Wk 6 Discussion (Due in 1 day) Urgent/..Wk 6 Discussion (Required Assignment) Due in 1 day.docx Required Annotated Bibliography (3-4 Pages) & it’s response (1 Paragraph for student post) (Must be 4 Pages wriiten, in page count/Length Not included Title 1st Page) Must be 100% Original Work Assignment must be follow Rubric Superior Criteria. Plz read My Note, Important tips (Wrote on 2nd Page) and also sample paper attached. Must be use attached Three Article NOTE: I hv attached 3 Articles & include each Article have (3 para) three paragraph summary, Analysis and application to the study. New Selected topic: Strategies Used by Agency Leaders to Safeguard Rosewood Trade (Annotated Bibliography must be write on related this topic & Apply) MY Notes: (Must see sample paper) Sample Annotated Bibliography attached so must be follow & minimum 3 pages required & three (3) peer-reviewed sources (no older than 5 years). (4 Pages required )Must be include Abstract/Intro like in sample Course: DDBA – Doctoral Study Mentoring Selected topic: Strategies Used by Agency Leaders to Safeguard Rosewood Trade Discussion 2: Annotated Bibliography In each week of this course, you will research and select three (3) peer-reviewed, scholarly sources to develop an annotated bibliography that you can use in your Doctoral Study. You will need to take the three sources and synthesize

 

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/228071161

Impact of Community-Based Forest Management on Forest Protection:

Evidence from an Aid-Funded Project in Ethiopia

Article  in  Environmental Management · June 2012

DOI: 10.1007/s00267-012-9887-5 · Source: PubMed

CITATIONS

56 READS

2,716

2 authors:

Some of the authors of this publication are also working on these related projects:

Sustainable Cocoa Production Program (SCPP) View project

Ryo Takahashi

Waseda University

10 PUBLICATIONS   265 CITATIONS

SEE PROFILE

Yasuyuki Todo

Waseda University

91 PUBLICATIONS   2,475 CITATIONS

SEE PROFILE

All content following this page was uploaded by Yasuyuki Todo on 21 July 2014.

The user has requested enhancement of the downloaded file.

 

 

Impact of Community-Based Forest Management on Forest Protection: Evidence from an Aid-Funded Project in Ethiopia

Ryo Takahashi • Yasuyuki Todo

Received: 13 September 2011 / Accepted: 20 May 2012 / Published online: 27 June 2012

� Springer Science+Business Media, LLC 2012

Abstract Many African countries have adopted com-

munity-based forest management (CBFM) to prevent

deforestation. However, empirical studies have not reached

a consensus on the effectiveness of CBFM. The purpose of

this study is to examine the impact of the establishment of

participatory forest management associations in Ethiopia.

We used remote sensing data to gauge the change in forest

area and employed a two-stage least squares model to

correct for possible biases. The results indicate that the

forest area managed by forest associations declines more in

the year of establishment than forest areas with no asso-

ciation. This finding suggests that villagers may engage in

‘‘last-minute’’ logging. However, 1 year after the estab-

lishment of the forest associations, the forest area of the

associations increased substantially, most likely because

the associations monitor illegal logging, enabling the

regeneration of open areas within the registered forest area.

On average, the forest area of the forest associations

increased by 1.5 % in the first 2 years, whereas forest areas

not managed as part of an association declined by 3.3 %.

The cumulative impact over 2 years yields a net increase in

the rate of change of 4.8 %. These results demonstrate that

it is important to improve the monitoring of forest areas

during the initial establishment of participatory forest

management associations to maximize the effects of asso-

ciation establishment.

Keywords Impact evaluation � Remote sensing � Forest protection � Community-based forest management � Ethiopia

Introduction

The deforestation of tropical forests is a widespread

problem in less developed countries, including those in

Sub-Saharan Africa (Achard and others 2002). For exam-

ple, 35 % of the total land area of Ethiopia was covered by

forest in the early twentieth century, but by the early 1950s,

forest cover had drastically declined to 16 % (Ethiopian

Forestry Action Programme 1994, cited in Urgessa 2003).

