Livestock Research for Rural Development 32 (10) 2020 LRRD Search LRRD Misssion Guide for preparation of papers LRRD Newsletter

Citation of this paper

Does cattle production contribute to improving welfare of poor ethnic minority households in Central Vietnam?

Truong Lam Do, Ho Ngoc Ninh, Tran Dinh Thao and Nguyen Xuan Trach1

Faculty of Economics and Rural Development, Vietnam National University of Agriculture, PO Box 131000, Hanoi, Vietnam
thaoktl@vnua.edu.vn
1 Faculty of Animal Science, Vietnam National University of Agriculture, P O Box 131000, Hanoi, Vietnam

Abstract

The present study measured the impact of cattle production on the welfare of ethnic minority households with the propensity score matching method and determined the factors explaining the participation in cattle production of ethnic minority households with generalized Poisson, negative binomial and binary logistic regression models. Cross-sectional data of 500 households collected in 2019 in five provinces of Central Vietnam, viz. Nghe An, Quang Binh, Quang Ngai, Kon Tum, and Dak Nong, were used for the analyses. Results showed that: (1) the welfare of the ethnic minority households with cattle production was higher than that of those without cattle production, and (2) the participation in cattle production of the ethnic minority households was affected by various factors such as government supports under the national program on sustainable poverty reduction, the age and education level of the household head, number of mobile phones, number of motorbikes, number of health shocks, and access to the national electricity system.

Keywords: propensity score matching, generalized Poisson regression, negative binomial regression, poverty


Introduction

Developing countries in tropical regions are encouraging the development of livestock production that is economically, socially, and environmentally sustainable (Ness et al 2007). Livestock in general and cattle in particular offer an alternative source of capital that the poor can accumulate as a “savings account” to hedge against income fluctuations (Kazianga and Udry 2006; Moll 2005). Keeping cattle is therefore considered an alternative form of insurance, providing the rural households with assets that can be sold in times of shocks (Hoddinott 2006; Mogues 2011). Cattle production also plays a very important role in the livestock-crop integrated farming systems, in which cattle provide crop production with draft power for land preparation and excreta as fertilizer while consuming crop byproducts as feed, thus reducing inputs from outside and minimizing environmental pollution. Moreover, beef consumption is increasing over time, particularly in the major urban centers, due to both tourism and increasing disposable income of the local population (Parsons et al 2013). That should offer the poor a good opportunity to improve their livelihood through participation in cattle production.

Vietnam has achieved remarkable economic growth and poverty reduction for recent years. Poverty reduction in Vietnam has been internationally recognized with the poverty headcount rate (measured in monetary expenditure) massively declining from 57% in the early 90s down to 13.5% in 2014 (UNDP 2018). The Government of Vietnam promulgated the national multidimensional poverty measurements for the application in the 2016-2020 period in November 2015. The multidimensional poverty measurements include both income and non-monetary dimensions. Multidimensional poverty is measured through five basic social services (or five dimensions): health-care, education, housing, water and sanitation, and information access. A household is considered multidimensionally poor if it is deprived in at least three indicators (Duc 2019). With supports from the Government of Vietnam through the national program in sustainable poverty reduction as well as other NGOs, private organizations, the multidimensional poverty rate of Vietnam dropped to 5.23% in 2018 (MOLISA 2019). However, ethnic minority households in areas such as Central Vietnam are still poor. The proportion of ethnic minority populations at the end of 2018 accounted for 12.51% of the total national population; nevertheless, the proportion of poor ethnic minority households accounted for 55.27% of the poor households of Vietnam. The achievements of poverty reduction are therefore not sustainable, and the rate of re-poverty is still high (Government of Vietnam 2019).

So far, there have been numerous studies on household poverty in Vietnam, including those focusing on ethnic minorities ( Van de Walle and Gunewardena 2001; Dong et al 2005; Baulch et al 2007; Baulch et al 2011; Imai et al 2011; Pham et al 2011; Baulch et al 2012; WB 2012; Nguyen 2012; Tuyen 2014; Nguyen et al 2017; Do et al 2019; etc.). There have also been several studies on cattle production in Vietnam, including Central Vietnam (Ha et al 2014; Huyen et al 2011; Huyen et al 2013; Parsons et al 2013; Tra et al 2010; Werner Stür et al 2013; etc.). However, compared with the previous studies, the present study had three different features. Firstly, it focused on the poor group of ethnic minorities in the Central of Vietnam, using cross-sectional data. Secondly, it used propensity score matching techniques to estimate the impact of cattle production on the welfare of the poor ethnic minority households to examine whether raising cattle could help the poor households in the region improve their welfare with reduced poverty. Thirdly, it determined factors affecting cattle production of the poor ethnic minority households in the Central of Vietnam.


