HOUSEHOLDS‘ WILLINGNESS TO PAY FOR RESTORING

ENVIRONMENTAL RESOURCE: A CASE STUDY OF

FOREST RESOURCE FROM DIRE DAWA AREA, EASTERN,
ETHIOPIA

Alem Mezgebo1

Abstract

This paper assesses households‘ perceptions on forest lose and presents empirical
results of households‘ willingness to pay for restoring forest resource in Dire Dawa
area, Ethiopia. Probit and Tobit models were applied to determine the mean and
factors affecting willingness to pay for forest restoration, respectively. A sample of 393
households was randomly selected, and the survey was used a face to face interviews.
However, after checked for sample selection bias 10 protest bidders were excluded
from the data set. The descriptive analysis shows that the forest resources have been
cleared. That is, 82% of the respondents reported that the reasons attributed to the
forest lost were population pressure, overgrazing, soil and water degradation and
agricultural expansion. The econometric result shows that the mean willingness to pay
from double bound elicitation method was computed at 94.09 ETB with the total
willingness to pay 2,026,604.51 ETB (1 US$=18.44 ETB) per annum for five years.
Whereas, the mean willingness to pay from open ended elicitation method was
computed at 64.82 ETB with aggregation value of 1,396,157.98 ETB per year. The
result from double bounded elicitation method is greater than open ended elicitation
method. This might be due to anchoring effect from the double bounded method.
The result suggested that any forest restoration intervention in the study area needs to
consider monthly income, initial bids, perception, educational level, ownership type
and access to extension services for successful forest restoration activities. Total farm
land holding and sex are also significant variables needs to consider.

Key Words: Willingness to Pay, Contingent Valuation Method, Forest Restoration,

Probit Model, Tobit Model
JEL Classification: Q00; Q20; Q50; Q57; Q59.

1
College of Agriculture and Environmental Science, P O Box, 336, Haramaya University, Ethiopia; Email:
adeway12@gmail.com; or yene1212@gmail.com; Tel: +251-913563182
The Author is grateful for the research grant from Ethiopia Economic Association (EEA) and the
International Growth Center (IGC). Without their support, completion would have not been real.

33
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

1. Introduction

Forest is a minimum area of land with a tree crown cover of more than 10-30% and
0.05-1 hectare with the trees height of 2 to 5 meter at maturity in situ (FAO, 2006).
Forests, like other natural resources perform a set of functions to meet the needs of
people (Cavatassi, 2004; Pak et al. 2010). Forests are providing various direct and
indirect benefits to human welfare (Dogru, 2001; WWF and IUCN, 2001; Chaudhury,
2006 and Wu et al. 2010). Some of these benefits include increase precipitations,
recreation, timber production and fuel wood (Bishop 1999; WWF and IUCN 2001
and Turner et al. 2004). On the other hand, forest resources are using for protecting
watershed, reducing erosion and removing greenhouse gas from the atmosphere
(WWF and IUCN, 2001). However, the degree of deforestation and forest degradation
is more fast and huge especially in the developing tropical countries (FAO, 2005). It is
estimated that 350 million hectares of tropical forest land have been severely damaged
(ITTO, 2002). The attenuation and degradation of forest enhance soil erosion,
decreasing water quantity and household income, and hence increase poverty (Azene
Bekele-Tesemma, 2002 and Maginnis and Jackson, 2003).

In Ethiopia a data on forest resources showed that it is among countries with forest
cover of 10‐30%. According to this report Ethiopia‘s forest cover is 12.2 million
hectare. It further indicated that the forest cover shows a decline from 15.11million ha
in 1990 to 12.2 million ha in 2010, during which 2.65% of the forest cover was
deforested (FAO, 2010). This showed that the country is characterized by high rate of
deforestation. The major causes of deforestation are expansion of agricultural land, and
increasing need for fuel wood, and overgrazing. Consequently, in Ethiopia as well as in
the study area because of deforestation households have been faced shortage of fuel
wood, land productivity problems, flood, low income and shortage of water. In general,
deforestation and over-exploitation of forest resource leads to low economic benefits,
unsustainable economic development and hence poverty.

Consequently, restoration of forest resource is very important from both a

socioeconomic and environmental angle. Because, forest restoration can be used to
reverse some of the more severe impacts of forest loss and degradation by providing a
range of forest products (Maginnis and Jackson, 2003). Moreover, forest restoration
may mitigate global climate change by reducing carbon stocks. For example, restoring
forest resource effectively in the non cultivated land of the study area may continue to
provide the economic valuable services of forest resource to the people living around

34
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

the resource. However, in the study area no attempt was made to estimate the
economic value of forest restoration using acceptable environmental valuation
techniques. Fail to estimate the economic value of forest restoration enhance the
complexity of forest restoration and management decision.

Yet, in recent decades, concerns have arisen about the proper valuation of
environmental resources and progress has been achieved in developing valuation
methods (Kramer et al. 1997 and Wu et al. 2010). An economists use the concept of
willingness to pay (WTP) to determine consumers WTP for improved and avoiding
deterioration of environmental resource (Agudelo, 2001 and Pearce, 2002). In this
study therefore, an attempt was done to estimate the economic values of forest
restoration in the non cultivated land of Dire Dawa area. Besides, the study assessed
the level of households perception on problems of forest lose, and determined the
factors affecting households WTP. It is believed that the study plays a key role in
formulation of a successful forest policy and determination of the real contributions of
forest resources to sustainable economic development. The study also helps the
government and concerned body to identify salient households‘ features that would
increase the targeting and subsequent success of forest restoration activities in the study
area as well as in other area with similar characteristics. Contingent valuation (CV)
method was used to identify households‘ WTP for forest restoration.

2. Theory of Welfare Economics

In the case of welfare economics the main purpose of any economic activities is to
increase the well being of the responding individuals or economic agents. In this study,
the basic assumption is an individual makes a decision to maximize their utility from
restoring forest resource given income constraints. Following this, measurement of the
economic values of forest restoration is depends on the effect of the hypothesis project
on households‘ welfare.

A Pareto criterion is the best way to explain welfare. It is indicating that policy changes
make at least one person better off without making any one worse off. Besides, Pareto
improvement noted public intervention is good for efficient resource allocation. If the
cost of the public action is less than the sum of the benefits from a public action, it is
considered worthwhile by the criterion (Haab and McConnell, 2002). The applied side
of modern welfare economics works a variant of the Pareto criterion by trying to find
ways to place a dollar value on the improvement and deterioration from environmental

35
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

changes. This allows the calculation of net gain or loss from a policy change, and
determination of whether the change is potentially Pareto improvement or not.
Changes in environmental quantity and quality may affect individual‘s welfare. Because
environmental change may lead to changes in prices an individual pay for marketed
goods and inputs. Moreover, it may changes the quantities or qualities of environmental
goods such as forest resource, in our case.