This decline has continued in recent years, with the total

forest cover falling from 13.8 % in 1990 to 12.7 % in 2005

(United Nations Statistical Division 2010).

Among the many factors driving deforestation, agricul-

tural expansion due to poorly defined land ownership and

land-use rights are the most common in Sub-Saharan

Africa (Geist and Lambin 2002). To prevent deforestation,

African countries have recently promoted community-

based forest management (CBFM) rather than centralized

forest management (Ribot and others 2010) in response to

studies showing that local communities manage resources

more sustainably than government agencies (Agrawal and

Yadama 1997; Ostrom 1999; Adams and others 2003).

This effectiveness of local management is more likely in

less developed countries, where most local societies are

based on trust and cooperation among members of a

community, leading to the monitoring and punishment of

irresponsible users (Pretty and Ward 2001; Hayami and

Godo 2005; Tole 2010).

A large number of empirical studies have found evi-

dence of the effectiveness of CBFM on forest conservation

(Edmonds 2002; Somanathan and others 2005; Dalle and

others 2006; Matta and Alavalapati 2006; Ellis and Porter-

Bolland 2008; Lund and Treue 2008). In contrast, a

research study conducted by Kijima and others (2000)

found that private ownership was a more efficient way of

R. Takahashi (&) � Y. Todo Department of International Studies, University of Tokyo,

5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8563, Japan

e-mail: ggt000@hotmail.com

123

Environmental Management (2012) 50:396–404

DOI 10.1007/s00267-012-9887-5

 

 

managing forests than common ownership. Thus, there is a

lack of consensus in the literature. Moreover, these studies

focused only on the post-implementation effects, and the

effects during the process of CBFM implementation were

overlooked. Alemagi (2010) noted that the establishment of

community-based procedures is considerably time con-

suming; for example, this process took an average of

approximately 18 months in Cameroon and 18–24 months

in Canada. Such intervals may influence the behavior of

residents in the community, thereby affecting the rate of

change in forest area.

The objective of this study, therefore, is to quantify the

impact of CBFM on forest protection as well as the effect

of CBFM during the process of its implementation. As a

case study, we use Ethiopia, where deforestation is severe,

because most of the existing studies on deforestation in less

developed countries target Asia or South America, and the

data for Africa is insufficient. The form of community-

based forest management analyzed is forest associations,

which have grown to number over 100 in Ethiopia.

In 2007, the Participatory Forest Management Project

(hereafter, ‘‘the project’’) was implemented in the Belete–

Gera Regional Forest Priority Area (RFPA), located in the

southwest of Ethiopia, by the Japan International Cooper-

ation Agency (JICA), a Japanese foreign aid agency that

establishes participatory forest management associations

(hereafter, ‘‘forest associations’’). Forest associations were

established at the sub-village level, the lowest level of

residence, to identify the border between forest and

homestead/farmland and to prevent logging in the forest

area. After establishment, the forest associations are tasked

with managing the registered forest area.

Description of the Project

The project targets the RFPA located in the Gera District and

the Shabe Sombo District in the Oromia Region (Fig. 1).

In 2007, the project began establishing participatory for-

est management associations (‘‘WaBuB’’ in the local lan-

guage) at the sub-village level to undertake forest protection

in the RFPA, together with income-generation activities,

such as farmer field schools and coffee certification (Ta-

kahashi and Todo 2011; Todo and Takahashi 2011). Table 1

shows the number of forest management associations

established by year. Preparation for the establishment of an

association usually takes 1 year, beginning in October and

ending around August or September in the following year.

