Pastoralism

Research, Policy and Practice

Pastoralism Cover Image
Open Access

Impacts of Prosopis juliflora invasion on livelihoods of pastoral and agro-pastoral households of Dire Dawa Administration, Ethiopia

  • Negussie Zeray1Email author,
  • Belaineh Legesse2,
  • Jema Haji Mohamed2 and
  • Mengistu Ketema Aredo2, 3
PastoralismResearch, Policy and Practice20177:7

DOI: 10.1186/s13570-017-0079-z

Received: 18 July 2016

Accepted: 7 February 2017

Published: 12 April 2017

Abstract

This paper examines impacts of Prosopis juliflora (hereinafter may be referred as ‘Prosopis’) invasion on livelihoods of agro (pastoral) households using detailed household data in rural Dire Dawa Administration of Ethiopia. Cross-sectional data were collected from a total of 450 sample households whereby 250 were from Prosopis-invaded households and the remaining 200 from non-Prosopis-invaded households. The major research question of the study was, ‘what would be the livelihoods of Prosopis-invaded households had they not been invaded by Prosopis?’ To answer this question, descriptive and econometric tools were employed. The study results revealed that family size, dependency ratio and access to irrigation had negative and significant relationships with Prosopis invasion. On the other hand, age of the household head, Tropical Livestock Unit (TLU) and engagement in food for work programmes were positively related to Prosopis invasion. The results of this study show that the positive effects of Prosopis invasion were pronounced in terms of education expenditure, average annual income from crop production, off-farm income, food and non-food expenditure and physical food consumption, whereas there has been negative effects of Prosopis invasion due to reduced income generation from milk products. In this study, possible recommendations were drawn for stakeholders in order to reverse Prosopis adverse effects and optimize the positive impacts of this species in the study area.

Keywords

Pastoral Agro-pastoral Prosopis juliflora Propensity score matching

Introduction

Ethiopia’s natural resource base is the foundation of economic development, food security and other basic necessities of its people. Agriculture is the dominant sector of the country’s economy that provides nearly 42% of GDP, 77% of employment, and 84% of exports (ATA 2015). In addition, majority of the agriculture sector consists of smallholder farmers who make their living from less than 2 ha of land. The sector is also a front-line victim of different environmental shocks including invasion of noxious plant species (FDRE, 2010). According to Stefan (2005), the impacts of invasive species on livelihoods are more pronounced in developing nations as the majority of their populations consist of small-scale farmers who are dependent on natural resources for their survival.

Invasive alien species out-compete native organisms for food and habitat, spread through their new environment, increase their population and harm ecosystems in their introduced range (Liba and Gretchen 2009). They cause ecological, economic and social impacts and are key drivers of global change. They inflict serious damage on ecosystem goods and services which are the lifeline of the human enterprise (Ross et al. 2014).

The Federal Government of Ethiopia has identified a number of major invasive plant species in the country and declared the need for their control and eradication (Taye et al. 2007). These include Parthenium (Parthenium hysterophorus), water hyacinth (Eichhornia crassipes), mesquite (Prosopis juliflora) and Lantana camara and Acacia species, such as Acacia drepanolobium and Acacia mellifera. The Environment Policy of Ethiopia, the Forest Resource Strategy and the National Biodiversity Strategy and Action Plan recognize invasive plant species to be growing threats to the biodiversity of the country and socio-economic welfare of the people (Anagae et al. 2004). At the regional level, however, there is a lack of clearly defined policy or strategy for the control and management of invasive species (Anagae et al. 2004; Dubale 2008) and little attempt has been made in terms of their research and management. This is the same for the Dire Dawa Administration.

Prosopis is one of the world’s worst invasive alien species, causing severe environmental degradation to the arid and semi-arid lowlands of the Horn of Africa, andthreatening the livelihood and food security of pastoral and agro-pastoral communities (Tilahun et al. 2014). Prosopis has invaded most of the pastoral and agro-pastoral areas of Afar Regional State and Dire Dawa Administration. In Ethiopia, Prosopis has covered an area of one million hectares and more than 12,000 ha in Dire Dawa Administration (BoARD 2009). In the Middle Awash, about 30,000 ha of grassland, rangelands, water points and croplands are estimated to be occupied by Prosopis (Mehari 2008). Prosopis is now causing concern in Ethiopia. It has become a serious challenge to pastoralists and agro-pastoralists, the result of which it is considered as a burning issue for regional government and city administration, researchers, academia, development practitioners and policy-makers at various levels. Therefore, the primary objective of this paper is to analyse the impact of Prosopis invasion on livelihoods of pastoral and agro-pastoral households of the rural Dire Dawa Administration, Ethiopia. This research attempts to answer the following question: What would be the livelihood of Prosopis-invaded households had they not been invaded?

Study area

The study was conducted in the west of Dire Dawa city, and it covers a total area of 26,437 ha, which is nearly 20% of the area under the Administration. The study area has an estimated population of 17,800, nearly 16.5% of the rural population in the Administration. Agriculture (both crop and livestock production) is the mainstay of the economy in the study area. Subsistence mixed farming is practised by 93% of the farm households. Selling of firewood is the most common non-farm income-generating activity in the study area. Dire Dawa Administration is found in the eastern part of Ethiopia within the Awash Basin at about 515 km away from Addis Ababa and 333 km from the international port of Djibouti. The Administration is geographically located between 09° 25′ to 09° 50′ N latitude, and 41° 35′ to 42° 25′ E longitude (WWDSE 2003).

The climatic condition of Dire Dawa seems to be greatly influenced by its topography, which lies between 894 and 2,284 m above sea level and characterized by a warm and dry climate with a relatively low level of precipitation. The mean annual temperature of Dire Dawa is about 25.4 °C. The average maximum temperature of Dire Dawa is 31.4 °C, while its average minimum temperature is about 18.2 °C. The region has two rain seasons, that is, a small rain season from March to April and a big rain season that extends from August to September. The aggregate average annual rainfall that the region gets from these two seasons is about 604 mm. On the other hand, the region is believed to have an abundant underground water resource (IDP 2006).

Methods

Sampling and data collection methods

The study employed a multi-stage sampling procedure to collect primary data. Firstly, Dire Dawa Administration was chosen purposively since it is the second most Prosopis-infested area in Ethiopia next to the Afar National Regional States. In the second stage, among 38 rural Kebele 1 in the Administration, a census-type selection method was employed to select all the four Prosopis-invaded Kebeles and a purposive sampling method was used to select three non-invaded Kebeles based on their demographic, socio-economic and geographic similarity with the invaded Kebeles. Thirdly, 250 and 200 agro-pastoral households were selected from invaded Kebeles and non-invaded Kebeles, respectively, applying probability proportionate to size (PPS) technique to have a total of reasonable comparison with 450 agro-pastoral households. Data was collected using pre-tested survey questionnaire. The survey questionnaire constitutes different parts from which respondents were asked about their socio-demographic and economic profiles and characteristics. The survey was administered using experienced enumerators who could speak the local language. In addition, published and unpublished research reports from various governmental and non-governmental organizations were also reviewed.

