loss. It would be particularly interesting to test this in prime GMAs
where the number of reported incidents is higher, looking at the households that reported
the highest level of crop damage to have an indication of what the worst case scenario

could be.

Linked to the issue of crop damage, additional research could look into the capacity of
scouts to protect wildlife while simultaneously protecting farmers from crop damage. The
findings indicate that there is a positive relationship between the number of scouts hired
in the community and the probability of crop loss in GMAS, which could indicate on one
hand success in protecting wildlife, hence the proliferation of incidents, but could also
indicate that scouts are not able to contain wildlife and prevent them from destroying the
fields. A review of the scouts’ mandate could help clarify the role they are given in terms
of community protection. If this is only focused on wildlife protection then there might

be a need by ZAWA to review the mandate to reflect the need to protect crops as well as

96

wildlife, and thus minimize conflict. In any case, it would be worth knowing whether
hiring larger numbers of scouts would make a difference in protecting farmers or would
instead make their situation worse, as the scouts may be more successful in their anti-

poaching activities than in their efforts to control offending animals.

97

APPENDICES

98

 

APPENDIX A

ANALYSIS OF THE MIGRATION VARIABLE

In Model 1 of this study, a dummy variable that indicates migration to and from GMAS
was generated based on the following question from the IGMAW survey: where was
your household residing 5 years ago? The dummy takes the value of one if the household
was residing in a different locality (within the same district, in a different district within

the same province, in a different province or in a different country), and zero otherwise.

Overall, 11.6% of households surveyed migrated in the past five years. 12.8% of
households in GMAs had migrated there within the past five years and 9.9% of those
outside GMAs had migrated within the past five years (Table A1). In both cases, the
magnitude of the percentage is similar and not large enough to create concern about
sample selection bias. There seems to be slightly higher movement to GMAS compared to
non-GMAs, which can also be observed when analyzing the migration variable by park
systems and GMAS (Table A2). In all park systems with the exception of Bangweulu, the
percentage of households in GMAS that had migrated over the past five years is higher
than the percentage of households in non-GMAS that had migrated during the past five
years. South Luangwa and Kafue attracted the most migrants, which could be related to
the development of the tourism industry in those areas (South Luangwa and Kafue were

the two parks that hosted the most tourists in 2006).

99

Table A 1. Mean comparison of the migration variable

 

Total Sample GMA non-GMA
Number of observations 2717 1574 1143

Mean 1 1.60% 9.90%
Source: Calculated from IGMAW survey data

 

Table A 2. Mean comparison of the migration variable by park system

 

Park Systems Total Sample GMA non-GMA

Bangweulu 7.9% 7.6% 8.5%
Kafue 13.3% 15.5% 11.6%
Lower Zambezi 10.6% 12.9% 8.4%
South Luangwa 14.7% 17.0% 10.1%

 

Source: Calculated from IGMAW survey data

Another way to analyze whether migration could lead to endogeneity is to test whether
there are structural differences between the subsamples of households that migrated vs.
those that did not. The Chow test examines whether parameters (slopes and intercepts) of
one group (households that migrated) are different from those of other groups
(households that did not migrate) (Chow, 1960). If the parameters are equal for both
groups, the OLS results are less likely to be biased. For the test, the unrestricted model is
built by including in Model 1 the migration dummy plus all the right hand side variables
interacted with the migration dummy. The null hypothesis, H0, is that the two groups

have equal parameters.

The F statistic is constructed using the formula: F=[(RU2 — RR2)/q] / [(1- RU2)/(n-k-1)]

Where: RU2 is the goodness of fit for the unrestricted model=0.22
RR2 is the goodness of fit of the restricted model (Model 1)=0.21
q =d.f. restricted model- d.f. unrestricted model=16
n is the total number of observations=2264
k is the number of parameters of the unrestricted model=31

100

 

The F statistic is 1.99. The critical value (c) for an F with d.f.=16, 2233 = 2.46 (at 99%
confidence level). Since F < c, the null hypothesis cannot be rejected, therefore the
parameters are assumed to be equal for both groups. We can conclude that households

that migrated are statistically identical to those that did not migrate.