Before the implementation of the project, the decision to

extract wood from the forest and expand farming or grazing

land into the forest area was left to the individual judgment of

the villagers. However, once a forest management associa-

tion is established, the border between homestead/farmland

and forest is clearly identified and marked with paint by

association members; using the registered forest area for

activities such as expanding farmland, extracting wood, or

planting trees is strictly prohibited, except in the case of

necessary thinning. Members of the associations regularly

monitor and evaluate the conditions of the forest. When they

find open spaces in the registered forest area, the forest

Fig. 1 The Belete–Gera Regional Priority Area. The participatory forest management associations (‘‘WaBuB’’ in the local language) were established at the sub-village level, which is shown in yellow. The background image is a false-color image of Landsat 5 imagery taken in January 2001. Active vegetation appears red-pink, bare soil and fallow fields are green, and urban structures are bluish-white

Environmental Management (2012) 50:396–404 397

123

 

 

association holds a meeting to plan for the care of these

spaces and prevent the degradation of the forest. In addition,

forest associations plant trees along the boundary of the

forest area to demarcate the boundary and distinguish the

forest area from areas of agroforestry.

Data

To estimate the forest area at the sub-village level, we use

data from satellite images of Landsat 7 with a resolution of

30 m. Landsat images from path/row 170/55 for January in

the following 5 years were used for our analysis: 2006,

2007, 2008, 2009, and 2010. Because the image for 2008

was affected by cloud cover, we build a composite image

using data from December 2007.

To distinguish between forest areas and non-forest areas,

we utilize the Normalized Difference Vegetation Index

(NDVI), a measure of vegetation commonly used in remote

sensing studies, such as those by Tucker and others (1985),

Davenport and Nicholson (1993), and Tucker and others

(2001). The NDVI takes values of between -1.0 and 1.0

and increases with the degree of vegetation biomass

(Jensen 1996). Following Southworth and others (2004), we

determined a threshold value of the NDVI for forest areas

based on information from satellite images and fieldwork.

We conducted ground truthing to collect locational data for

17 points at boundaries between forest and non-forest areas

that were in place during the study period (according to

interviews with several local residents) and chose the point

with the highest NDVI value for each year as the threshold

value for forest areas. Forest areas are defined as areas that

function as forest either physically or socially for local

communities (Southworth and Tucker 2001). Non-forest

areas include agricultural lands, fallow fields, rangelands,

cleared areas, bare soil areas, and urban areas.

Although this methodology has been used in previous

studies, such as those by Southworth and others (2004) and

White and Nemani (2006), there may be some errors in

estimating the NDVI threshold value, and these errors may

lead to errors in the rate of change in forest area. However,

because the same error would affect any locational unit

within the same year, the rate of change in the forest area for

sub-villages with and without a forest association would be

over- or underestimated to the same extent. Therefore, the

possible error in the estimation of forest area from satellite

images does not lead to a bias in the estimation of the impact

of forest associations on the rate of change in forest area.

The boundaries of the registered forest areas were recor-

ded by with a global positioning system (GPS) device. The

total number of forest associations and the average and total

area are provided in Table 2. The total area brought under the

management of forest associations between 2007 and 2009

(a total of 92 forest associations) was approximately

60,000 ha. A comparison of the environmental characteris-

tics of the Gera and Belete Forest Areas is listed in Table 3.

The forest area decreased by an average of approximately

1.7 % annually between 1995 and 2006 in the areas studied.

Empirical Framework

This section provides an overview of the empirical frame-

work employed in this study. First, to control for time-

invariant unobservable characteristics, we use the difference

in the rate of change in forest area as a dependent variable.