Data analysis

The empirical data were analysed using descriptive, inferential and econometric tools. The descriptive statistics which were used for analysing data for this study include mean, standard deviation, percentages and tables. Besides, t test and chi-square test were used to test whether there is a statistically significant difference between the two groups (invaded and non-invaded households) in terms of continuous and categorical variables, respectively.

Impact estimation techniques

Impact estimation was conducted using an econometric model called propensity score matching (PSM) approach. We used the PSM developed by Rosembaum and Rubin (1983) to assess if invasion by Prosopis has a significant negative or positive effect on the livelihood of the invaded communities. PSM uses a statistical model to calculate propensity of invasion on the basis of the set of observable characteristics. Invaded and non-invaded households are then matched on the basis of similar propensity scores. The idea behind the PSM approach is to find control observations (that is, non-invaded households) having initial observable characteristics similar to the invaded households, to serve as valid surrogates for the missing counterfactuals.

This involves estimating a logit model that predicts the probability that each household is invaded as a function of observed household and community characteristics using a sample of invaded and non-invaded households. The model specification is checked to test equality of the means of these observed characteristics across the invaded (treatment sample) and non-invaded (control sample).

In estimating the logit model, the dependent variable is invasion by Prosopis, which takes the value of 1 if a household is invaded and 0 otherwise. The mathematical formulation of the logit model is as follows:
$$ {p}_i\kern0.5em =\kern0.5em \frac{e^{z_i^i}}{1\kern0.5em +\kern0.5em {e}^{z_{i_i}}} $$
(1)
where p i is the probability that the ‘ith’ households is invaded by Prosopis, Z i is a linear function of m explanatory variables (X) and is expressed as
$$ {Z}_i={a}_o+{a}_1{x}_{1 i}+{a}_2{x}_{2 i}+{a}_3{x}_{3 i}+\cdots +{a}_m{x}_{m i}+{U}_i $$
(2)

where

i = 1, 2, 3… m

a o  = intercept

a i  = regression coefficients to be estimated

U i  = a disturbance term, and

X i  = pre-intervention characteristics.

The probability that a household belongs to the non-invaded group:
$$ 1-{p}_i\kern0.5em =\kern0.5em \frac{1}{1\kern0.5em +\kern0.5em {e}^{z_{i_i}}} $$
(3)

According to matching theory (Rosembaum and Rubin 1983; Bryson et al. 2002; Jalan and Ravallion 2003), the logit model via which the propensity score is generated should include predictor variables that influence the selection procedure and the outcome of interest. In this respect, the study was undertaken by including as many explanatory variables as possible to minimize the problem of unobservable characteristics in our evaluation of the impact of Prosopis invasion.

This approach assumes that after controlling for all pre-invasion observable household and community characteristics that are correlated with invasion and the outcome variables, non-invaded households have the same average outcome as invaded households would have had if they were not invaded. If not feasible to control for these characteristics, PSM estimation becomes biased. Having control households from the same communities as treatment helps to reduce the risks of such bias by providing a similar distribution of unobserved community characteristics.

It is also assumed that for each invaded household and for all observable characteristics, a comparison group of non-invaded household with similar propensity scores exists. Heckman et al. (1998) emphasize that the quality of the match can be improved by ensuring that matches are formed only where the distribution of the density of the propensity scores overlap between treatment and comparison observations or where the propensity score densities have ‘common support’. Common support is improved by the dropping of all observations whose propensity scores are smaller than the minimum and larger than the maximum of the treatment and control, respectively (Caliendo and Kopeinig 2005).

The balancing property of the logit specification is tested to ensure that the sample of invaded households and the sample of non-invaded households have similar mean propensity scores and observables at various levels of propensity scores (Becker and Ichino 2002). Hence, the results are presented based on specifications that passed the balancing tests. Related to the balancing property of p-score is the conditional independence assumption (CIA), which states that the existence of Prosopis is random and uncorrelated with household income, once the set of observable characteristics, X, are controlled for. Sensitivity analysis was also undertaken to check if the influence of unobserved variables on the selection process is so strong as to undermine the matching procedure.

The general formula for the mean impact of Prosopis invasion on livelihoods of pastoral and agro-pastoral households is given by
$$ \varDelta C\kern0.5em =\kern0.5em \frac{1}{N_T}{\displaystyle \sum_{i\varepsilon \left\{ D=1\right\}}\left[{y}_i,\kern0.5em i-{\displaystyle \sum_j w\left( i,\kern0.5em j\right){y}_0,\kern0.5em j}\right]} $$
(4)

where 0 < w(i, j) ≤ 1, {D = 1} is the set of treated individual, j is an element of the set of matched comparison units, N T is the number in the treated group and i is the treated individual. Different matching estimators are generated by varying the choice of w(i, j).

Variable definitions, measurement and working hypothesis

Variable definitions, measurement and working hypothesis of this study are found in Table 1.
Table 1

Variable definitions and measurement

Variable

Type and definition

Measurement

Dependent variable

 TREATMENT

Dummy, invasion by Prosopis

1 for invaded households, 0 otherwise

Outcome variables

 EDUEXPE

Continuous, education expenditure

Birra

 FOaNFEXPE

Continuous, food and non-food expenditure

Birr

 HEALEXPE

Health expenditure

Birr

 CALINTAKE

Continuous, calorie per day per AE

Cal

 AAILS

Continuous, average annual income from livestock sale

Birr

 AAIMP

Continuous, average annual income from milk production

Birr

 AAICP

Continuous, average annual income from crop production

Birr

 AAFW/CP

Continuous, average annual income from fuel wood/charcoal production

Birr

 OFFFARMIN

Continuous, average annual income from off-farm activities

Birr

Explanatory variables

Hypothesis

 AGEHH

Continuous, age of household head

Number of years

(+)

 EDUHH

Dummy, education of household head

1 if illiterate, 0 otherwise

(+)

 TOTHH

Continuous, family size

Number of household members

(−)

 TOTALAND

Continuous, landholding size

Ha

(−)

 IRRGTN

Dummy for access to irrigation

1 if has access to irrigation, 0 if not

(+/−)

 SEXHH

Dummy for sex of household head

1 if male, 0 otherwise

(+)

 DEPENRATIO

Continuous, dependency ratio

Ratio

(−)

 TLU

Continuous, livestock-holding size

Tropical livestock units

(−)

 CREDITACC

Dummy for access to rural credit service

1 if access to credit service, 0 if not

(+/−)

 FARMEXPHH

Continuous

Number of years

(−)

 SAFTYNT

Dummy for access to safety net programme

1 if a household head is engaged in safety net programme, 0 if not

(+)

 DRTEF

Dummy for effect of drought

1 if a household’s income from livestock and crop production decreases when compared to the normal year and 0 if it remains the same

(+/−)

 DISROAD

Continuous, distance from main road

Km

(−)