101

 

APPENDIX B

OLS RESULTS BY WELFARE LEVEL AND PARK SYSTEM

Table B 1. Results of the OLS by welfare level

 

 

 

Variable descrijx ion Coefficient Significance Standard
level error
lmercept 13.120 *** (0.12)
Age of household head in years -0.003 (0.00)
Sex of horsehold head (=1 if male) 0.066 * (0.06)
Maximum education of hh member (in years) 0.042 (0.01)
Number of chilcren (<1 5 years) 0.018 (0.01)
Number of female adults (15-60 years) 0.113 (0.03)
Number of male adults (15-60 years) 0.070 *** (0.03)
Distance to nearest al l-weather road (in km) -0.005 (0.00)
Cropped area (in hectares) 0.038 (0.02)
Value of consumption dtrable assets (in ZMK) 0.020 (0.00)
Value of production durable assets (in ZMK) 0.010 *** (0.00)
Population dersity (in sq km) 0.001 (0.00)
Types of existing Infrastructure (#) 0.033 "”" (0.01)
Tourist lodge in the commnity (=1) 0.198 (0.10)
Households in lowesttwo quintiles and living in GMA] 0.033 *** (0.10)
Horseholtt in lowest three quintiles and living in GMA] 0.031 (0.05)
Householth in upper two quintiles and living in GVIAl 0.046 "‘ (0.02)
Horseholtb in lowesttwo quintiles and living in GMA2 -0.059 (0.07)
Horseholds in lowest three quintiles and living in GMA2 0.040 * *"‘ (0.04)
Householch in upper two quintiles and living in GMA2 -0.008 (0.02)
Dependant variable: Total horsehold income (in logarithm)
Observatiors 2264
R-squared 0.212

 

Source: Calculated from IGMAW survey data

102

Table B 2. Results of the OLS by park system

 

 

 

Variable descrirx ion Coefficient Si gnificance Standard
level error
Intercept 13.090 W (0.12)
Age of household head in years 0.003 * (0.00)
Sex of household head (=1 if male) 0.066 (0.06)
Nbximurn education of hh member (in years) 0.042 *** (0.01)
Number of children (<15 years) 0.020 (0.01)
Number of female adults (15-60 years) 0.114 *** (0.03)
Number ofmale adults (15-60 years) 0.073 ** (0.03)
D'stance to nearest all-weather road (1n km) 0005 *** (0.00)
Cropped area (in hectares) 0.039 * (0.02)
Value of consumption durable assets (in ZVIK) 0.020 *** (0.00)
Value of production durable assets (in ZMK) 0.010 *** (0.00)
Population density (in sq km) 0.001 *** (0.00)
Types of existing Infiastructure (#) 0.(B1 *** (0.01)
Tourist lodg inthe community (=1) 0.199 * (0.10)
I-buseholds living in GMA within Luangwa Park system 0.177 ** (0.08)
Households living in GMA within Kafue Park system 0.135 * (0.08)
I-buseholds living in GMA within Zambezi Park system 0.046 (0.08)
Dependant variable: Total household incorre (in logarithm)
Observatiors 2264
R-squaed 0.212

 

Source: Calculated from IGMAW survey data.

103

APPENDIX C

CRAGG RESULTS BY WELFARE LEVEL

Table C 1. Results of Cragg model on wage employment by welfare level (lower two

quintiles)

 

Variable description

Intercept

Age of household head in years

Sex of household head (=1 if male)
Maximum education hh member (years)
Number of children (<15 years)
Number of female adults (15-60 years)
Number of male adults (15-60 years)
Distance to nearest road (km)
Cropped area (in hectares)

Value of consumption assets (ZMK)
Value of production assets (ZMK)
Population density (in sq km)
Infrastructure (#)

Tourist lodge in the community (=1)
Prime GMA (=1)

Secondary or specialized GMA (=1)

Tier 1 Sig. Tier 2 Sig.
16.820 ***
-0.l39 *
-4.996 "'

-l.980 **"'
-0.010 "'

0.181

0.078 ***

0.005

-0.400 **

0.118
-0.004
0.025
0.014
-0.012

0.001 "'
0.048 *

0.039

0.618 **"‘
0.655 ***

0.183
-0.469
-0.812

3.738 ***

-0.052
-0.416
0.324

0.245 *

-0.001

-0.665 "”"

-l .121
0.380
2.055

Marginal Effects

 

Probit Sig. CAPE Sig. UAPE Sig.