Second, we apply a two-stage least squares (2SLS) model to

control for possible biases and estimate the impact of the

establishment of the forest associations on the rate of change

in forest area. The estimation equation is specified as follows:

log yit � log yit�1 ¼a þ b1EST0it�1 þ b2EST 1 it�1

þ b3EST 2 it�1 þ b4PERI

07 it�1

þ b5PERI08it�1 þ b6Xit þ eit; ð1Þ

where yit is the forest area of sub-village i in year t; there-

fore, log yit – log yit-1 indicates the rate of change in forest

area of sub-village i between the years t and t – 1. To

Table 1 The establishment of forest management associations

Total number

of sub-villages

in RFPA

Number of associations established

in

2007 2008 2009 2010

Gera Forest 80 2 19 35 24

Belete Forest 45 1 13 22 9

Total 125 3 32 57 33

Table 2 Summary statistics of the area under management by

forest associations by year of

establishment

Number of

forest associations

Area (ha)

Mean SD Median Min. Max. Total area

Established in 2007 3 2,814 3,762 977 323 7,141 8,442

Established in 2008 32 671 738 421 57 2,782 21,468

Established in 2009 57 549 617 338 20 2,932 31,272

Total 92 665 944 378 20 7,141 61,182

398 Environmental Management (2012) 50:396–404

123

 

 

examine the impact of the establishment of forest associa-

tions and the effect of their establishment, we create three

dummy variables at three different points in time. EST0it�1 takes a value of 1 if sub-village i was in the process of

establishing a forest association in year t – 1. EST1it�1 and

EST2it�1 take a value of 1 if sub-village i has had a forest

association for 1 or 2 years, respectively, in year t – 1.

We also include the dummy variables PER107it�1 and

PER108it�1; which take a value of 1 if sub-village i is located

next to an area where a forest association was established

in 2007 and 2008, respectively. Some previous studies

found that the establishment of forest associations induces

deforestation in the surrounding forest area (Chakraborty

2001; Ostrom and Nagendra 2006; Balooni and others

2007). These dummy variables would capture such a neg-

ative leakage effect if the establishment of forest associa-

tions causes deforestation in the surrounding forest area.

We refer to the previous studies and employ environ-

mental variables for the estimation (Cropper and Griffiths

1994; Chen and others 1999; Cropper and others 1999; Zhang

and others 2005). Xit denotes the environmental factors of

sub-village i during the observation period and includes the

following: the number of households in the sub-village, the

average slope, and the proportion of acrisol present.

The independent variables EST0it�1; EST 1 it�1; and EST

2 it�1

in Eq. (1) are endogenous because they are correlated with

eit; as a result, we take two steps to reduce endogeneity, as suggested in Duflo and Pande (2007).

First, we use a multinomial logistic model to identify the

determinants of the establishment of forest associations:

ProbðESTi ¼ jÞ¼ expðc0jZi þ d

0 jXiÞ

1 þ P2

j¼1 expðc0jZi þ d

0 jXiÞ

; ð2Þ

where ESTi represents the year of establishment in sub-

village i: 2007 (j = 2), 2008 (j = 1), or 2009 (j = 0). We

assume c 0 0 ¼ 0 and d

0

0 ¼ 0 for normalization.

Zi represents variables that determine the selection of

sub-villages for forest associations but do not determine the

change in forest area, including the distance to the closest

project office (the project has two offices: in Jimma City

and Gera District), the deforestation rate between 1995 and

2006, and the dummy variable for Gera District.

Second, to reduce biases due to endogeneity, we cal-

culate the predicted probability of the establishment of

forest associations for each year from Eq. (2) and employ

the predicted probability as instruments for a 2SLS model

of Eq. (1). The predicted probabilities of the establishment

of associations are closely related to the dummies for the

actual establishment, but they are controlled by a nonlinear

function of geographic variables. Therefore, the predicted

probabilities are most likely to be unrelated to the error

term in Eq. (1) and are able to serve as instruments to

estimate impact of the forest associations.