Source: own definition

1USD = 19.25 Ethiopian Birr

aThe unit of currency in Ethiopia

Result

Descriptive results

As already noted, this study is based on data collected from a total of 450 sample households, of which 250 households belong to the four Kebeles invaded by Prosopis and the remaining 200 households belong to the three non-invaded rural Kebeles of the administration. As indicated in Table 2, the average age of respondents in Prosopis-invaded Kebeles was 42.49 years, whereas it is 38.58 years for the non-invaded Kebeles. Average family size of the households in Prosopis-invaded Kebeles was 6.302 persons or 5.578 in adult equivalent while that of the non-invaded households was 6.270 persons or 5.407 in adult equivalent. t test results for differences in age distribution was significant with a t value of −3.694, and family size between the two groups was insignificant (t = −0.135).
Table 2

Socio-economic characteristics of sample households by household type

Pre-interv. variables

Invaded N = (250)

Non-invaded N = (200)

Sample household N = (450)

Means difference

t value

Mean

STD

Mean

STD

Mean

STD

Mean

STD

AGEHH

42.490

10.208

38.58

11.004

40.545

10.777

3.91

−0.796

−3.694***

Family size age between 15 and 64

3.411

1.802

2.970

1.756

3.192

1.791

0.441

0.047

−2.484**

DISROAD

1.840

2.421

2.475

4.027

2.155

3.330

−0.635

−1.606

1.918*

DISMKT

3.824

3.872

13.736

5.618

8.756

6.914

−9.912

−1.746

20.614***

FARMEXPHH

26.109

10.902

20.135

8.210

23.137

10.098

5.974

2.692

−6.202***

TOTHH

6.302

2.496

6.270

2.254

6.286

2.376

0.032

0.242

−0.135NS

DEPENRATIO

1.030

0.072

1.056

0.075

−0.026

−0.003

1.288

1.051

2.678***

HH size in AE

5.578

2.152

5.407

1.894

5.493

2.027

0.171

0.258

−0.843NS

TOTALAND

2.621

2.369

2.148

2.211

5.493

2.027

0.474

0.158

−2.073**

TLU

5.033

3.214

4.610

3.271

4.820

3.246

0.423

0.057

1.308NS

Source: own survey data, 2015

*, ** and *** mean significant at the 10%, 5% and 1% probability levels, respectively

Not significant (NS)

Proximity of households to market centre, the city and main road of Dire Dawa were also assessed, and the result showed that the average distances of the invaded and non-invaded households from the city were 3.824 and 13.736 km, respectively, while their corresponding distances from the main road were 1.840 and 2.475 km. t test results showed that there is a significant difference at 1% and 10% level of significance (t = 20.614 and t = 1.918), in proximity of households to market centre and main road, between the two groups, respectively (Table 2).

A closer look at economically active family members in the sampled households (15 to 65 years old) showed that the treatment group of households (invaded households) had relatively larger active family members, accounting to about 3.411 persons, than the control group of households which was found to be about 2.790 persons. This implies that, on average, the households from the invaded Kebeles have relatively more labour force than those households from the non-invaded Kebeles. However, the treatment group of households showed a slightly lower average dependency ratio (1.030) than households from the control group (1.056). This implies that every economically active person in the Prosopis invaded households and non-invaded households had to support more than one economically inactive person.

Farm experience of the sampled households was also assessed, and the result showed that, on average, Prosopis-invaded households had 26.109 years of farm experience while non-invaded had an average of 20.135 years farm experience. t test results, as showed in Table 2, indicated that there is a significant difference in farm experience of household heads at 1% level of significance (t = −6.202) between the two groups.

Livestock holding, as a wealth variable, indicates the capacity of agro-pastoral household to involve in high-return income sources. As shown in the table below, livestock holding in TLU was 5.033 for the treatment group and 4.610 for the control group. This implies that livestock per capita in the invaded areas was higher than that in the non-invaded areas, even if the t test result presented in Table 3 shows the difference is insignificant (t = 1.308).
Table 3

Farm holding size of sample households by household type

Land size (ha)

Non-invaded

Invaded

Total

Number

%

Number

%

Number

%

No farm land

2

0.4

6

1.3

8

1.7

<0.5

14

3.1

11

2.4

25

5.6

=0.5

49

10.9

50

11.1

99

22

0.5 < x < 1

44

9.8

45

10.0

89

19.8

=1

46

10.2

47

10.4

93

20.7

>1

45

10.0

91

20.2

136

30.2

Total

200

44.4

250

55.6

450

100.0

Source: own survey data, 2015

As is the case in the majority of farm households in the country, land is one of the scarce factors of production in the study area. Land holdings of sampled households varied from no farm land (1.7%) to greater than 1 ha (30.2%) out of which 20.2% belonged to the treatment (invaded) areas and the remaining 10% belonged to the control (non-invaded) group. This implies that the largest share (49.1%) of the households in the study area own less than one hectare of farm land.

Survey results also showed that 3.1% from the control (non-invaded) group and 2.4% from the treatment group held less than 0.5 ha of farmland. About 10.9% from the control and 11.1% from the treatment group owned 0.5 ha of farmland, while 10.2% of households in the control group and 10.4% of households in the treatment (invaded) group owned 1 ha of farmland. The small size of plots or farmlands accompanied with some environmental stresses in the study areas further aggravated the already existing agricultural practice and of course adversely affected crop production.

As far as educational status of the respondents is concerned, nearly 80% and 66% of households from Prosopis-invaded and non-invaded Kebeles were illiterate, respectively. Only 19.80% from the invaded and 34% from the non-invaded Kebeles attended primary education. The result of chi-square test showed that there was a significant difference in the educational status of household heads at 1% level of significance between the two groups (χ 2 = 10.311). As the result indicates, majority of households who are invaded by Prosopis were illiterates.

Survey results showed that, out of the total 450 sampled households, 85% and 15% were male headed and female headed, respectively (Table 4). The results also indicated that 88% and 12% of Prosopis-invaded households and 82% and 18% of non-invaded households were male headed and female headed, respectively. t test result for differences in gender of the household head between the two groups was significant (t = −2.963)
Table 4

Socio-economic characteristics of sample households by household type

Pre-interv. variables

Description

Invaded

Non-invaded

Total

χ 2 value

(N = 250)

(N = 200)

(N = 450)

No.

%

No.

%

No.