 

n/a *** n/a n/a
-0.001 * -0.125 -0.019
0.010 -4.467 0.014
0.005 *** 0.164 0.104
0.000 -0.420 -0.013
-0.024 ** -0.726 -0.531
0.007 3.343 0.304
0.000 -0.047 -0.007
0.001 -0.372 0.014
0.001 0.289 0.031
-0.001 0.219 0.028
0.000 * -0.001 0.002
0.003 * -0.595 0.031
0.002 -1.003 0.001
0.057 *** 0.340 0.784 **
0.057 *** 1.838 0.900 **

 

Number of observations: 918
Log likelihood: -274.9l
Source: Calculated from IGMAW survey data

Note: Significant levels of CAPE and UAPE only computed for GMA variables

104

Table C 2. Results of Cragg model on wage employment by welfare level (upper two

quintiles)

 

Variable description

Tier 1 Sig. Tier 2 Sig.

Marginal Effects

 

Probit Sig. CAPE Sig. UAPE Sig.

 

Intercept

Age of household head in years

Sex of household head (=1 if male)
Maximum education hh member (years)
Number of children (<15 years)
Number of female adults (15-60 years)
Number of male adults (15-60 years)
Distance to nearest road (km)
Cropped area (in hectares)

Value of consumption assets (ZMK)
Value of production assets (ZMK)
Population density (in sq km)
Infrastructure (#)

Tourist lodge in the community (=1)
Prime GMA (=1)

Secondary or specialized GMA (=1)

-3.900 *** 18.272 ***

-0.004
0.101
0.230 ***
-0.016
0.066
0.025
-0.011 *
-0.004
0.023 *"‘*
-0.011 ***
0.002 ***
0.034
0.602 ***
0.762 ***
0.424 **

0.028
-5.184 * * *
0.1 18
-0.326
-0.036
1.881 ***
-0.001
-0.242
0.010
-0.017
0.001
-0.103
1.076
-1 .406
-0.290

n/a tint
0.001
0.013
0.032 ***
-0.002
0.009
0.003
-0.002 *
0.000
0.003 ***
-0.002 ***
0.000 ***
0.005
0.118 ***
0.144 ***
0.071 **

n/a
0.028
-5.161
0.118
-0.324
—0.036
1.873
-0.001
-0.241
0.010
-0.017
0.001
-0.103
1.071
-1.400
-0.288

n/a
-0.007
-0.473

0.597
-0.087
0.163
0.326
-0.028
-0.043
0.060
0.057
0.006
0.071
1 .671
1.726 ***
1.031 **

 

Number of observations: 1346
Log likelihood: -852.7

 

Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

105

Table C 3. Results of Cragg model on self employment by welfare level (lower two
quintiles)

 

Marginal Effects

 

 

Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept -0.135 10.615 *** n/a n/a n/a
Age of household head in years -0.012 *** 0.041 * -0.005 *** 0.039 -0.053
Sex of household head (=1 if male) 0.136 1.247 * 0.053 1.183 1.342
Maximum education hh member (years) -0.013 -0.173 -0.005 -0.164 -0.149
Number of children (<15 years) 0.052 * -0.244 0.020 "' -0.231 0.211
Number of female adults (15-60 years) 0.113 * 2.232 *** 0.044 * 2.116 1.617
Number of male adults (15-60 years) 0.043 1.990 *** 0.017 1.888 1.095
Distance to nearest road (km) -0.001 -0.014 0.000 -0.014 -0.009
Cropped area (in hectares) -0.160 *** 0.969 ** -0.063 "* 0.919 -0.559
Value of consumption assets (ZMK) -0.010 0.065 -0.004 0.062 0032
Value of production assets (ZMK) 0.006 * 0.009 0.002 * 0.009 -0.056
Population density (in sq km) 0.000 0.011 ** 0.000 0.010 0.004
Infrastructure (#) 0.065 *** -0.022 0.026 *** -0.021 0.381
Tourist lodge in the community (=1) 0.176 -0.562 0.070 -0.533 0.825
Prime GMA (=1) 0.360 ** -1.l67 0.143 *"‘ -1.107 1.680 *
Secondary or specialized GMA (=1) -0.223 * -1.446 -0.087 "‘ -1.37l -l.953 **

 

Number of observations: 918
Loglikelihood: -1901.3
Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

106

Table C 4. Results of Cragg model on self employment by welfare level (upper two

quintiles)