Estimation Results

The estimation results from the multinomial logistic model

are presented in Table 4, with columns 1 and 2 showing the

results for forest associations established in 2007 and 2008,

respectively. As expected, the distance to the closest pro-

ject office has a negative and significant effect on the

establishment of forest associations in 2008, implying that

being closer to a project office increases the likelihood of

early selection. In contrast, the result for this variable is

insignificant for 2007; however, this finding may be an

effect of the small number of observations; that is, only

three forest associations were established in 2007. The size

of a sub-village as indicated by the number of households

has a significant positive impact on a sub-village being

selected in 2007 but not in 2008, indicating that forest

associations were established in larger sub-villages earlier

in the implementation of the project. Finally, in 2008, the

deforestation rate between 1995 and 2006 had a weakly

Table 3 Summary statistics of the environmental characteristics of Gera and Belete Forests

Gera Forest (n = 56) Belete Forest (n = 36)

Mean SD Min. Max. Mean SD Min. Max.

Average annual rate of change in forest

area between 1995 and 2006

-0.014 -0.032 -0.165 0.043 -0.021 -0.070 -0.179 0.201

Average area of forest associations (ha) 752 1,076 38 7,141 529 684 20 2,932

Number of households 158 82 33 529 106 47 22 208

Average elevation (m) 2,101 283 1,408 2,725 1,971 217 1,534 2,380

Average slope (%) 8.2 5.0 0.8 20.0 13.0 6.2 1.6 30.4

Proportion of acrisol (%) 0.1 0.3 0 1.7 0.2 0.3 0 1.5

Acrisol is a type of soil classified by the Food and Agriculture Organization (FAO) and the United Nations Educational, Scientific and Cultural

Organization (UNESCO) with a subsurface accumulation of low activity clays and low base saturation. As such, acrisol is relatively unfertile

Environmental Management (2012) 50:396–404 399

123

 

 

significant (P 0.10) positive effect on the selection of forest associations. This finding indicates that the higher

the rate of deforestation, the more likely a village was to be

selected in 2008.

We then perform 2SLS to estimate the impact of the

forest associations using the predicted probability of the

establishment of forest associations in each period as an

instrument. The results of the first stage of 2SLS are pre-

sented in columns 1–3 in Table 5, showing that the pre-

dicted probabilities are significantly correlated with the

actual establishment and presence of forest associations.

Therefore, the instruments appear to be robust for the 2SLS

estimation.

The results of the second stage of the 2SLS estimation

are presented in column 4. The results indicate that the

dummy for the year of establishment is negative and sig-

nificant at the 5 % level of significance. The value of the

coefficient indicates that the rate of change in forest area

for sub-villages with a forest association was 12 % lower

than that in sub-villages without a forest association.

In contrast, we find that the dummy for 1 year after the

establishment is positive and significant at the 10 % level

of significance. We acknowledge that the 10 % level of

significance found in these estimations is not very high.

However, we check the robustness of the results by varying

the independent variables in Eq. (1), and we still find the

same statistical significance (see column 5 in Table 5).

This result indicates that even though the rate of change

decreased during the establishment period, once a forest

association had been set up, the rate of change in the areas

with forest associations is 16.9 % points higher than in

areas without forest associations. In the case of the dummy

for 2 years after the establishment, we find no significant

effect; however, as above, this finding may reflect the small

number of observations for forest associations established

in 2007 (i.e., three).

To highlight the results, let us consider what would

happen if a forest association were to be established in a

hypothetical sub-village with a forest area of 1 km 2

and the

average characteristics of all sub-villages. The average

annual instantaneous change rate in forest area (i.e., the

first difference in the log of the forest area) in this study

area is approximately -1.7 %, causing the forest area of

the sub-village to be reduced to 0.983 km 2

(=e -0.017

)

1 year later and 0.967 km 2

(=e -0.017-0.017

) 2 years later

without a forest association. However, if an association

were to be set up, the forest area would be reduced to

0.872 km 2

(=e -0.017-0.120

) 1 year later. It is important to

note that the dependent variable in our estimation is the

first difference in the log of the forest area, i.e., the

instantaneous change rate defined as (dyit/dt)/yit, rather than

the discrete-time change rate defined as (yit – yit-1)/yit-1.