%

SEXHH

MALE

220

88

164

82

384

85

2.963*

FEMALE

30

12

36

18

66

15

EDUHH

Illiterate

200

80

132

66

332

73

10.311***

Literate

50

20

68

34

118

27

Yes

71

28.4

145

72.50

216

50.45

25.604***

No

179

71.6

94

46.50

273

59.05

 

CREDITACC

Yes

103

41.2

68

33.66

171

37.43

2.153NS

No

147

58.8

132

65.35

279

62.08

DRTEF

Yes

175

70

190

94.06

365

82.03

43.879***

No

75

30

10

4.95

85

17.475

Number of food secured people

Food secure

126

50.4

81

40.5

207

45.45

4.048**

Not food secure

124

49.6

119

59.5

243

54.55

SAFTYNT

Yes

161

64.4

102

50.5

263

57.45

7.346***

No

89

35.6

98

48.51

187

42.06

IRRGTN

Yes

22

8.8

56

27.72

78

18.26

24.383***

No

228

91.2

144

71.29

372

81.25

Source: own survey data, 2015

*, ** and *** mean significant at the 10%, 5% and 1% probability levels, respectively

Not significant (NS)

In both Prosopis-invaded and non-invaded areas, it was observed that 64.4% and 50.5% of the respondents were engaged in food for work or the so-called safety net programme, respectively. This implies that more than half of the households are not in a position to feed themselves from income generated from crop and livestock production. Besides, the average number of food secured and food insecure households was assessed based on the calorie intake levels of sample respondents, and the results revealed that 59.5% of the respondents in non-invaded areas and 49.6% of the respondents in the invaded areas were food insecure. This means that in both cases, there is almost food uncertainty. The result of the chi-square test showed that there is a significant difference among the two groups in number of food secured people at 5% level of significance (χ 2 = 4.048) (Table 4).

In the study area, both perennial and annual crops are grown through rain-fed and irrigation agriculture. Subsistence mixed farming is the dominant farming type constituting 81.25% of the households in the study area. Irrigation accounts for about 27.72% of the households in the control (non-invaded) areas and only 8.8% in the treatment (invaded) areas. The chi-square test was also undertaken to check whether there is a significant mean difference in households’ access to irrigation between the two groups. The result showed that the difference was significant at 5% (χ 2 = 6.544) (Table 4).

In the study area, it was observed that 41.2% of the respondents in the invaded areas and 33.66% of the respondents in the non-invaded areas have access to rural credit service. As can be seen from Table 4, the chi-square test result indicated that there is insignificant difference in access to rural credit service between the two groups.

As far as cooperative membership is concerned, nearly 71.6% from Prosopis-invaded Kebeles and 46.50% from the non-invaded Kebeles are not members of cooperative respectively. The result of chi-square test showed that there was a significant difference in household heads’ membership to a cooperative, at 1% level of significance between the two groups (χ 2 = 25.604).

Table 5 shows that there were significant differences between the two groups of sample respondents with respect to expenditure on education, milk income, income from livestock sale, income from Prosopis, average annual income from crop production, food and non-food expenditure, health expenditure and calorie per day per Adult-Equivalent (AE). However, there were no statistical significant differences between the two groups of sample households with respect to off-farm income. The explanation for the significant outcome variables are presented below.
Table 5

Current income, expenditure and calorie intake of sample households by household type

Pre-interv. variables

Invaded N = (250)

Non-invaded N = (200)

Sample household N = (450)

Means difference

t value

Mean

STD

Mean

STD

Mean

STD

Mean

STD

EDUEXPE

562.644

485.170

366.620

301.501

465.119

415.608

169.024

183.669

−4.860***

AAILS

1,719.946

2,651.149

2,161.128

2,267.842

1,939.439

2,474.703

−441.182

383.307

1.792*

AAIMP

656.990

1,130.953

941.050

1,663.434

798.313

1,426.359

−284.06

−532.481

2.004**

AAFW/CP

1,235.460

2,223.456

561.200

1,529.271

900.008

1,937.154

674.26

694.185

−3.539***

AAICP

3,053.03

2,350.367

2,215.006

1,881.04

2,636.106

2,168.163

838.031

469.326

−3.944***

OFFFARMIN

6,173.871

9,412.567

5,036.040

14,727.954

1,137.831

−5,315.39

5,607.786

12,344.151

−0.9240NS

FOaNFEXPE

9,879.327

5,684.264

6,429.45

5,636.152

8,162.970

5,911.239

3,449.877

48.112

−6.110***

HEALEXPE

243.158

252.378

153.170

295.857

198.388

278.198

89.988

−43.48

−3.282***

CALINTAKE

2,704.443

1,718.556

1,859.700

1,165.135

2,284.173

1,527.393

844.743

553.422

−5.763***

Source: own survey data, 2015

*, ** and *** mean significant at the 10%, 5% and 1% probability levels, respectively

Not significant (NS)

Expenditure on education includes money spent on exercise books, pens, pencils, uniform, school fees and other education materials of the previous year. The mean education expenditure of the total sample households is Birr 465.119 (USD 24.162) per annum. The average education expenditure for Prosopis-invaded and non-invaded households is Birr 562.644 (USD 29.228) and Birr 366.620 (USD 19.045), respectively. The t test result shows that there is statistically significant difference among the two groups at 1% probability level.

The quantity of milk from the households who own livestock was estimated on the basis of the number of milking animals and the amount of milk that the average cows, goats, sheep and camels produced. Explicit costs including annual average cost for medication and supplementary feed cost were also considered with the assumption that income from milk production is equal to domestic consumption plus sale. The descriptive result shows that there is significant difference between the two groups of households in that the mean income from milk production by Prosopis-invaded and non-invaded households is Birr 656.990 (USD 34.129) and Birr 941.050 (USD 48.886), respectively. The t test result reveals that there is statistically significant difference between the two groups at 5% probability level. According to the respondents, reduced pasture availability (caused by mesquite invasion) and frequent rain shortage caused the problem. Pasture availability is of course one of the factors determining milk productivity. A comparative study conducted by Mugasi et al. (2000) showed that livestock reared under relatively highly encroached grazing fields yielded less milk than those reared under less encroached fields.

Income from livestock sale is the sum of money households obtained from sale of live livestock. The mean income from livestock sales in Prosopis-invaded and non-invaded households is Birr 1719.946 (USD 89.348) and Birr 2161.128 (USD 112.266), respectively. The t test result reveals that there is statistically significant difference between the two groups at 10% probability level. This study is in line with Pasiecznik (1999) which reported that mesquite invasion forms impermeable, dense thickets. The same source also stated that the invasion reduces grass cover of grazing lands and consequently affects stocking density.

Income from fuel wood and charcoal is a sum of money households obtained from fuel wood collection and charcoal making. The mean income from fuel wood and charcoal in Prosopis-invaded and non-invaded households is Birr 1235.460 (USD 64.180) and Birr 561.200 (USD 29.153), respectively. The t test reveals that there is statistically significant difference in income generation from firewood collection and charcoal making.

Explicit costs such as costs for fertilizer, pesticides, seeds and labour, taking a minimum wage of Birr 20 per day, was used in the calculation of net income from crop production. These quantities were converted to values (Birr) using households’ self-report of sales price. Based on this estimation, the mean annual income generated from crop production in Prosopis-invaded households is Birr 3053.03 (USD 158.599) while it is Birr 2215.006 (USD 115.065) for non-invaded households. There is a significant difference in income generated from crop production between invaded and non-invaded households, and the t value of −3.944 suggests that this variable is significant at 1% probability level. This study is also in line with Haji and Mohammed (2013) who confirmed that Prosopis is attributed to have resulted in increased crop yields by 29%. Prosopis have positive effects on the soil layer (Ilukor et al. 2014). These positive soil characteristics can be used for consecutive planting of native species on cropping fields. In some studies, the physico-chemical property of soil under mesquite canopy was found to be better than the adjacent open field (e.g. El Fadl, 1997, cited in Esther and Brent 2005) which may be due to nitrogen fixation, leaf litter addition and change in soil structure due to deep tap root system (Pasiecznik et al. 2001).