Such welfare changes can be measured using ordinary consumer‘s surplus,

compensating variation, compensating surplus, equivalent variation and equivalent
surplus. However, ordinary consumer‘s surplus does holds income constant but not the
level of utility. On the other hand, compensating variation and compensating surplus
measures of the gains or loss and hold utility constant at the initial level, while
equivalent variation and equivalent surplus measures welfare change and hold utility
constant at some specified alternative level. Generally, depending on the consumers‘
property position vis-à-vis the good in question (in this study forest restoration) all these
Hicks welfare measures involve either payment or compensation to maintain utility at
the specified level (Randall and Stall (1980), cited from Mitchell and Carson, 1989). If
the proposed change is welfare increasing (restoring forest resource), which is the focus
of this study the appropriate welfare measure, is the compensating surplus. This
measure can be interpreted as the consumer‘s WTP for the environmental resource
which maintains their initial utility level constant (Mitchell and Carson, 1989).

In Hicksian welfare measure estimation of the economic benefits of the environmental

goods requires identification of the actual demand function for the improvement of the
environmental goods. However, it is very tricky to estimate the actual demand curve
since it requires accurate market data. Therefore, to fill this problem we should use an
alternative method which requires the creation of hypothetical market scenario. The
alternative method is that a CVM and this method can generate the WTP data, which
will be used to value the forest restoration without having to estimate the actual demand
curve.

This concept can be further emphasized from the relationship between the expenditure
function and Hicksian compensated surplus measure (CS). According to Haab and
McConnell (2002), the expenditure function that provides the theoretical structure for
welfare estimation is specified as:

(1)

36
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Where: M is the minimum amount of income needed to maintain utility level given the
price and public good vectors; q= is the vector of environmental goods; p= is a vector of
prices; u=is level of utility when , x= is the vector of private goods and y=
income.

Let represent some initial level of those respective arguments and

represent some succeeding levels. The compensation surplus can be
specified by:

(2)

is preferred to for proposed new project brings welfare gain (just like in the case
of this study). In this case, the compensated surplus (CS) measure tells us the
consumers‘‘ WTP for welfare gain. Contingent valuation is capable of obtaining the
appropriate Hicksian measure for a proposed change in the public good (Mitchell and
Carson 1989). It can be viewed as a way of estimating the change in the expenditure
function (Haab and McConnell 2002). Coming to the case of forest restoration of this
study the value of forest resource could be determined through household WTP.

3. Empirical Reviews

Contingent valuation surveys have been widely applicable methods in valuing use and
non use values of environmental goods and services (like forest resource) (Whittington
et al., 1990; Whittington 1998). There have been a large number of studies for valuing
non-market benefits of forests in monetary terms using contingent valuation technique.
A few selected case studies pertaining to certain forests in the world as well as in
Ethiopia are discussed below.

Bin Ramlan et al. (2011) were estimated the economic value of forest research institute
Malaysia‘s canopy walkway from visitors using CVM in the form of dichotomous
choice elicitation methods. The authors used a Logit and Probit models to estimate the
visitor‘s WTP responses for the access to the walkway. Based on the estimation results,
the calculated mean of WTP ranged from MYR5.33 to MYR213.32 for the logit model,
whereas the value ranged from MYR5.39 to MYR13.02 for the probit model based on
95 % confidence interval. The study had shown that visitors to forest research institute

2
MYR refers Malaysian Ringgit which is the Malaysian currency.

37
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

Malaysia‘s canopy walkway are willing to pay about MYR7.61 for the entrance permit.
The study concluded that the entrance fee collections are used as additional funds for
the costs of maintenance and conservation of forest research institute Malaysia‘s canopy
walkway.

A study by Carlsson et al. (2004) applied a contingent valuation method to estimate the
economic values of community plantations trees in the highlands of Ethiopia. A
discrete-continuous elicitation format was applied. The survey covered a total of 1520
households from both East Gojam and South Wollo of Amhara region. The mean
WTP from sample respondents is estimated to 10 ETB3 for the closed-ended
responses. The study found that there is a problem in applying a closed ended
elicitation format because it would exaggerate the respondent willingness to pay for
community plantations trees. Besides, the analysis of the bid function shows that
women in villages without any existing community plantation are significantly more
interested in the establishment of a plantation than men. The authors recommended
that separate interviews are made with heads and spouses when it comes to valuation of
local natural resources. The result of the study also showed that the aggregate
willingness to pay vary dramatically between villages. Therefore, the concerned body
needs to develop good tools for the selection of locations for community plantations if
they seek to maximize their contribution to welfare.

Chukwuone and Okorji (2008) were conducted a study to determines households

willingness to pay for systematic management of community forests in the rainforest
region of Nigeria. The study used the contingent-valuation method in the form of
discrete choice with open-ended follow-up question. The study was used a Tobit
model, and found that a variables wealth category, occupation, formal education and
number of females in a household positively and significantly influence WTP. Male
headed households, initial bid, number of males in a household and distance from
home to forests area negatively and significantly influence WTP. The authors
concluded that organizing the local community in systematic management of
community forests for forest conservation will enhance participation and hence poverty
alleviation.

A study by Garrod and Willis (1997) estimated the mean WTP of the public for the
non-use biodiversity value of remote coniferous forests in Britain using contingent

3
ETB refers Ethiopian birr which is the Ethiopian currency

38
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

ranking method and contingent valuation in the form of double bounded elicitation
method. It estimated the public‘s WTP for a number of forest management standards
that could be adopted to improve levels of biodiversity in remote upland coniferous
forests, which the respondent would never visit. The authors estimated the value of
marginal changes in biodiversity of remote upland coniferous forests, rather than the
total value of biodiversity in remote upland coniferous forests as a whole. The value for
increasing biodiversity of these forests using contingent valuation at the margin was £10
to £11 per household per year for biodiversity for a 30% increase of the area of this
forest type. Whereas, the value for increasing biodiversity of these forests using a
contingent ranking method was £0.30 to £0.35 per household per year for a 1%
increase in these forests.

Solomon (2004) used a contingent valuation method in the form of open ended, single
and double bounded elicitation format to elicit households WTP for multi-purpose
tree resources in three selected Districts of Arsi Zone, Ethiopia. The study considered
two groups of Trees including the common eucalyptus tree type and three indigenous
trees (Guniprus Excise, Acacia Abisinica and Acacia Seyal). The study found that the
mean WTP for eucalyptus tree from open ended, single and double bounded
elicitation format are computed at 22.79 ETB, 45.81 ETB and 38.06 ETB,
respectively. Using similar procedure the mean willingness to pay for indigenous trees
using open, single and double bounded probit model were also computed at 22.14
ETB, 44.31 ETB and 26.96 ETB, respectively. The author found that the mean WTP
from all elicitation methods were greater for eucalyptus than for indigenous trees. This
could be indigenous trees are not, in most cases, fast growing like eucalyptus tree. The
results of study also show that age, types of ownership, access to credit, the value of
livestock owned by the family and bid are significant influences households‘ WTP for
eucalyptus and indigenous trees. The study concludes that labor was the most preferred
payment vehicle than cash and kind.