Therefore, the forest area 1 year after the establishment of

the forest association is not 1 – 0.017 – 0.120 = 0.863,

but e -0.017-0.120

= 0.872, although the two values are

close to each other. Two years after the establishment of

the forest association, the forest area would become

1.015 km 2

(=e -0.017-0.120-0.017?0.169

). Therefore, our

results suggest that the establishment of a forest association

increases the forest area by 1.5 % in the first 2 years

compared with a decrease of 3.3 % without the association.

Thus, the establishment of forest associations leads to a net

increase in forest area of 4.8 %.

Another important finding is that both dummy variables

for areas located near forest associations established in

2007 and 2008 are insignificant, which suggests that the

establishment of forest associations does not exert greater

pressure on surrounding forest areas.

Furthermore, we test for the endogeneity of the three

dummy variables in Eq. (1) using Durbin–Wu–Hausman

(DWH) tests. The null hypothesis is rejected (P 0.01); in other words, these dummy variables are endogenous.

Table 4 Results from the multinomial logistic estimation

t Statistics are in parentheses

* and ** indicate statistical

significance at the 10 and 5 %

levels, respectively

Variables Established

in 2007

Established

in 2008

(1) (2)

Number of households in sub-village 0.8611* (1.7986) -0.5566 (-1.2289)

Average slope 0.2143 (1.1635) 0.0746 (1.3947)

Proportion of acrisol -18.5952 (-0.0840) -19.2497 (-0.2415)

Deforestation rate between 1995 and 2006 11.2334 (0.5582) 8.6530* (1.6869)

Distance to the closest project office -0.2014 (-1.0318) -0.0799** (-2.2814)

Dummy of district (1 = Gera) -3.2898 (-0.7082) -1.7849* (-1.8377)

Constant -0.6576 (-0.1154) 1.7689 (1.3154)

Observations 92

Log likelihood -61.655

Pseudo R2 0.136

400 Environmental Management (2012) 50:396–404

123

 

 

T a

b le

5 R

e su

lt s

fr o

m th

e in

st ru

m e n

ta l

v a ri

a b

le e st

im a ti

o n

In d

e p

e n

d e n

t v

a ri

a b

le s

(1 )

(2 )

(3 )

(4 )

(5 )

2 S

L S

2 S

L S

1 st

S ta

g e

1 st

S ta

g e

1 st

S ta

g e

2 n

d S

ta g

e

N u

m b

e r

o f

h o

u se

h o

ld s

in su

b -v

il la

g e

0 .0

0 2

5 (0

.1 1

0 2

) –

0 .0

0 1

8 (-

0 .1

0 6

5 )

– 0

.0 0

1 0

(- 0

.1 9

5 6

) –

0 .0

0 6

6 (-

0 .3

9 8

7 )

A v

e ra

g e

sl o

p e

0 .0

0 1

0 (0

.2 8

6 1

) –

0 .0

0 0

4 (-

0 .1

4 7

6 )

– 0

.0 0

0 0

(- 0

.0 0

3 0

) 0

.0 0

0 5

(0 .1

8 8

3 )

P ro

p o

rt io

n o

f a c ri

so l

– 2

.9 6

6 6

(- 0

.5 0

0 4

) 0

.0 1

9 2

(0 .0

0 4

4 )

0 .0

0 3

3 (0

.0 0

2 4

) –

2 .2

2 6

0 (-

0 .5

1 9

1 )

D u

m m

y fo

r a re

a s

lo c a te

d n

e a r

a fo

re st

a ss

o c ia

ti o

n e st

a b

li sh

e d

in 2

0 0

7

– 0

.0 7

8 2

(- 0

.6 6

8 2

) –

0 .0

0 0

1 (-

0 .0

0 1

3 )

0 .0

0 0

3 (0

.0 0

9 5

) –

0 .0

9 6

5 (-

1 .1

3 6

3 )

D u

m m

y fo

r a re

a s

n e a r

a fo

re st

a ss

o c ia

ti o

n

e st

a b

li sh

e d

in 2

0 0