An effort has also been made to study the expenditure pattern among the households which is presented below.

Total household expenditure includes food and non-food expenditure of households in the previous year. The mean total household expenditure of invaded and non-invaded households is Birr 9879.327 (USD 513.212) and Birr 6429.45 (USD 333.997), respectively. The t test result reveals that there is a statistically significant difference between the two groups at 1% probability level.

Health expenditure is a sum of money spent for medication by the household for family members and their animals. The mean health expenditure of Prosopis-invaded and non-invaded households is Birr 243.158 (USD 12.6316) and Birr 153.170 (USD 7.957), respectively. The t test result reveals that there is statistically significant difference between the two groups at 1% probability level.

Findings of the study show that invaded households and non-invaded households had calorie intake of 2704.443 and 1859.7 cal per AE, respectively. As already mentioned, two weeks’ data on available food for sample households’ consumption, from purchase and/or stock, were converted to kilocalories and the figures were divided by the total AE in each households AE. Finally the computed kilocalorie per AE in each household was compared with the minimum subsistence energy requirement per AE per day, 2,100 kcal. Invaded households’ calorie intake is higher than non-invaded households (see Table 5).

Although education expenditure, income from fuel wood collection and charcoal making, income from crop production, food and non-food expenditure, health expenditure and calorie per day per AE of the invaded households were higher than those of non-invaded households, milk income and income from livestock sale in the invaded households were less than those of non-invaded households. However, this does not mean that the mean difference is exclusively because of the invasion of Prosopis. Since comparisons are not yet restricted to households who have similar characteristics, it is impossible to attribute the difference in the above variables between the two groups exclusively to Prosopis invasion. Therefore, to handle this shortcoming, a further analysis was conducted using propensity score matching techniques to strengthen the findings.

Estimation results

This section describes the steps followed to measure the impact of Prosopis invasion. More precisely, it presents the estimation of propensity scores, matching methods, common support region and balancing test. It also explains the treatment effect of Prosopis invasion across the invaded households.

Propensity scores

The results of a propensity score are obtained as the probability scores of individuals from the fitted simple logistic regression model. Logistic regression is applied when the dependent variable is dichotomous. The model is estimated with STATA 13 computing software using the propensity score matching algorithm developed by Leuven and Sianesi (2003). In the estimation process, data collected from the two groups, namely Prosopis-invaded households and non-invaded households, were pooled such that the dependent variable takes a value 1 if the household was invaded by Prosopis and 0 otherwise. Prior to running the regression model, the explanatory variables were checked for the existence of multi-collinearity and heteroscedasticity.

The Variance Inflation Factor (VIF) values of the variables in the model, as shown in Tables 12 and 13 in the Appendix, are less than the critical values showing that there is no problem of collinearity.

Heteroscedasticity was checked, and no problem was noticed in the data. As mentioned earlier, the variables included in the model are assumed to affect neither a household’s invasion with Prosopis nor the outcome of Prosopis invasion.

The overall model goodness of fit represented by model count R 2 is very good, and about 88% of the variation in outcome is explained by the explanatory variables in the model. The log likelihood ratio which follows the chi-square distribution was found to be 224.935 which is significant at 1% probability level, which implies that the entire explanatory variables together explained the variation in the dependent variable.

As can be seen from the estimated coefficients of all the 14 explanatory variables, Prosopis invasion is significantly influenced by six of them. More precisely, family size, dependency ratio and access to irrigation water had shown negative and significant relationship with Prosopis invasion. On the other hand, age of the household head, TLU and engagement in food for work programme were positively and significantly related to Prosopis invasion.

Dependency ratio of the respondent household was found to significantly and negatively affect Prosopis invasion, which is also in line with the hypothesis. This means that higher dependency ratio leads to lesser probability of the households to be invaded by Prosopis.

Households, who have large family size, have a significant and at the same time negative relationship with Prosopis invasion. This means that larger family size contributes large family labour to use Prosopis as source of income or more labour is available in the house for fuel wood collection. Therefore, households who have larger family sizes have a lower probability to be invaded by Prosopis.

In line with the hypothesis, age of the household head was found to be positively related with Prosopis invasion. This means that as the age of household head increases, his/her potential ability to eradicate Prosopis decreases due to limited working capacity.

Households who participated in food for work (safety net) programme were found to have a significant and positive relationship with Prosopis invasion, which is also in line with the hypothesis. This means households benefiting from this programme favour Prosopis as they will not use it as a source of income.

Livestock holding as a wealth variable indicates the capacity of the agro-pastoral household to involve in high-return income sources. The result, as shown in the table below, for livestock holding is not in line with the hypothesis. It is interesting that for the households who owned larger number of livestock, the higher the probability to be invaded by Prosopis. This is because, when agricultural/farm land of the households are invaded by Prosopis, they will find other income source for their living. Particularly agro-pastoral households want to expand livestock production as the result of invasion. The model output is depicted in Table 6.
Table 6

Results of the logistic regression model

Variables

Coef.

Std. err.

z

P > z

AGEHH

0.0715

0.0206

3.48***

0.001

SEXHH

0.3138

0.5473

0.57

0.566

EDUHH

−0.1205

0.4881

−0.25

0.805

DISROAD

0.0448

0.0588

0.76

0.446

FARMEXPHH

−0.0129

0.0246

−0.52

0.601

TOTHH

−0.2301

0.0905

−2.54**

0.011

DEPENRATIO

−0.4194

0.04724

−8.88***

0.000

TOTALAND

−0.2254

0.0794

−1.25

0.245

TLU

0.2149

0.0614

3.50***

0.000

IRRGTN

−1.8196

0.5134

−3.54***

0.000

CREDITACC

0.5186

0.4623

1.12

0.262

SAFTYNT

1.5612

0.4841

3.22***

0.001

DRTEF

−0.9156

0.5816

−1.57

0.115

_cons

0.6698

1.1556

0.58

0.562

Number of obs = 450

    

LR χ 2(13) 352.80

    

Prob > χ 2 = 0.0000

    

Log likelihood = −102.24075

    

Pseudo-R 2 = 0.6331

    

Source: own estimation result

*, ** and *** mean significant at the 10%, 5% and 1% probability levels, respectively

Figure 1 shows the distribution of households with respect to estimated propensity scores. As can be seen in the figure, most of the treatment households are found to the right of the distribution, while most of the control households are found in the left side of the distribution.
Figure 1

Distribution of households with estimated propensity scores

Matching of Prosopis-invaded households and non-invaded households

As noted earlier, before implementing the matching task, three main steps must be followed and these are presented as follows: First, predicted values of propensity scores should be estimated for all treated and control households. Second, a common support condition should be imposed on the propensity score distributions of household with and without the Prosopis invasion. Finally, observations whose predicted propensity scores fall outside the range of the common support region should be discarded.