 

Marginal Effects

 

 

Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept 0.004 12.686 n/a n/a n/a
Age of household head in years -0.005 0.075 -0.002 0.069 0.002
Sex of household head (=1 if male) 0.164 -0.096 0.065 -0.088 1.065
Maximum education hh member (years) -0.04l ** -0.289 -0.016 * -0.264 -0.415
Number of children (<15 years) 0.013 0.218 0.005 0.199 0.189
Number of female adults (15-60 years) 0.06] 2.183 0.024 1.993 1.438
Number of male adults (15-60 years) 0.065 0.592 0.026 0.541 0.719
Distance to nearest road (km) 0.005 -0.031 0.002 -0.029 0.020
Cropped area (in hectares) -0.072 ** -0.197 -0.029 ** —0.180 -0.581
Value of consumption assets (ZM K) 0.009 ** -0.008 0.004 ** -0.007 0.058
Value of production assets (ZMK) -0.002 -0.001 -0.001 0.000 0.062
Population density (in sq km) 0.001 ** 0.005 0.000 ** 0.005 0.010
Infrastructure (#) 0.032 * -0.108 0.013 * -0.099 0.010
Tourist lodge in the community (=1) 0.149 0.648 0.059 0.592 1.314
Prime GMA (=1) -0.151 1.435 -0.060 1.310 -0.352

Lea-LI.“—

‘- 15' M .l‘l".""

Secondary or specialized GMA (=1) -0.106 0.728 -0.042 0.665 -0.379
Number of observations: 918
Log likelihood: -2234.1

Source: Calculated from IGMAW survey data

Note: Significant levels of CAPE and UAPE only computed for GMA variables

107

 

Table C 5. Results of Cragg model on crop income by welfare level (lower two

 

 

 

quintiles)
Marginal Effects

Variable description Tier 1 Sig. Tier2 Sig. Probit Sig. CAPE Sig. UAPE Sig.
Intercept 0.574 ** 16.619 *** n/a n/a n/a
Age of household head in years 0.003 -0.039 0.000 -0.030 -0.018
Sex of household head (=1 if male) -0.107 -1.527 -0.010 -1.l61 -1.396
Maximum education hh member (years) -0.022 0.086 -0.002 0.065 -0.015
Number of children (<1 5 years) 0.025 0.589 * 0.002 0.448 0.486
Number of female adults (15-60 years) -0.104 0.292 -0.010 0.222 -0.154
Number of male adults (15-60 years) -0.046 -0.039 -0.004 -0.030 -0.182
Distance to nearest road (km) 0.002 0.035 0.000 0.026 0.030
Cropped area (in hectares) 1.191 *** 3.699 “* 0.114 *** 2.812 6.530
Value of consumption assets (ZMK) 0.023 * -0.049 0.002 "‘ -0.037 0.043
Value of production assets (ZMK) 0.019 ** 0.195 *** 0.002 *“' 0.148 0.209
Population density (in sq km) -0.001 -0.013 0.000 -0.010 -0.011
Infrastructure (#) -0.002 -0.211 -0.0002 -0.161 -0.150
Tourist lodge in the community (=1) -0.359 -0.175 -0.045 -0.133 -1.330
Prime GMA (=1) -0.001 -1.027 -0.0001 -0.781 -0.702
Secondary or specialized GMA E1) -0.399 *"' -5.661 "* -0.047 ** -4.304 -5.189 ***

 

Number of observations: 918
Log likelihood: -3395.8

Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

108

Table C 6. Results of Cragg model on crop income by welfare level (upper two

 

 

 