The study by Tefera (2006) applied contingent valuation method in the form of double
bounded elicitation method to estimate the economic values of improved natural forest
in Wondo-Wosha Sub-catchment, Ethiopia. A questionnaire survey was conducted on
148 respondents from six peasant associations surrounding the natural forest. The
mean WTP for a single household was about 30 ETB (US$ 1 = 8.7 ETB) per year.
The mean willingness to accept for a single household was 44.6 birr per year.
Moreover, about 72% of the respondents gave the value of the forest at a price of 30
birr or more per year while 18% of the respondents agreed to pay nothing assuming

39
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

that they have traditional rights to the land and/or have low level of income. The bid
function analysis suggested that household income has minimum influence on WTP.
Therefore, it can be concluded that even the poor households were willing to pay the
average values in terms of time or labor contribution to save the natural forest. On the
other hand, community valuations for forest resources in the catchment do not vary
much in magnitude when the payment vehicle was changed from cash to labor
contribution. Therefore, people are concerned about conservation issues, what so ever
their status and the situations. The concept of valuing forests is well supported by the
community.

Tegegne (1999) also applied this method to elicit people‘s valuation for environmental
protection in terms of both cash requirement and labor contribution. He concluded
that households in the study area are willing contribute in terms of labor than cash.
Moreover, education, age, sex and family size turned out to be factors affecting the
willingness to pay in terms of labor.

It may be observed from these empirical studies that there are large numbers of direct
and indirect benefits of forests. Using CV method for valuation across regions, different
estimates of economic values of tangible and intangible benefits are obtained. The
variations in the estimates could be partly on account of different socio-economic
variables and the scope of the studies. Furthermore, the literatures above suggested that
contingent valuation method is viable techniques to quantify households‘ WTP for
non-marketed goods in the developing countries. Thus the given literature above
provided some sound footings to this study to value households WTP for forest
restoration in Dire Dawa area.

4. Research Methodology
4.1 Description of the Study Area

The study was conducted in Dire Dawa Administrative Council located 515 kms far
from Addis Ababa between 9027‘ and 9049‘ latitudes north and 41038‘ and 42019‘
longitude east. The landscape of the study area is varies with an altitude ranging
between 950 and 2260 m.a.s.l. The total land coverage of the study area is 128,802
hectares. About 13.47% of the total area of the administration is cultivable land; where
as 2.22% and 84.31% of the total area is urban land and non cultivated land,
respectively. Specifically, within the four sample kebeles 9,730 hectares is non
cultivated land, which is the target area. The total population of the study area is

40
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

342,827 with 171,930 male and 170,897 females (FDREPCC, 2008). The livelihood of
the population typically depends on the production of perennials and annual crops,
and rearing animals. The average farmland holding per household of the rural
households is less than 0.5 hectare (FDRE, 2001).

Figure 1: The Study Area (Source: Own sketch)

The annual temperature and rainfall is ranges between 24.8°c and 31.4°c, and 500-850
mm, respectively. The natural forest has been cleared to satisfy the demands of the ever
increasing population. The remaining high forests cover less than 1% while the majority
of the landmass is covered with shrub lands.

4.2 Sampling Techniques and Method of Data Collection

A three-stage sampling technique was used when selected sample respondents. In the
first stage 38 rural kebeles were purposively selected out of the 47 kebeles based on
identified as their livelihood is more attached to the environmental resources than the
urban kebeles. Secondly, 4 kebeles (Adada, Adigaflema, Eja Aneni and Harela) were
randomly selected out of the 38 kebeles. In the third stage, proportionally with
population percentage, a total of 393 households were selected randomly. The data was
collected using face to face interview with the heads or working members of the
households. The author and five local enumerators was administered the survey.

41
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

The CVM method was also employed to elicit households WTP for restoring forest
resource. In CV surveys, there are about four major elicitation methods, namely
payment cards, discrete choice (single bounded dichotomous choice), discrete choice
CV with follow-up questions (double-bounded dichotomous choice) and open ended.
However, due to limited experience the payment card approach is not used especially
in rural areas of developing countries (Venkatachalam, 2004) including our country
Ethiopia. The other elicitation method is single bounded dichotomous choice format.
It is easier for respondents to make willingness to pay decisions than open-ended
questions (Bennett and Carter, 1993). However, the double-bounded dichotomous
choice format is better than single-bounded in three ways. It makes clear bounds on
unobservable true WTP and sharpens the true WTP, and hence efficiency gains (Haab
and McConnell, 2002). Finally, the double-bounded dichotomous choice format is very
vital to collect more information about WTP of each respondent‘s (Hanemann et al.
1991 and Arrow et al. 1993). Therefore, this study employed the double-bounded
dichotomous choice format to elicit households‘ WTP in Dire Dawa area.

4.3 Preliminary Survey and Bids

Before the final survey a pre-test was done using 40 randomly selected households.
Then based on the pilot results the starting point prices identified for WTP were 25,
50, 70 and 100 birr per year for five years. Given this, the actual survey was undertaken
by dividing the total sampled households randomly into four groups. The field survey
was successfully completed with 10 protest zeros. The criteria for selecting protest zero
was based on the report of the NOAA Panel on contingent valuation by Arrow et al.
(1993). The mean difference of households‘ socio-economic variables of protest and
valid responses was compared using t-test and chi square test. Our sample t-test and
chi-square test indicated that the mean difference of the households‘ socio-economic
variables of protest and valid responses was insignificant. Therefore, these protest
respondents were excluded from the data set.

4.4 Data Analysis Methods

4.4.1 Probit Model

The linear regression analysis is widely accepted to analyze the relationship between the
dependent and explanatory variables (Aldrich and Nelson, 1984). In the linear
regression analysis the dependent variable is a continuous variable, while the
explanatory variables can be either dummy or continuous variables. However, when the

42
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

dependent variable in a regression model is binary (0, 1) the analysis can be carried out
using either linear probability model, logit and/or probit models (Pindyck and
Rubinfeld 1981). But, the results of linear probability model may face the following
problems. Firstly, the linear probability model may generate predicted values out of the
interval zero and one, which violate the basic principles of probability. Secondly, the
coefficient of determination (R2) is likely to be much lower than one, and it is
questionable to use R2 as measure of goodness of fit (Gujarati, 2004). The third
problem with linear probability model is that the partial effect of any explanatory
variable is constant (Maddala, 1992).