The estimated propensity scores vary between 0.0496 and 0.996 (mean = 0.643) for Prosopis-invaded or treatment households and between 0.0163 and 0.965 (mean = 0.361) for non-invaded (control) households (Figures 2 and 3). The common support region would then lie between 0.0496 and 0.965. This means households whose estimated propensity scores are less than 0.0496 or greater than 0.965 were not considered for the matching exercise. As a result of this restriction, 10 households from treatment were discarded from the analysis (see Table 7).
Figure 2

Kernel density of propensity scores of non-invaded households

Figure 3

Kernel density of propensity scores of invaded households

Table 7

Distribution of sample households by estimated propensity scores and household type

Group

Obs

Mean

Std.

Minimum

Maximum

Total households

450

0.502

0.263

0.0163

0.996

Treatment households

250

0.643

0.239

0.0496

0.996

Control households

200

0.361

0.204

0.0163

0.965

Source: own estimation result

Choice of matching algorithm

In order to get the best matching estimators for matching the treatment and control households in the common support region, the nearest neighbour, caliper and kernel matching estimators were tested (Table 8). Finally, the matching estimators make use of different criteria such as equal means test referred to as the balancing test (Dehejia and Wahba 2002), pseudo-R 2 and matched sample size. In selecting a given matching estimator, a matching estimator which balances all explanatory variables (i.e. results in non-significant mean differences between the two groups) bears a low R 2 value and also results in a large matched sample size which is preferable. Consequently, a kernel matching with no bandwidth was found to be the best estimator for the data.
Table 8

Comparison of the three matching estimates by performance criteria

Matching estimator

Performance criteria

Balancing testa

Pseudo-R 2

Matched sample size

NN

 1 neighbour

8

0.109

440

 2 neighbour

8

0.082

440

 3 neighbour

9

0.067

440

 4 neighbour

10

0.064

440

Caliper

 0.01

12

0.038

396

 0.25

12

0.055

440

 0.5

7

0.134

440

KM

 With no bandwidth

13

0.033

440

 Bandwidth of 0.1

11

0.042

440

 Bandwidth of 0.25

11

0.050

440

 Bandwidth of 0.5

8

0.084

440

Source: own estimation result

aNumber of explanatory variables with no statistically significant mean differences between the matched groups of Prosopis-invaded households and non-invaded households

Table 9 depicts the estimated results of matching quality tests. Based on the selected best estimator (kernel matching with no bandwidth), the balancing test of covariates, before and after the matching of Prosopis-invaded and non-invaded households, shows that there are no statistically significant mean differences between the two groups in terms of age of the household head, distance to the main road, sex of the household head, education level of the household head, farm experience, family size of the household, dependency ratio, farm size, access to irrigation, participation in safety net programme and drought faced during the last three years.
Table 9

Results of the balancing tests of covariates using the kernel matching estimator

Variables

Before matching (450)

After matching (440)

Treated

Control

t value

Treated

Control

t value

N = (250)

N = (200)

N = (240)

N = (200)

AGEHH

42.490

38.580

3.690***

42.032

43.368

−1.220

SEXHH

0.881

0 .820

1.720*

0.871

0.890

−0.570

EDUHH

0.198

0.340

−3.240***

0.204

0.321

−0.120

DISROAD

1.840

2.475

−1.920*

1.806

1.552

0.770

FARMEXPHH

26.109

20.135

6.200***

25.317

26.357

−0.990

TOTHH

6.302

6.270

2.560**

6.274

6.856

−0.130

DEPENRATIO

1.149

1.428

−2.680***

1.146

1.261

−1.160

TOTALAND

2.621

2.1475

2.070**

2.534

3.0487

−1.560

TLU

5.033

4.610

2.310**

4.788

4.661

0.400

IRRGTN

0.089

0.280

−5.080***

0.0914

0.114

−0.720

CREDITACC

0.411

0.340

1.470

0.362

0.325

0.680

SAFTYNT

0.644

0.510

2.730***

0.577

0.535

0.740

DRTEF

0.698

0.950

−7.000***

0.731

0.653

1.640

Source: own estimation result

*, ** and *** means significant at 1%, 5% and 10% probability levels, respectively

Treatment effect on the treated

The effect of Prosopis invasion on households’ livelihoods was meticulously analysed. The estimated results show supportive evidence of statistically significant impact of Prosopis invasion on household expenditure on education, milk income, average annual income from crop production, off-farm income, food and non-food expenditure and food poverty measured in calorie intake. Pre-intervention differences (demographic, location, institutional and asset endowment characteristics) among the Prosopis-invaded and Prosopis non-invaded households were controlled.

The finding of this study revealed that mean difference in education expenditure, milk income, calorie intake per AE and average annual income obtained from fuel wood collection and charcoal making between the two groups of sample households is significant at 1% probability level. For food and non-food expenditure, the two groups of sample households showed significant mean difference at 5% probability level. Results also show that average annual income from milk, average annual incomes from crop production and off-farm income are significant at 5% probability levels. In other words, on average, Prosopis invasion has boosted education expenditure, average annual income from crop production, off-farm income, food and non-food expenditure, physical food consumption and income from fuel wood and charcoal in Prosopis-invaded household by Birr 213.053 (USD 11.068), Birr 529.43 (USD 27.503), Birr 3831.295 (USD 199.028), Birr 2607.103 (USD 135.434), 805.421 cal and Birr 2133.399 (USD 110.826), respectively, over the non-invaded households. On the other hand, Prosopis invasion has reduced milk income by Birr 410.667 over the non-invaded households (Table 10).
Table 10

Average treatment effect on the treated (ATT)

Variables

Treated

Controls

Difference

t value

EDUEXPE

572.646

359.593

213.053

3.63***

FOaNFEXPE

9,894.328

7,287.226

2,607.103

2.95**

HEALEXPE

260.349

215.813

44.536

0.48NS

CALINTAKE

2,673.691

1,868.270

805.421

3.96***

AAILS

1,676.303

1,796.247

−119.945

−0.33NS

AAIMP

677.241

1,087.908

−410.667

−1.68*

AAICrP

4,436.468

3,907.037

529.430

1.70*

OFFFARMIN

6,030.369

2,199.075

3,831.295

1.78*

AAFW/CP

2,257.849

124.450

2,133.399

4.57***

Source: own estimation result

*, ** and *** means significant at 1%, 5% and 10% probability levels, respectively

Not significant (NS)

Sensitivity analysis

In this section, the issue whether or not the final evaluation results are sensitive with respect to the choice of the balancing scores is addressed. Matching estimators work under the assumption that a convincing source of exogenous variation of treatment assignment does not exist. Likewise, sensitivity analysis was undertaken to detect if the identification of conditional independence assumption was satisfactory or affected by the dummy confounder or the estimated Average Treatment Effect on the Treated (ATT) is robust to specific failure of the CIA. Sensitivity analysis using the bounding approach was employed, and this involves calculating upper and lower bounds, using the Wilcoxon signed rank test. These rank tests test the null hypothesis of no-treatment effect for different hypothesized values of unobserved selection bias. Table 11 reveals the sensitivity result of outcome ATT values to the dummy confounder.
Table 11

Sensitivity analysis of the estimated ATT

Outcome variables

Percentage change

EDUEXPE

4.63

AAIMP

13.74

AAICrP

4.68

OFFFARMIN

12.47

FOaNFEXPE

8.23

AAFW/CP

2.19

CALINTAKE

5.45

Source: own estimation result

Regarding milk income, off-farm income, income from crop production and income from Prosopis, the average treatment effect on the treated is found to be insensitive to the dummy confounder. Similarly, education expenditure, food and non-food expenditure and calorie per day per AE were also found to be robust or insensitive to the dummy confounder. This shows how strongly an unmeasured variable influences the selection process in order to undermine the implications of matching analysis.