quintiles).
Marginal Effects
Variable description Tier 1 Sig. Tier 2 Sig. Probit Sig. CAPE Sjg. UAPE Sig.
Intercept 0.719 * 23.693 *** n/a n/a n/a
Age of household head in years 0.001 0.014 0.000 0.012 0.014
Sex of household head (=1 if male) 0.135 -l.292 0.001 -1.089 -0.642
Maximum education hh member (years) -0.007 -0.172 -0.00004 0145 -0.154
Number of children (<15 years) 0.015 0.427 0.0001 0.360 0.373
Number of female adults (15-60 years) -0.046 0.086 -0.0003 0.072 -0.057
Number of male adults (15-60 years) 0.011 -0.646 0.0001 -0.545 0473
Distance to nearest road (km) -0.007 -0.002 -0.00004 -0.001 -0.019
Cropped area (in hectares) 2.061 *** 2.597 *** 0.012 *** 2.189 7.543
Value of consumption assets (ZMK) -0.010 0.078 -0.0001 0.066 0.035
Value of production assets (ZMK) 0.006 0.090 **"‘ 0.00003 0.076 0.044
Population density (in sq km) -0.001 -0.027 *** -0.00001 -0.023 -0.023
Infrastructure (#) 0.014 -0.517 ** 0.0001 -0.436 -0.363
Tourist lodge in the community (=1) -0.447 * 1.837 -0.005 * 1.5486 0.228
Prime GMA (=1) -0.060 -0.661 -0.0004 -0.557 -0.673
Secondary or specialized GMA (=1) -0.392 ** -9.113 *** -0.003 -7.683 -8.125

 

Number of observations: 921
Log likelihood: -3500.l

Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

109

Table C 7. Results of Cragg model on crop damage by welfare level (lower two

quintiles)

 

Variable description

Tier 1 Sig. Tier 2 Sig.

Marginal Effects

 

Probit Sig.CAPE Sig.UAPE Sig.

 

Intercept -2.3l4 *" 6.936 “‘* n/a n/a n/a
Age of household head in years -0.002 0.022 0.000 0.021 0001
Sex of household head (=1 if male) -0.113 0.858 -0.015 0.815 -0.091
Household size (#) -0.027 0.127 -0.003 0.121 -0.030
Distance to nearest road (km) 0.010 ** -0.020 0.001 ** -0.019 0.014
Cropped area (in hectares) 0.044 0.079 0.006 -0.403 0.026
Value of consumption assets (ZM K) -0.010 -O.121 *** -0.001 0.075 -0.008
Value of production assets (ZMK) -0.005 -0.424 -0.001 -0.115 -0.031
Population density (in sq km) 0.00004 0.001 0.00001 0.0005 0.0001
Infrastructure (#) -0.020 -0.150 -0.003 -0.142 -0.050
Number of scouts (#) 0.041 * 0.137 0.005 "‘ 0.131 0.084
Total value of harvest (in ZMK) 0.042 **"‘ 0.204 "‘** 0.005 *** 0.194 0.094
Prime GMA (=1) 0.922 *** -1.246 0.1852 **" -1.185 1.388 ***
Secondary (wecialized GMA (=1) 0.544 *" -0.864 0.087 "* —0.821 0.804 **

 

Number of observations: 919

Log likelihood: -528.4

Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

Table C 8. Results of Cragg model on crop damage by welfare level (upper two
quintiles)

 

Marginal Effects
Tier 1 Sig. Tier 2 Sig. Probit Sig.CAPE Sig. UAPE Sig.

 

Variable description

 

Intercept -2.598 *** 3.587 * n/a n/a n/a
Age of household head in years 0.003 0.028 0.001 0.024 0.011
Sex of household head (=1 if male) 0.037 0.762 0.006 0.671 0.186
Household size (#) -0.029 -0.028 -0.005 -0.025 -0.064
Distance to nearest road (km) 0.000 -0.032 0.000 -0.029 -0.004
Cropped area (in hectares) 0.082 * -0.061 0.013 * -0.371 0.110
Value of consumption assets (ZMK) -0.007 0.000 -0.001 -0.054 -0.023
Value of production assets (ZMK) -0.004 -0.421 -0.001 0.000 -0.014
Population density (in sq km) 0.000 -0.004 0.000 -0.003 0.000
Infrastructure (#) 0.019 0.031 0.003 0.028 0.043

Number of scouts (#) 0.026 0.084 0.004 0.074 0.067
Total value of harvest (in ZMK) 0.042 *** 0.222 *** 0.007 *** 0.196 0.120
Prime GMA (=1) 0.635 *** 1.080 0.1317 *** 0.951 1.473 ***
Secondary or specialized GMA (=1) 0.721 *** 2.072 * 0.156 *** 1.826 1.794 ***
Number of observations: 921
Loglikelihood: -684.8
Source: Calculated from IGMAW survey data
Note: Significant levels of CAPE and UAPE only computed for GMA variables