The limitations of the linear probability model can be solved by applying either logit or
probit models or both (Amemiya, 1981). The two models generate predicted values
between 0 and 1, which follow the basic principles of probability. The main difference
between logit and probit is that the conditional probability approaches zero or one at a
slower rate in logit than in probit model (Pindyck and Rubinfeld, 1981 and Gujarati,
2004). Secondly, the error term in the logit model are assumed to follow the standard
logistic distribution, whereas in probit model error term is assumed to follow the
standard normal distribution. Thirdly, the probit model works well for bivariate models
than logit model. However, in most cases the two models are statistically similar (Park,
2008). This statistical similarity between the two models makes a choice of the models
depends on the availability and flexibility of computer program, personal preference
and experience (Ibid). Therefore, in this study probit model was used to determine the
factors that are affecting the WTP of households. Following Cameron and Quiggin
(1994), the probit model can be specified as:

(3)

Where: = vector of unknown parameters of the model

= vector of explanatory variables
= unobservable households‘ actual WTP for forest restoration
= discrete response of the respondents for the WTP
= the offered initial bids assigned arbitrarily to the ith respondents
= unobservable random component distributed

43
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

The bivariate probit model was used to estimate the mean WTP from the double
bounded dichotomous elicitation method. But, when the estimated correlation co-
efficient of the error terms in bivairate probit model are assumed to follow normal
distributions with zero mean and distinguishable from zero (see Equation 4 below), the
system of equations could be estimated as seemingly unrelated bivariate probit
(SUBVP) model (Haab and McConnell 2002). Therefore, in this study SUBVP was
employed to estimate the mean WTP. According to Greene (2003), a bivariate probit
model can be specified as:

(4)

Where: = ith respondent unobservable true WTP at the time of the first bid offered.

= ith respondent implicit underlying point estimate at the time of the second bid offered.

The first and second bids offered to the respondents respectively.

= Error terms for the first and second equations of Equation 4 above
= Coefficients of the first and second bids offered

The respondents know their own maximum WTP, but to the researcher it is a
random variable with a given cumulative distribution function (cdf) denoted by G ( ,
θ) where θ represents the parameters of this distribution, which are to be estimated on
the basis of the responses to the CV survey. The log-likelihood function for the
responses to a CV survey was specified as:

(5)

44
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Where if the ith response is (Yes, Yes) and 0 otherwise; if the ith
response is (Yes, No) and 0 otherwise; if the ith response is (No, Yes) and 0
otherwise; if the ith response is (No, No) and 0 otherwise.

4.4.2 The Censored Regression (Tobit) Model

The study employed Tobit model to investigate results of the open-ended question,
which is used as a second elicitation technique in the CV questionnaire of this study to
model the actual household‘s WTP for restoring forest resource. In probit and logit
model the dependent variable ( ) is not observed, what we observe is the dummy
variable. However, in Tobit model the dependent variable, or the WTP, is partially
observed and the dependent variable ( ) assumes zero values for a substantial part of
the sample. That is, is observed if and is not observed if . If y* and
xi were observed for everyone in the population, there would be nothing new, and we
could use standard regression methods (ordinary least squares (OLS)) (Maddala, 1992).
However, in this study since we deal with maximum WTP for forest restoration which
is partly observed, therefore, using OLS leads bias and hence, this study employed
Tobit model. In general, the censored regression (Tobit) models generally apply when
the variable to be explained is partly continuous. According to Maddala (1992) the
equation for Tobit model is specified as:

(6)

where: is latent or unobserved willingness to pay for forest restoration; is a

household‘s actual maximum willingness to pay for forest restoration; is vector of
explanatory variables; is a parameter vector common to all households; α is the
intercept; and assuming the random error is independent and normally distributed
across respondents, . Some of the households interviewed did not have
any WTP, whereas, some of them had WTP for restoring forest resource. For those
not undertaking WTP is zero in Tobit model the WTP is a random variable and has
probability distribution and it is possible to determine each observations probability.

(7)

45
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

Where p is probability distribution and F( ) is cumulative density function.

The model parameters can be estimated by maximizing the Tobit likelihood function
of the following form.

(8)

f and F are the density probability function and cumulative distribution function of ,
respectively. means that the product over those i for which , and
* *
y 0 y 0

means the product over those i for .

The Tobit coefficients do not directly give the marginal effects of the explanatory
variables on the dependent variable. But their signs show the direction of change in
probability of WTP as the respective explanatory variables changes. Therefore, it is not
reasonable to interpret in the same way as the one interprets coefficients in an
uncensored linear model (Johnston and Dinardo, 1997). Hence, we should estimate
the marginal effect of the Tobit model. Following Long (1997) and McDonald and
Maffitt (1980) the following techniques could be used to identify the effects of
explanatory variables on the probability of WTP and the amount of households‘ WTP
(the whole and willing respondents only).

The change in the probability of willingness to pay for forest restoration as explanatory
variables changes was estimated by:

9)

The marginal effect of an explanatory variable on the expected value of willingness to

pay was estimated by:

(10)

Similarly, the change in the probability of willingness to pay with respect to a change in
explanatory variable among willing respondents was estimated by:

(11)

46
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Where, , F(z) is the cumulative normal distribution of Z, f(z) is the value of the
derivative of the normal curve at a given point( that is, unit normal density), Z is the Z-
score for the area under the normal curve, is the vector of Tobit maximum
likelihood estimates and is the standard error of the error term.

5 Results and Discussion

5.1 Characteristics of Sample Households

A total of 393 sampled households were interviewed. However, ten respondents were
protested zero bidders. We compare the means of households‘ socio-economic
variables of protest and valid responses using independent samples T test and
crosstabs- chi square test. The mean of the households‘ socio-economic variables of
protest and valid responses was insignificant. Thus, only the valid response that is, 383
households were included in the analysis. Of the total 383 respondents, 54% were
males while 46% were female respondents. The age of these sampled respondents‘
ranges from 16 to 78 years with an average of 37.59 years old. The survey results also
showed that 71% of the respondents were married and the rest 29% were un-married.
A total number of 2731 family sizes were recorded with a minimum of 2 persons and a
maximum of 15 persons per households. On average, about 7 persons per household
were recorded which was above the national average of 4.7 persons (FDREPCC, 2008).
The result on the status of the respondents showed that 68% of the respondents were
head of the households, and the rest 32% were working member of the households.

Educational attainment is another parameter considered in our empirical models. The

educational status of the sampled respondents ranges from zero (illiterate) to 10+3
years of schooling with an average of about 6 years of schooling. The household survey
found that 13.58% of the respondents were illiterate, and most of the respondents
(64.23%) never went beyond elementary level. Among the sampled respondents, only
22.19% attained higher level of education beyond elementary school. The total farm
land holding of the sampled households was also estimated at 187.75 hectar with
average cultivated farm size per household of 0.49 ha. This indicated that the average
farm size of the study area is lower than the national average of 0.8 ha (CSA, 1995).
Moreover, the result shows that access to extension service of sample respondents.
Data with regard to access to extension service showed that 89% of the households‘
access to extension service whereas the remaining 11% were not.

47
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

5.2 Income Sources of Surveyed Households

Households in the study area were engaged in production of cash crop, annual crop
and rearing animals. They were also engaged in off farm activities (working in
governmental and nongovernmental organizations, trade). The total monthly income of
sampled households from the on farm and off farm activities was computed at
629,293.45 ETB with minimum 258 and maximum of 5850 ETB. On average the
income of the surveyed households were estimated 1643.06 ETB per month. Taking
the average family size of 7, the average per capital incomes was ETB 230.44 per
month. This is approximately four times higher than ETB 62.7 monthly average per
capital income reported by the IMF at country level (IMF, 2001). Higher average per
capita incomes in the study area could be due to production of cash crops like chat and
livestock production.