Conclusions

In this study, cross-sectional data from Prosopis-invaded and non-invaded Kebeles of rural Dire Dawa Administration were used to explore the impacts of Prosopis invasion on households’ welfare. The main question that this research attempted to answer was, ‘what would be the livelihoods of Prosopis-invaded households had they not been invaded by Prosopis?’ Answering this question requires observing outcomes with and without the Prosopis invasion for the same household. However, it is impossible to observe the same object in two states simultaneously. In other words, the fundamental problem in any social programme evaluation is the missing data problem. While the programme evaluator observes the factual for an object, it is impossible to observe the counter-factual for the same object.

This study used the propensity score matching technique to eliminate the possible sample selection bias since the data were from a survey study. Kebeles were grouped into two strata. Stratum one, which represent the treatment group, consists of four Kebeles that are severely invaded by Prosopis. The other three Kebeles which are not invaded by Prosopis are categorized under stratum two representing the control group.

The present study is a comparison between Prosopis-invaded Kebeles forming four in number and Prosopis non-invaded Kebeles numbering three. As expected, a household’s invasion by Prosopis was determined by a combination of factors. These are age of the household head, family size of the household, dependency ratio, access to irrigation water, TLU and engagement in food for work programme. Finding a reliable estimate of the Prosopis impact thus necessitates controlling for all such factors adequately. In doing so, propensity score matching has resulted in 240 treated households to be matched with 200 controlled households. In other words, a matched comparison of all outcome variables was performed on these households who shared similar pre-intervention characteristics except the Prosopis. The resulting matches passed a variety of matching quality tests and were fit for addressing the main objectives of this study.

After controlling for other characteristics, it has been found that Prosopis invasion had significantly boosted the Prosopis-invaded household, over the non-invaded households on education expenditure by Birr 213.053 (37.23%), average annual income from crop production by Birr 529.43 (11.93%), off-farm income by Birr 3831.295 (63.53%), food and non-food expenditure by Birr 2607.103 (26.35%) and physical food consumption by 805.421 cal per AE (30.12%). On the other hand, Prosopis invasion reduces milk income of the Prosopis-invaded household by Birr 410.667 (60.64%) over the non-invaded households. Although this study tried to capture some aspects of the Prosopis invasion in the rural Dire Dawa Administration, more aspects of Prosopis invasion such as more benefits and social costs were not examined, making the total economic evaluation incomplete.

Recommendations

Based on our results, we provide the following recommendations for the sustainable management and control of Prosopis in the Dire Dawa Administration.

First, any projects or programmes that aim to tackle the challenges of Prosopis invasion and optimize the positive impacts of this species need to consider the Prosopis invasion effects explicitly, as there is huge market failure and externalities relative to the Prosopis invasion. Moreover, active involvement of the local community is important in the design and success of Prosopis invasion control/eradication programmes. So raising awareness of the local community about the pros and cons of Prosopis effects on households’ welfare will play a great role for the success of community-based Prosopis invasion control/eradication programmes. Hence, before trying to mobilize the community in control/eradication of Prosopis invasion, an awareness creation activity has to be done.

Second, age of the household head was found to be positively related to the probability of Prosopis invasion. Therefore, encouragement must be given to older household heads on the control/eradication of Prosopis invasion. In other words, development agencies and policy-makers should target Prosopis control/eradication technologies on the basis of age.

Third, households with large family size found to have lesser probability to be invaded by Prosopis because the family could have contributed higher labour force invasion can be an opportunity. Projects/programmes that work on efficient family labour allocation and family planning need to be encouraged in maintaining and minimizing household size to the level of household income capability. The implementation of family planning programmes should be supported with the current health-extension package of giving high emphasis on local people’s perception of household size and on their attitudes towards family planning facilities.

Fourth, livestock rearing is the most important economic activity for the agro-pastoral households in the study area. Projects like dairy cow credits, sheep and goat credits, camel credit and fattening need to be supported through establishment of organized credit facilities, together husbandry skill and knowledge training for improvement of livestock management to increase family income can gain capacity to control/eradicate Prosopis invasion. Management of herds (stocking and restocking) and utilization of improved feed and fodders in combination with Prosopis as a feed source need to be given due attention. Side by side, the extension service needs to give due attention to problems of animal health and critical feed shortage. In addition, necessary inputs should be provided on time, as per the requirement of the farmers.

Fifth, food security programmes are positively related to Prosopis invasion. Therefore, government and NGO intervention with regard to food aid and different food security programmes and projects have to take into consideration the negative impacts of Prosopis invasion. Since some invaded households gain a good source of income, promoting Prosopis utilization in a planned way through adoption of appropriate and sustainable management is advisable in the study area.

Sixth, the study results also indicated that access to irrigation was negatively related to Prosopis invasion. Therefore, in order to address the Prosopis invasion problems in the study area, attention should given by GOs and NGOs to developing small-scale irrigation or water harvesting technology, by organizing farmers into watershed cooperatives.

Seventh, due to the fact that the local communities in the invaded area have a record of better harvest in lands cleared of Prosopis, converting some of the invaded grazing lands into cultivable land (especially sorghum) is also another point of recommendation.

Last, but not the least, further research is needed on the total economic valuation of Prosopis invasion on the administration so as to gain more insight into the impact of Prosopis invasion.

Footnotes
1

A Kebele is the smallest administrative unit of Ethiopia similar to a ward.

 

Declarations

Acknowledgements

We acknowledge the Ethiopian Ministry of Education for financing the Ph.D. study. We are also grateful for the comments provided by the anonymous reviewers.

Authors’ contributions

All authors of this manuscript have directly participated in the planning and execution of this study, and therefore, we confirm that the manuscript has been read and approved by all named authors.