110

APPENDIX D

ENDOGENEITY TEST, CRAGG MODEL ON CROP DAMAGE

Table D 1. Results of OLS on number of scouts

 

Variable description

Coefficient Significance Standard

 

 

level error

Intercept 13.090 "‘* (0.213)
Age of household head in years 0.032 (0.003)
Sex ofhousehold head (=1 ifmale) 0.019 (0.106)
Household Size -0.002 (0.018)
Distance to nearest all-weather road (in km) -0.007 ** (0.003)
Cropped area (in hectares) -0.026 (0.042)
Value of consumption durable assets (in ZMK) 0.003 (0.004)
Value of production durable assets (in ZMK) -0.003 (0.003)
Population density (in sq km) 0.000 (0.001)
Types of existing Infiastructure (#) 0.058 *** (0.019)
Total value of harvest (in ZMK) 0.006 ** (0.003)
Primary GMA (=1) 1.709 *** (0.163)
Secondary or specialized GMA (=1) 1.110 *** (0.122)
VAG received funds from ZAWA 1.821 *** (0.114)
Luangwa Park system -0.812 *** (0.162)
Zambezi Park system 0401 *** (0.140)
Kafue Park system -0.821 *** (0.126)
Dependant variable: number of scouts

Observations 2266

R-squared 0.25

 

Source: Calculated from IGMAW survey data.

111

Table D 2. Results of the ivtobit on crop damage

 

 

 

Variable description Coefficient Significance Standard
level error
Intercept -2.860 *** (0.263)
Number of scouts -0.028 (0.062)
Age of household head in years -0.002 (0.003)
Sex of household head (=1 if male) -0.045 (0.112)
Household size -0.049 ** (0.020)
Distance to nearest all-weather road (in km) 0.006 "' (0.003)
Cropped area (in hectares) 0.082 ** (0.040)
Value of consumption durable assets (in ZMK) -0.003 (0.004)
Value of production durable assets (in ZMK) -0.005 * (0.003)
Population density (in sq km) 0.000 (0.001)
Types of existing Infrastructure (#) 0.007 (0.025)
Total value of harvest (in ZMK) 0.050 *** (0.004)
Primary GMA (=1) 1.024 *** (0.164)
Secondary or specialized GMA (=1) 0.848 *** (0.142)
Dependant variable: dummy for crop damage
lnstrtunented variable: number of scouts
Observations 2266
Log likelihood -5804.8

 

Wald test of exogeneity (/alpha = 0): chi2(l) = 0.94 Prob > ch12 = 0.3327

P value = 0.333

112

BIBLIOGRAPHY

113

BIBLIOGRAPHY

Agrawal, A. & Gibson, C. (1999). Enchantment and Disenchantment: The Role of
Community in Natural Resource Conservation. World Development 27, (4), 629-
649.

Ahmad, J ., Bergstrom, C., Bryceson, 1., Child, 8., Francis, J ., Khan, P., Ousmane, B.,
Price, T.L., Senaratna, S., Tareen, N. &. van Dam C. (2001). Case Studies in
Sustainable Use. International Union for the Conservation of Nature and Natural
Resources. Available from http\\www.cbd.int.

Ashley, C., Meyer, D., Roe, D., & Goodwin, H. (2004). Defining Pro-Poor Tourism. Pro-
Poor Tourism Partnership. Sheet No.1.

Bandyopadhyay S. and Tembo, G. (2009). Household consumption and natural resource
management around national parks in Zambia Forthcoming in Journal of Natural
Resources Policy Research.

Barrett, C. & Arcese, P. (1995). Are Integrated Conservation-Development Projects
(ICDPS) Sustainable? On the Conservation of Large Mammals in Sub-Saharan
Africa. World Development, 23 (7), 1073-1084.

Barrett, C., Reardon, T. & Webb, P. (2001). Nonfarm Income Diversification and
Household Livelihood Strategies in Rural Africa: Concepts, Dynamics, and
Policy Implications. Food Policy, 26 (4), pp. 315-331.

Beyene, AD. (2008). Determinants of off-farm participation decision of farm households
in Ethiopia. Agrekon, 47 (1), 140-161.

Burke, W.J. (2009). Estimating and Interpreting Cragg’s Tobit Alternative using Stata
Stata Journal (in press).

114

Bwalya, SM. (2003). Understanding Community-Based Wildlife Governance in
Southern Africa: A Case Study from Zambia African Journal of Environmental
Assessment and Management, 7, 41-60.