Major sources of income in the study area are from on farm activities, primarily from
production of cash crops and livestock production. The total monthly income of these
households from on farm activities was computed at 496,452.12 ETB. On the other
hand, the monthly income of the households obtained from off farm activities were
also computed at ETB 132,841.33. The fact that off-farm incomes contribute smallest
to the total family income, it explains that most of the surveyed household can rely
mainly on agricultural activities with relatively narrow landholding size for their
livelihood. It also indicates that large portion of the households engaged in on farm
income generating activities to meet their family livelihood needs. Data related to
livestock owned by the respondents was also collected in terms of TLU4. The survey
result shows that on average 1.95 TLU with a minimum of 1.56 and maximum of 3.87
was recorded per households.

5.3 Causes of forest lose and Its Protection Measures

In the study area vegetation is not found in contiguous form covering large area; rather
it is seen as fragmented patches of bush land, shrub land and trees in agricultural sites
and hillsides. The natural forest has been cleared to satisfy the demands of the ever
increasing population such as construction material, fuel wood, fodder and agricultural
expansion. The majority of the landmass is covered with shrub lands with the
remaining high forests of less than 1%. Therefore, household perception on the

4
Conversion factors used in estimation of tropical livestock unit (TLU) were Donkey = 0.7; Cow, Bulls
and Ox=1; Calf = 0.25; Sheep and Goats= 0.13; Chicken=0.013 and Camel = 1.25

48
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

problems of forest loses has a direct effect on households willingness to value the forest
restoration. Most of the respondents are aware about the availability of the resource.
About 82% of the respondents were known the goods to be valued properly. They have
an experience of using the resource for fire wood, charcoal, construction, shelter, feed,
and source of income.

Perceived respondents reported that the availability of forest resource is decreasing

from time to time, and the reasons attributed to the problem of forest lost were
population pressure, overgrazing, agricultural expansion, and soil and water
degradation. More specifically, 22.91% of the respondents frequently mentioned
population pressure as the first environmental problem followed by overgrazing and
soil and water degradations. On the other hand, about 18.28% of the respondents did
not perceive the problems of forest lose (See Table 1 below).

Table 1: Environmental problems of the study area listed by the respondents

Problems Number of households %
Population pressure 85 22.91
Agricultural expansion 65 16.97
Soil and water degradations 81 21.15
Overgrazing 82 21.41
None 70 18.28
Total 383 100%
Source: Survey result, 2012

Suggestions were also elicited from the aware respondents to overcome the problem of
forest lost. These include massive tree planting, strong government regulation, soil and
water conservation, and training forest users (see Table 2 below).

Table 2: Alternative Protection Measures

Protection Measures Frequency %
Massive tree planting 81 21.15
Soil and Water Conservation 75 19.58
Strong government regulation 79 20.63
Training forest users 78 20.37
None 70 18.28
Total 383 100%
Source: Survey data, 2012

49
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

5.4 Households Willingness to Pay for restoring forest resource

The results revealed that about 89% of the total 383 sample households were willing to
pay for forest restoration and their WTP is positive. The double bounded dichotomous
choice format was used to estimate the mean WTP from responses of both the first and
the second bids offered. The analysis was conducted using seemingly unrelated bivariate
probit model. The estimation result of model is reported in Table 3 below.

Table 3: Estimation results of the bivariate probit model

Variable Coef Z
Initial bids -0.0153904 -6.12 0.000
Constant 1.51082 8.37 0.000
Second bids -0.0215554 -9.8 0.000
Constant 1.978898 11.66 0.000
ρ*** 0.926 3.53 0.000
Number of obs 383
Log likelihood -398.86335
Wald chi2(2) 122.05
Prob > chi2 0.000
Likelihood-ratio test of rho=0:chi2(1)= 34.351 Prob>chi2=0.000
5
Mean WTP = 94.09 birr ( At 95% CI, 98.17 to 90 ETB)

The mean willingness to pay for forest restoration was computed at 94.09 ETB per
year per household for five years horizon. At 95% confidence interval the WTP varies
between 98.17 to 90 ETB (see Table 3 above). The result shows that the mean WTP
from double bound format was greater than the mean value from the open ended
response which was computed at 64.82 ETB per year per household for five years.
This may indicate the existence of anchoring effect from the double bounded
elicitation method. In other word, the open ended elicitation method has an advantage
to avoid the anchoring effect. This result is consistent with the various studies (Alemu,
2000; Köhlin, 2001; Carlsson et al., 2004 and Solomon, 2004).

5
The mean WTP from bivariate probit model was computed using the formula specified by Haab and
Mconnell (2002) ; is a coefficient for the constant term, and is a
coefficient for offered bids to the respondents

50
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

5.5 Determinants of Households‘ WTP

5.5.1 The probit model estimation result

The estimated result on factors affecting the households‘ WTP for forest restoration is
presented in Table 4 below. The sign of all the explanatory variables were as expected.
It could be seen that monthly income of the respondent shows positive and significant
relationship with the households‘ WTP. This positive effect indicated that respondents
with higher monthly income were more likely to say yes to the first bid than households
with lower income. This indicated that households with higher income have a greater
ability to pay than the households with lower income. A study by Alemu (2000) and
Tefera (2006) recognizes significant association between households‘ income and
WTP. The result also shows that keeping the influences of other factors constant at
their mean value, a one birr increase in income of the respondent, increase the
probability of accepting the first bid by about 0.02%.

The result of the probit model showed that education level of the respondents is
positively and significantly related to WTP. One possible reason could be that literate
individuals are more concerned about forest resource than illiterate one. Educated (or
literate) individuals relatively know more about the significance of resources, and they
are concerned more about the environmental resource. This is consistent with the
findings of Tegegne, (1999) and Carlsson et al. (2004). Furthermore, the result revealed
that keeping the influences of other factors constant, a one year increase in the
educational level of the households, the probability of accepting the first bid increase by
1.39%.

Total farm land holding was one of the hypothesized variables in the probit model, and
the effect of which on the WTP was estimated by the model. The sign of total farm
land holding turned out to be consistent with the prior expectation, and it was positively
and significantly related with the WTP. The significant result indicated that households
who have higher cultivated land were more likely to say yes response for the initial bid
than the respondents with small cultivated land. This is probably due to the fact that
larger farm size earns more income from crop production especially cash crops. The
result showed that an increase in family size increases probability of saying yes to the
offered prices by about 13.10%. Solomon, 2004 also found the same result.