Authors’ information

Negussie Zeray (Ph.D. fellow): Ph.D. fellow since 2006 in the field of Agricultural Economics; MSc degree in Agricultural Economics from Alemaya University of Agriculture, Ethiopia, in 2010; and BSc degree in Agricultural Extension from Alemaya University of Agriculture, Ethiopia, in 1998. Belaineh Legesse (Associate Professor): Ph.D. degree in Rural Development Studies from Swedish University of Agricultural Sciences (SLU) in April, 2003; MSc degree in Rural Development Studies from Swedish University of Agricultural Sciences (SLU) in 2000; MSc degree in Agricultural Economics from Alemaya University of Agriculture, Ethiopia, in 1991; and BSc degree in Agricultural Economics from Alemaya University of Agriculture, Ethiopia, in 1984. Jema Haji Mohamed (Associate Professor): Ph.D. degree in Economics from Swedish University of Agricultural Sciences (SLU) in April, 2008; MSc degree in Economics from Swedish University of Agricultural Sciences (SLU) in 2004; MSc degree in Mathematics from Addis Ababa University, Ethiopia, in 1999; BSc degree in Mathematics from Addis Ababa University, Ethiopia, in 1989.

Mengistu Ketema Aredo (Associate Professor): Ph.D. Agricultural Economics.

Competing interests

The authors declare that they have no competing interests.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Department of Agricultural Economics, Dilla University
(2)
Department of Agricultural Economics and Agribusiness, Haramaya University
(3)
Haramaya University

References

  1. Anagae, A., F. Reda, G. Tesfaye, A. Admasu, and Y. Ayalew. 2004. Policy and stakeholder analysis for invasive plants management in Ethiopia, 3. Ethiopia: Ethiopian Agricultural Research Organization. Report submitted to CAB International under the PDF-B Phase of the UNEP/GEF-Funded Project: Removing Barriers to Invasive plants Management in Africa.
  2. ATA (Ethiopian Agricultural transformation Agency). 2015. Agricultural transformation progress report. Addis Ababa.
  3. Becker, S.O., and A. Ichino. 2002. Estimation of average treatment effects based on propensity scores. The Stata Journal 2(4): 358–377.Google Scholar
  4. BoARD (Bureau of Agriculture and Rural Development). 2009. Annual report. Dire Dawa.
  5. Bryson, A., R. Dorsett, and S. Purdon. 2002. The use of propensity score matching in the evaluation of labour market policies, working paper no. 4.. Department for Work and Pensions. London School of Economics.
  6. Caliendo, M., and S. Kopeinig. 2005. Some practical guidance for the implementation of propensity score matching, the discussion 1588. Cologne: University of Cologne.
  7. Dehejia, R.H., and S. Wahba. 2002. Propensity score matching methods for non-experimental causal studies. The Review of Economics and Statistics 84(1): 151–161.View ArticleGoogle Scholar
  8. Dubale, Admasu. 2008. Invasive plants and food security: The case of Prosopis juliflora in the Afar region of Ethiopia, FARM-Africa for IUCN. Bangor: University of Wales.
  9. Esther, M., and S. Brent. 2005. Invasion of Prosopis juliflora and local livelihoods: Case study from the Lake Baringo area of Kenya, ICRAF working paper no. 3, 66. Nairobi: World Agroforestry Center.Google Scholar
  10. Federal Democratic Republic of Ethiopia. 2010. Ethiopia’s agriculture sector policy and investment framework (PIF) 2010-2020. Ethiopia.
  11. Haji, J., and A. Mohammed. 2013. Economic impact of Prosopis juliflora on agro-pastoral households of Dire Dawa Administration, Ethiopia. African Journal of Agricultural Research 8(9): 768–779.Google Scholar
  12. Heckman, J.J., H. Ichimura, J. Smith, and P.E. Todd. 1998. Characterizing selection bias using experimental data. Economics 66(5): 1017–1098.View ArticleGoogle Scholar
  13. IDP (Integrated Development Plan). 2006. Annual report. Dire Dawa, 2006/7 2010/2011.Google Scholar
  14. Ilukor, J., R. Birner, M. Tilahun, and S. Getu. 2014. A socialeconomic assessment of the impact of Prosopis juliflora invasion andbparticipative management approaches in the Afar Region, Ethiopia. In anaging Prosopis Juliflora for better (agro-) pastoral Livelihoods in the Horn of Africa Proceedings of the Regional Conference May 1 - May 2, 2014, Addis Ababa.
  15. Jalan, J., and M. Ravallion. 2003. Estimating the benefit incidence of an antipoverty program by propensity-score matching. Journal of Business and Economics Statistics 21(1): 19–30.View ArticleGoogle Scholar
  16. Leuven, E., and B. Sianesi. 2003. PSMATCH2. Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. Software, http://ideas.repec.org/c/boc/bocode/s432001.html.
  17. Liba, Pejchar and Gretchen C. Daily. 2009. The impact of invasive species on ecosystem services and human well-being. Stanford University. Trends in Ecology and Evolution 24(9): 497-50.
  18. Mehari, Z. 2008. Invasion of Prosopis (SW.) DC and rural livelihoods. The case of Afar pastoralists at Middle Awash Area of Ethiopia. MSc. thesis, 3. Oslo: Norwegian University of Life Science.
  19. Mugasi, S.K., E.N. Sabiit, and B.M. Tayebwa. 2000. The economic implications of bush encroachment on livestock farming in rangelands of Uganda. African Journal of Range and Forage Science 17: 64–69.View ArticleGoogle Scholar
  20. Pasiecznik, N.M. 1999. Prosopis-pest or providence, weed or wonder tree?.. (Newsletter No. 28): European Tropical Forest Research Network.
  21. Pasiecznik, N.M., P. Felker, P.J.C. Harris, L.N. Harsh, G. Cruz, J.C. Tewari, K. Cadorer, and L.J. Maldonado. 2001. The Prosopis juliflora-Prosopis pallida complex: A monograph, 172. UK: HARDCOVER EDITION IS ALSO AVAILABLE Coventry.
  22. Rosembaum, P.R., and D.B. Rubin. 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70(1): 41–55.View ArticleGoogle Scholar
  23. Ross T. Shackleton, David C. Le Maitre, Nick M. Pasiecznik and David M. Richardson. (2014). Prosopis: A global assessment of the biogeography, benefits, impacts and management of one of the world’s worst woody invasive plant taxa. AoB PLANTS 6: plu027; doi:10.1093/aobpla/plu027.
  24. Stefan, A. 2005. Spread of the introduction tree Prosopis juliflora (Sw.) DC in the Lake Baringo area, Kenya. Instituionen for skoglig vegetationsekologi, Grafiska enheten, SLU, Umeå.
  25. Taye, T., R. Fessehaie, and Y. Firehun. 2007. Invasive alien weed species in Ethiopia: Biology, distribution and importance, and available control measures. The Rio-Earth Summit, 1992. Rio de Janeiro: Convention on Biodiversity Agenda 21.Google Scholar
  26. Tilahun, Mesfin, Regina, Birner, and John, Ilukor. (2014). Households’ demand for mitigation of Prosopis juliflora invasion in the Afar Region of Ethiopia: A contingent valuation. In Managing Prosopis Juliflora for better (agro-) pastoral Livelihoods in the Horn of Africa Proceedings of the Regional Conference May 1- May 2, 2014. Addis Ababa, Ethiopia.
  27. WWDSE (Water Works Design and Supervision Enterprise). 2003. Water sector development program (project ETH/98/001). Addis Ababa: Ministry of Water Resources.Google Scholar

Copyright

© The Author(s). 2017