Central Statistics Office, Zambia. 2006. Living Conditions Monitoring Survey (LCMS).
Available at http:www.zamstats.gov.zm

Child, B. & Dalal-Clayton, B. (2004). Transforming Approaches to CBNRM: Learning
from the Luangwa Experience. Book by McShane, T. O. and M. P. Wells (eds),
Getting Biodiversity Projects to Work: Towards More Effective Conservation and
Development. New York: Columbia University Press.

Chow, GO, (1960). Tests of equality between sets of coefficients in two linear
regressions. Econometrica, 28 (3), 591—605.

C hundama, M., Gumbo, D., Mulonani, D., & Kalyocha, G. 2004. Institutional and Policy
Analysis for Protected Areas under the Reclassification of Zambia ’s Protected
Areas Systems Project. World Wide Fund for Nature Southern Africa Regional
Conservation Programme WWF Zambia Report to the Ministry of Tourism
Environment and Natural Resources.

Cong, R. and D. M. Drukker, (2001). Treatment Effects Model. Stata Technical
Bulleting, 10 (55).

Cragg, J. G. (1971). Some Statistical models for Limited Dependent Variables with
application to the demand for durable goods. Econometrica. 39 (5).

Dalal-Clayton,.B. & Child, B. (2003). Lessons from Luangwa: The story of the Luangwa
Integrated Resource Development Project, Zambia. Wildlife and Development
Series No. 13. London: International Institute for Environment and Development

De Janvry, A. & Sadoulet, E. (2001). Income Strategies Among Rural Households in
Mexico: The role of Off-farm Activities. World Development. 29 (3), 467-480.

Fynn, J ., & Haggblade, S. (2006). Impact of the Kwacha Appreciation and Proposed Tax
Provisions of the 2006 Budget Act on Zambian Agriculture. Working Paper 16.
Lusaka: Food Security Research Project.

115

Gibson C.C. & Marks, S.A.(1995). Transforming Rural Hunters into Conservationists:

An Assessment of Community-Based Wildlife Management Programs in Africa.
World Development, 23 (6), 941-957.

Government of the Republic of Zambia, 2002. Poverty Reduction and Strategy Paper.
Ministry of Finance and National Planning. Lusaka, Zambia.

Government of the Republic of Zambia, 2006. Fifth National Development Plan 2006-
2010. Ministry of Finance and National Planning Lusaka, Zambia.

Gujarati, D. (2003). Basic Econometrics: Fourth Edition. Boston, MA: McGraw-Hill
Higher Education.

Haggblade, S., Hazell, P. & Reardon, T. (2007). Transforming the Rural Nonfarm
Economy. Opportunities and Threats in the Developing World. Johns Hopkins
University Press.

Haggblade, S., Hazell, P. & Brown, J. (1989). F arm-nonfarrn linakes in rural sub-saharan
Africa. World Development, 17 (8), 1173-1201.

Hamilton, K., Tembo, G., Sinyenga, G., Bandyopadhyay” 8., Pope, A., Guillon B.,
Muwele B., & Pavy J. (2007). The real economic impact of Nature Tourism in
Zambia. Natural Resources Consultative Forum.

Hausman, J .A. (1987). Specification tests in econometrics. Econometrica, 46 (6), 1251-
1271.

Honey, M. (1999). Ecotourism and sustainable development: who owns paradise?
Washington, DC: Island Press

International Institute of Environment and Development (IIED). (1994). Understanding
Whose Eden? An Overview of Community Approaches to Wildlife Management.

Jayne, T.S., Govereh, J ., Chilonda, P., Mason, N., Chapoto, A. & Haantuba, H. (2007).
Trends in agricultural and rural development indicators in Zambia. Working
paper No. 24. Food Security Research Project. Lusaka, Zambia.

116

Jayne, T.S., Yamano, T., Weber, M.T., Tschirley, D., Benfica, R., Chapoto, A., & Zulu,
B. (2003). Smallholder income and land distribution in Africa: implications for
poverty reduction strategies. Food Policy. 28 (3), 253-275.

Leader-Williams, N. & Albon, SD. (1988). Allocation of resources for conservation
Nature, 336(8), 533-535.