Perception has an expected positive related to likelihood of saying yes to the first bid.
That is, households having higher awareness about the forest lose and environmental

51
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

problems are willing to pay positively. The coefficient of this variable was significant (at
5%), keeping other things constant changing the dummy from 0 to 1 will increase
probability of accepting the initial bid by about 18.16%. The coefficient of ownership
also has positive relation to the likelihood of saying yes to the initial bid. That is,
keeping the effect of other variables constant changing ownership from 0 to 1 will
increase probability of accepting the initial bid by about 14.61%. This is because the
respondents feel secure of their right to use the resource after plantation. This finding is
consistent with Solomon, 2004, and Lindhjem and Navrud, 2008. Besides, changing
access to extension from 0 to 1 would increase probability of accepting the initial bid by
about 80.26%. This is because access to extension is an important source of
information, knowledge and advice to smallholder farmers. Similar result was found by
Azami et al. 2012.

The coefficient of starting bid price has negative sign and significant at 5% level of
significance. The negative sign and the significance of this coefficient indicated that as
the starting bid price increases by one unit, the probability of household‘s willingness to
pay in birr reduces by 0.3%. This may indicate there is income scarcity or cash poverty.
Besides, the result shows that demand for forest restoration is decrease as price
increases. This is consistent with the findings of Solomon (2004); Carlsson et al. (2004);
Bin Ramlan et al. (2011); and Mousavi and Akbari, (2011).

52
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Table 4: The probit model estimation results of households‘ WTP

Dependent variable: Choice in the valuation question (no = 0, yes = 1); 383 observations
Explanatory Variables Coef. Marginal effects z
Households income 0.001 0.0002*** 4.13
Age of the respondents 0.007 0.0019 1.05
Educational level of the respondents 0.052 0.0139** 1.98
Sex of the respondents -0.101 -0.0270 -0.39
Ownership types 0.533 0.1461** 2.54
Respondents marital status 0.153 0.0422 0.53
Respondents status -0.368 -0.0928 -1.31
Total family size 0.028 0.0076 0.83
Tropical livestock unit 0.226 0.0606 1.56
Perception 0.588 0.1816** 2.33
Access to extension services 2.60 0.8026*** 5.07
Total farm land holding 0.488 0.1310** 2.01
Initial bids -0.011 -0.0030** -2.39
Constant -3.978 -4.63
Number of observation 383
LR chi2(13) 202.48
Prob>chi2 0.000
Pseudo R2 0.431
Log likelihood -133.648
*significant at 10%; ** significant at 5%; *** significant at 1% significance levels Source: own
survey

5.5.2 The Tobit model estimation result

The result from Tobit model was different from probit model in the significance level
and some variables. That is, 5 of the 13 explanatory variables were statistically
significant. The other 8 explanatory variables are insignificant effect on the amount of
WTP for restoring forest resource. Except for sex the significant variables have a
positive effect on the amount of WTP. However, the interpretation of the censored
regression model is not straightforward. That is, the marginal effects cannot be
adequately explained from the estimated coefficients of the Tobit model (see Table 5
below). Therefore, for interpretation of the Tobit model this paper report three sets of
marginal effects: the effect on the probability of a positive WTP, the effect on
conditional WTP, and the effect on unconditional WTP.

53
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

To be more specific, households‘ monthly incomes have positive and significant

association with the households WTP for restoring forest resource. That is, when the
income of the household increase by one birr, it would increase the probability of
willingness of a household to pay for forest restoration by about 0.001%. Besides,
when income of the household increase by one birr their willingness to pay would
increase, on average, by about 0.005 ETB for all observation and 0.0047 ETB for
willing respondents‘, ceteris paribus. This shows that forest restoration is a normal
economic good whose demand changes in the direction of income change.
Respondents with higher education levels were more likely to state positive WTP, and
on average, they actually stated higher conditional and unconditional WTP than
respondents with lower educational levels. This result suggests that investing in
education of people might help to restore forest resource in degraded environment.
The marginal effect of the result shows that the respondent being educated, the
probability of willingness to pay for forest restoration increases by 0.22%. Also, as the
years of education increases by one year, the amount of cash the household is willing to
pay for forest restoration increase by 1.024 birr for the whole sample of study, and 0.91
birr for the willing respondents, ceteris paribus. The variables ownership type and
access to extension services also have positive and significant effect on the amount of
WTP. In terms of ownership type a unit changes from 0(government ownership) to 1
(community ownership) the probability being willing to pay increases by 28.53%. That
is, the marginal effect result shows that a unit changes from 0(government ownership)
to 1 (community ownership), the willingness to pay increased by 12.5 birr and 11.04
birr for the whole and willing respondents respectively, ceteris paribus. This is because
the households may feel secure of their right to use the resource after restoration.
Access to extension service was another variable found to be significant. Since the
parameter estimate is positive, it implies that households having access to extension
service increases the probability of WTP by 36.07%. This enhanced the awareness of
the respondents on the forest lost and restoration. The male respondents would have a
2.4% probability less than a female to pay for forest restoration. That is, female
respondents were willing to pay approximately 10 birr more than the males, ceteris
paribus.

54
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Table 5: The Tobit model estimation results of households‘ WTP

Dependent variable: Maximum Willingness to Pay; 383 observations
Marginal effects
Explanatory Variables Coef. t-value
A B c
Households income 0.006*** 3.96 0.000011 0.0047 0.005
Age of the respondents 0.146 1.03 0.00031 0.1252 0.141
Educational level of the respondents 1.058** 2.02 0.0022 0.9079 1.024
Sex of the respondents -11.62** -2.12 -0.0239 -9.997 -11.25
Ownership types 12.94*** 3.27 0.2853 11.04 12.50
Respondents‘ marital status 4.236 0.76 0.0093 3.613 4.094
Status of the respondents -1.671 -0.29 -0.0034 -1.437 -1.62
Total family size -0.261 -0.4 -0.00055 -0.2237 -0.252
Tropical livestock unit -0.201 -0.07 -0.00042 -0.1727 -0.195
Perception 3.039 0.56 0.0067 2.590 2.937
Access to extension service 59.08*** 8.25 0.3607 39.91 50.32
Total farm land holding 6.422 1.64 0.1344 5.513 6.22
Initial bids -0.600 -7.06 0.0012 0.515 0.5812
Constant -53.32 -3.87
( for all sample respondents)
*significant at 10%; ** significant at 5%; *** significant at 1% level of significance
a
Marginal effects on the probability of being censored
b
Marginal effects on the truncated expected value(for the willing respondents only)
c
Marginal Effects on the censored expected value
Source: survey result

5.6 Aggregate WTP for Restoring Forest Resource

An important issue related to the measurement of welfare using WTP is aggregation of

benefit (Alemu, 2000). According to Mitchell and Carson (1989) there are four
important issues to be considered regarding sample design and estimating a valid
aggregation of benefits: population choice bias, sampling frame bias, none response
bias and sample selection bias. Random sampling method was used in this study using a
list of households. Face to face interview methods was used and protest zero responses
were excluded from the analysis and expected protest zeros was accounted in the
estimation of the total aggregate benefit of forest restoration in this paper. Hence, none
of the above biases was expected in this paper. Mean WTP was used as a measure of
aggregate value of forest restoration in this study. The mean is perhaps better than the
median since the good dealt with is not a pure public good (Alemu, 2000), as there are

55
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

purely private benefits from forest restoration measures. As it is indicated in Table 6

below, the aggregate WTP was calculated by multiplying the mean WTP by the total
number of households in the population. Following this, the aggregate WTP for
restoring forest resource was computed at 2,026,604.51 birr per year for five years.
Whereas, from open ended questions the total WTP for restoring forest resource was
also computed at 1,396,157.98 birr per year.

Table 6: Estimation of Total Aggregate Benefits of forest restoration

Total Expected households Expected households Aggregate
Mean
households to have a protest with valid responses 8 Benefit
6
WTP
(Y) zeros (X) (Z) 7
(in Birr)9

22,091 552 21539 94.09 2,026,604.51

22,091 552 21539 64.82 1,396,157.98
Source: own survey, 2012

6 Conclusion and Recommendations

The purpose of this study was to assess the economic value of forest restoration in Dire
Dawa area using CVM. The descriptive analysis shows that 82% of the respondents
reported that the reasons attributed to the forest lost were population pressure,
overgrazing, soil and water degradation and agricultural expansion. In order to restore
the forest resource policy makers should encourage and provide technical advice to
households who are planting and maintaining tree resource, and practicing soil and
water conservation.

The results of the study on willingness to pay showed that the households were willing
to pay for restoring forest resource. The annual mean WTP value of households for
restoring forest resource based on the double bounded dichotomies choice was
computed at 94.09 birr per year for five years. Whereas, the annual total WTP from
open ended format was also computed at 64.82 birr per year. The study found that the

6
10(2.5%) of 393 sampled households were protest zeros. We excluded those protest zeros from further
analysis after we have tested for sample selection bias. So X is the expected number of households which
are expected to protest for the proposed project. It is calculated by the percentage of sampled protest zeros
(2.5%) by the total population 22091 (Y).
7
Y-X is the total households in the study area which are expected to have a valid response
8
Is the mean willingness to pay calculated from the double bounded dichotomous choice estimation and
open ended elicitation methods
9
Is mean multiplied by the number of total households which are expected to have valid response
(Z*Mean WTP) measured in ETB

56
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

value of forest restoration from open ended format was lower than double bounded
elicitation format significantly. The study conclude that there could be a problem in
applying a closed-ended elicitation format to forest restoration in developing countries
since a forest restoration typically implies a community based scenario and such a
scenario invites to anchoring effect. The well-known problem of anchoring effect
and/or compliance bias is also difficult to avoid in such settings. Application of a
closed-ended format under such circumstances would exaggerate the WTP for forest
restoration.

The empirical findings on the determinants of WTP indicated that monthly income,
access to extension service, ownership types and initial bids are key factors influencing
the willingness to pay in both probit and Tobit models. Besides, the study estimated
that there are statistically significant and quantitatively non-negligible effects of
perception and total farm land holding on the households‘ WTP in probit model. The
variable sex is also significant effects from Tobit model. Such differences are worth
considering when planning households contribution in projects for restoring forest
resource. Generally, the study leads us to conclude that understanding of socio-
economic characteristics that significantly influenced households WTP is a necessary
and first step to achieve restoring forest resource. Moreover, specific factors that
affected the valuation of a household for restoring forest resource in open ended
elicitation method should be analyzed separately from the valuation of households for
restoring forest resource in dichotomous choice elicitation method in order to restore
the forest resource.

57
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

References

Agudelo, J. I. (2001). The Economic Valuation of Water: Principles and Methods. Value of
water, Research Report Series No. 5. IHE DELFT, the Netherlands.
Aldrich, J. H. and F. D. Nelson. (1984). Linear Probability, Logit and Probit Models. Stage
Publications, London.
Alemu Mekonnen. (2000). Valuation of Community Forestry in Ethiopia: A Contingent
Valuation Study of Rural Households. Environment and Development Economics. 5:
289-308pp.
Amemiya, T. (1981). Qualitative Response Models: A Survey. Journal of Economic Literature
19:1483-1536pp.
Arrow, K., R., Solow, P. R., Portney, E. E., Leamer, R., Radner, H., Schuman. (1993). Report
of the NOAA Panel on Contingent Valuation. Federal Register. 58:4601-4614pp.
Azami M., S. A., Ghadimi, H., Fathi, R., Movahedi, N., Dowlati. (2012). The Role of
Education in People's Willingness to Pay for Conservation of Oak Forests in
Kurdistan. International Research Journal of Applied and Basic Sciences. 3(4): 729-
736pp.
Azene Bekele-Tesemma. (2002). Forest Landscape Restoration: Initiatives in Ethiopia,
Compiled by IUCN-EARO and WWF-EARPO.
Bennett, J. W. and Carter, M. (1993). Prospects for Contingent Valuation: Lesson from the
Southeast Forest. Australian Journal of Agricultural Economics, 37(2):79-93pp.
Bin Ramlan M. A., A., Radam, M. R. Yacob, N. A. Yahya. (2011). Willingness to Pay Towards
the Sustainability of Forest Research Institute Malaysia‘s (FRIM‘s) Canopy Walkway.
International Journal of Business, Management and Social Sciences. 2(3): 85-92pp.
Bishop J. T. (ed.) (1999).Valuing Forests: A Review of Methods and Applications in Developing
Countries. International Institute for Environment and Development (IIED), 3
Endsleigh Street, London WC1H ODD, UK.
Cameron, T. A. and Quiggin, J. (1994). Estimation using contingent valuation data from a
‗dichotomous choice with follow-up‘ questionnaire. Journal of Environmental
Economics and Management. 27(3):218-34pp.
Carlsson F., G., Köhlin, and Alemu M. (2004). Contingent Valuation of Community Plantations
in Ethiopia: A Look into Value Elicitation Formats and Intra-Household Preference
Variations. Working Papers in Economics no. 151, November 2004, Gothenburg
University.
Cavatassi R. (2004). Valuation Methods for Environmental Benefits in Forestry and Watershed
Investment Projects. ESA Working Paper No. 04-01.
Central Statistical Agency (CSA). (1995). Report on Land Utilization, statistical bulletin 132,
Addis Ababa, Central Statistical Authority, Addis Ababa (CSA).