Lewis, D. & Alpert, P. (1997). Trophy Hunting and Wildlife Conservation in Zambia.
Conservation Biology. 11 (1).

Li, TM. (2002). Engaging Simplifications: Community-Based Resource Management,
Market Processes and State Agendas in Upland Southeast Asia World
Development. 30 (2), 265—283.

Lin, T. & Schmidt, P. (1984). A test of the Tobit specification against an alternative
suggested by Cragg. The Review of Economics and Statistics, 66 (1), 174-177

Lungu, F .B. (1990). Living with Wildlife: Wildlife Resource Management with Local
Participation in Africa. World Bank Technical Paper number 130. Africa
Technical Department Series.

Marks, SA. (2001). Back to the Future: Some Unintended Consequences of Zambia's
Community-Based Wildlife Program (ADMADE) Africa Today. 48 (1),121-141.

Mudenda, M.M. (2006). The Challenges of Customary Land Tenure in Zambia. TS 50 -
Conflicts and Land Administration. Shaping the Change XXIII FIG Congress
Munich, Germany.

Mutandwa, E. & Gadzirayi, CT. (2007). Impact of community-based approaches to
wildlife management: case study of the CAMPF IRE programme in Zimbabwe.
International Journal of Sustainable Development & World Ecology.l4, 336-344.

Neumann, R. P. (1997). Primitive Ideas: Protected Area Buffer Zones and the Politics of
Land in Africa Development and Change, 28 (3), 559-582.

117

Ooi, Jin-Bee. 1982. Commonwealth Geographical Bureau. Natural Resources in Tropical
Countries. Singapure: Singapure University Press Proprietary, Limited.

Reardon, T. (1997). Using Evidence of Household Income Diversification to Inform
Study of the Rural Nonfarm Labor Market in Africa. World Development 25 (5),
735-747.

Simasiku, P., Simwanza, H.I., Tembo, G., Bandyopadhyay, S., & Pavy, J. (2008). Game
Management Areas in Crisis. The impact of wildlife management policies on
communities and conservation. Natural Resources Consultative Forum.

Songorwa, A.N. (1999). Community-Based Wildlife Management (CWM) in Tanzania:
Are the Communities Interested? World Development 27 (12), 2061-2079.

Swanepoel and Scandrol. Available from http://www.hunt-africa.com/zambia-huntim:
safaris-overview.php. Accessed on February 2009.

 

Tschirley, D.L. & Benfica, R. (2001). Smallholder agriculture, wage labour and rural
poverty alleviation in land-abundant areas of Africa: evidence from Mozambique.
The Journal of Modern African Studies, 39 (2), 333-358.

Tschirley, D.L. & Kabwe, S.. (2007). Cotton in Zambia: 2007 Assessment of its
Organization, Performance, Current Policy Initiatives, and Challenges for the
Future Working Paper No. 26 Food Security Research Project. Lusaka, Zambia.

Van Loenen, B. (1999). Land Tenure in Zambia. University of Maine. Department of
Spatial Information Engineering.

Wainwright, C. & Wehrmeyer, W. (1998). Success in Integrating Conservation and
Development? A Study from Zambia. World Development. 26 (6), 933-944.

Woldehanna, T., Lansink, A.O. & Peerlings, J.. (2000). Off-farm work decisions on
Dutch cash crop farms and the 1992 and Agenda 2000 CAP reforms. Agricultural
Economics. 22, 163-171.

118

 

Wooldridge, J.M. (2002). Introductory Econometrics: A modern approach. Business &
Economics. Thomson-South-Western.

Wright RM. (1995). Alleviating Poverty and Conserving Wildlife in Afi'ica: An
"Imperfect" Model from Zambia National Association for the Practice of
Anthropology. 15 (1), 19-32.

Yunez-Naude, A. & Taylor, J.E. (2001). The Determinants of Nonfarm Activities and
Incomes of Rural Households in Mexico, with Emphasis on Education. World
Development. 29 (3), 561-572.

Zambia Wildlife Authority. Zambia Wildlife Act (1998). Available from:
http://ww.za\~’a.og.21n.

 

Zambia Wildlife Authority (2008). Available from http// wvwvzawaorgzm.

119

I

S
E

R I
mmmmm
8 l
l..,
U I
H. |-.
Y l,
T H
N. ll
R H
E I

till

I

'lllljtlijjj

63 5548

030