58
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Chaudhury, P. (2006). Valuing Recreational Benefits of Urban Forestry: A Case Study of

Chandigarh city, PhD thesis, Forest Research Institute (Deemed University),
Dehradun.
Chukwuone, N. A. and C. E. Okorji. (2008). Willingness to Pay for Systematic Management of
Community Forests for Conservation of Non-Timber Forest Products in Nigeria‘s
Rainforest Region: Implications for Poverty Alleviation. Rob B. Dellink and Arjan
Ruijs (eds.), Economics of Poverty, Environment and Natural- Resource Use, 117-
137pp.
Dogru M. (2001). Planning and Management of Forest Resources in Turkey (Draft), Assistance
for the Preparation of a National Programme for Turkey. Earthscan. London.
FAO. (2005). Global Forest Resources Assessment. Progress towards Sustainable Forest
Management.
______. (2006) Choosing a forest definition for the Clean Development Mechanism: Forest and
Climate Change Working Paper 4, Rome, Italy (http//: www.
fao.org/forestry/11280‐1‐0.pdf, accessed date August 7 2012).
______. (2010) Global Forest Resources Assessment 2010 ‐ Country Report Ethiopia. Food
and Agriculture Organisation (FAO), Rome, Italy (www. fao.org/forestry/
fra/fra2010/en/, accessed date July 15, 2012).
Federal Democratic Republic of Ethiopia (FDRE). (2001). Regional Conservation Strategy. The
Resource Base, Its Utilization and Planning for Sustainability. Vol. 1. Dire Dawa,
Ethiopia.
Federal Democratic Republic of Ethiopia Population Census Commission (FDREPCC). (2008).
Summary and Statistical Report of the 2007 Population and Housing Census Results.
December 2008, Addis Ababa.
Garrod, G. D. and K. G. Willis. (1997). The Non-Use Benefits of Enhancing Forest
Biodiversity: A Contingent Ranking Study. Ecological Economics 21: 45-61pp.
Greene, W. H. (2003). Econometric Analysis. 5 edition. Pearson education, Inc., Upper
th

Saddle River, New Jersey, 07458.

Gujarati, D. N. (2004). Basic Econometrics. 4 Edition. McGraw-Hill, Inc.
rd

Haab, T. C. and K. E. McConnell. (2002). Valuing Environmental and Natural Resources, the
Econometrics of Non Market Valuation. Edward Elgar, Cheltenham U.K.
Hanemann, M., J., Loomis and B., Kaninen. (1991). Statistical Efficiency of Double Bounded
Dichotomous Choice Contingent Valuation. American Journal of Agricultural
Economics, Vol. 73, No. 4. (Nov., 1991), 1255-1263 pp.
International Monetary Fund (IMF). (2001). International Financial Statistics. Washington, DC
20431, USA.
ITTO. (2002). ITTO Guidelines for the Restoration, Management And Rehabilitation Of
Degraded And Secondary Tropical Forests. ITTO Policy Development Series.
Johnston J, and Dindaro, J. (1997). Econometric Methods. 4th edition. New York.
Köhlin, G. (2001). Contingent Valuation in Project Planning and Evaluation: The Case of Social
Forestry in Orissa, India, Environment and Development Economics, 6: 237-258.

59
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

Kramer, R., R. Healy, and R. Mendelsohn. (1997). Forest Valuation. Chapter 10. In Managing
the World‘s Forests, Narendra Sharma (ed). World Bank Natural Resources
Development Series, Arlington, VA.
Lindhjem, H. and S., Navrud. (2008). Asking for Individual or Household Willingness to Pay
for Environmental Goods? Implication for aggregate welfare measures, Norway.
Long S. (1997). Regression Models for Categorical and limited dependent Variable.
McDonald J. F. and Maffitt A. (1980). The Use of Tobit Analysis. Rev. Econ. Stat., 62(2): 318-
312.
Maddala, G. S. (1992). Introduction to Econometrics. 2 edition. Macmillan, Inc, University of
nd

Florida and Ohio state university, New York.

Maginnis S. and W., Jackson. (2003). The Role of Planted Forests in Forest Landscape
Restoration. UNFF Intersessional Experts Meeting on the Role of Planted Forests in
Sustainable Forest Management New Zealand: 25–27pp.
Markandya A., P. Harou, L. G. Bellu, and V. Cistulli (in association with the international bank
for reconstruction and development/ the World Bank). (2002). Environmental
Economics for Sustainable Growth: A Handbook for Practitioners. Edward Elgar
Publishing, Inc. Cheltenham, UK. Northampton, MA, USA.
Mitchell, R. C., Carson, R. T. (1989). Using Surveys to Value Public Goods: The Contingent
Valuation Method. Resources for the Future, Washington, 463 p.
Mousavi , S. N. and S. M. R. Akbari. (2011). Estimated value of forest conservation in Iran: A
case study of Fars Province. African Journal of Agricultural Research Vol. 6(30), pp.
6407-6411, 12 December, 2011.
Pak M., M. F., Türker and A., Öztürk. (2010). Total Economic Value of Forest Resources in
Turkey. African Journal of Agricultural Research. 5(15):1908-1916pp.
Park, H. M. (2008). Estimating Regression Models for Categorical Dependent Variables Using
SAS, STATA, LIMDEP, and SPSS. The Trustees of Indiana University.
Pearce D. W. (2002). Economic Valuation with Stated Preference Techniques: Queen‘s Printer
and Controller of Her Majesty‘s Stationery Office.
Pindyck, R. S. and D. C. Rubinfeld. (1981). Econometric Models and Econometric Forecasts.
nd
2 Edition, Mcgraw-Hill Book Co. New York.
Solomon Jebessa. (2004). Contingent Valuation of Multi-Purpose Tree Resources: The Case of
Arsi Zone, Ethiopia. A Thesis Submitted To the School Of Graduate Studies of Addis
Ababa University, Ethiopia
Tefera Mengistu. (2006). Frontier Community Valuation for Forest Patches: the Case of
Wondo-Wosha Subcatchment, Southern Nations, Nationalities and Peoples‘ Region,
Ethiopia. Ethiopian Journal of Natural Resources. 8 (2): 281-293pp.
Tegegne GebreEgziabher. (1999). Willingness to Pay for Environmental Protection: An
Application of Contingent Valuation Method (CVM) in Sekota District, Northern
Ethiopia. Ethiopian Journal of Agricultural Economics, 3(1).

60
 Ethiopian Journal of Economics, Vol XXI No 2, October 2012

Turner, K., S. Georgiou, R. Clark, R. Brouwer, and J. Burke. (2004). Economic Valuation of
Water Resources in Agriculture: From the Sectoral to a Functional Perspective of
Natural Resource Management, FAO 2004, Rome.
Venkatachalam, L. (2004). The Contingent Valuation Method: A Review. Environmental
Impact Assessment Review, 24: 89–124pp.
Verma, M. (2000). Economic Valuation of Forests of Himachal Pradesh‘, (mimeo), Indian
Institute of Forest Management, Bhopal.
Whittington, D. (1998). Administering Contingent Valuation Surveys in Developing Countries.
World Development. 26(1): 21-30pp.
Whittington, D., J. Briscoe, X. Mu and W. Barron. (1990). Estimating the Willingness to Pay
for Water Services in Developing Countries: A Case Study of the Use of Contingent
Valuation Surveys in Southern Haiti. Economic Development and Cultural Change,
38(2): 293–311pp.
Wu S., Y. Hou and G. Yuan. (2010). Valuation of Forest Ecosystem Goods and Services and
Forest Natural Capital of the Beijing Municipality, China. Unasylva 234/235, Vol. 61.
WWF and IUCN. (2001). Forest Landscape Restoration. Tanzania Country Report.

61
Alem Mezgebo: Households’ willingness to pay for restoring environmental…

62