Materials and methods

Conceptual framework

The sustainable livelihoods framework (Carney 1998; Ashley and Carney 1999; Nguyen et al 2018) was used as a conceptual framework to analyze cattle production as a livelihood strategy of the ethnic minority households (Figure 1). The choice of cattle production as a livelihood strategy by the ethnic minority households depends on their assets, vulnerability context, and local infrastructure. The ethnic minority households has 5 key livelihood assets, viz. social, physical, human, financial, and natural capitals. The vulnerability context includes income shocks such as storms, droughts, floods, ills, accidents, fires, etc. The vulnerability context has effects on livelihood assets, livelihood strategies, the local infrastructure, and livelihood outcomes of the ethnic minority households. The local infrastructure includes road systems, national electricity systems, distance from the household to the nearest town, etc.

Figure 1. The conceptual framework for the study

Practicing cattle production, the ethnic minority households get both direct and indirect incomes, which contribute to reducing their poverty. The direct income comes from selling live cattle or milk or from providing draft power for farming and transport (McMichael et al 2007). The indirect income comes from cattle waste used as manure to improve soil fertility, thus contributing to improved crop production for food and cash (Nguyen et al 2016), and as fuel for cooking and heating.

Study areas and data collection

The Central of Vietnam, consisting of the North Central and Central Coastal areas, and Central Highlands, is one of the poorest and most remote regions in the country. Most of the provinces in the region are inhabited by ethnic minorities, which can be categorized into two groups: the first group consists of ethnic minorities with a larger population such as Thai, Paco, Bru Van Kieu, Kadong, Xo Dang, M’ Nong, Bana, E De, etc., and the second group consists of those with a population less than 10,000 people (called very small minority group) such as Chut, Hre, etc. Most of poor households in the region belong to ethnic minorities (see details in Table A1 in the Appendix section). Therefore, the present study focused on five provinces with high poverty incidences of ethnic minorities but having potential for cattle production in Central Vietnam, viz. Nghe An, Quang Binh, Quang Ngai (in the North Central and Central Coastal areas), Dak Nong and Kon Tum (in the Central Highlands) (Figure 2).

Figure 2. Poverty map of Vietnam and the study sites

A three-stage procedure was applied for data collection, which was conducted at the beginning of 2019. The first stage was the selection of sample districts in each province. Using the district profiles and consulting local experts at the provincial level, two districts in each province were chosen based on the following criteria: (1) Representatives in terms of ethnic minorities; (2) High poverty incidence of ethnic minorities; (3) High potential for cattle production. As a result, ten representative districts chosen in the five selected provinces consisted of: Tuong Duong and Quy Chau districts of Nghe An province, Tuyen Hoa and Minh Hoa districts of Quang Binh province, Son Tay and Son Ha districts of Quang Ngai province, KonPlong and Kon Ray districts of Kon Tum province, and Tuy Duc and Dak Glong districts of Dak Nong province. Then, the second stage was identification of communes and villages to study based on the same criteria as set for the selection of districts. Two communes in each selected district and two villages in each selected commune were chosen. The last stage was selection of representative poor ethnic minority households from a list of poor households in the selected villages. The numbers of selected households were determined based on the proportion of the poor ethnic households in the villages. This allowed us to set up a cross-sectional data of 519 households, of which 111 households were in Nghe An, 122 households in Quang Binh, 115 households in Quang Ngai, 88 households in Kon Tum, and 83 households in Dak Nong. After excluding the households with missing quantitative information, the final data set for the analyses consisted of 500 households.

Data were collected with two types of questionnaires: household questionnaire and village questionnaire. The household questionnaire consisted of questions for demographic, economic, and social information on ethnic minority households as well as subsidy from governmental programs. The required information was focused on livelihood resources (human, physical, social, natural, and financial capitals) and livelihood activities (farming, non-farm self-employment, off-farm employment, and other income sources). The household questionnaire had a section on inputs and outputs of cattle production as well as household expenditure in 2018. Cattle production income was calculated as total revenue minus total costs. The total revenue was equal to the value of the liveweight gain of cattle raised for one year. Total costs consisted of expenses for seedstock animals, feeding, veterinary care, hired labor, housing and facilities depreciation, and interest. In addition, another section on shocks was designed, following Do et al (2019), to gather various perceived shock events that the households had faced during the last three years. These shocks were categorized into three groups, viz. weather shocks, health shocks, and other shocks (e.g. market shocks, etc.). Weather shocks contained floods, droughts, heavy rainfalls, landslides, and storms. Health shocks consisted of events such as illnesses or deaths of household members. Moreover, the information on the access of households to the supports from the state and local governments through the poverty reduction program, especially subsidy for livestock production were also included in the household questionnaire. In addition to the household questionnaire, the village questionnaire was designed to capture village-level data on population, infrastructure, and other socio-economic indicators of the village.

Data analysis
Measuring the impact of cattle production on household welfare

Dependent variables were compared between the treatment and control groups to estimate the average treatment effect on the treated (ATT) for examining the impact of cattle production on household welfare. The dependent variables included daily income per capita, total household income, and headcount indices. The treatment group was households with cattle production, while the control group was households without cattle production. To compare dependent variables between the two groups, the propensity score matching (PSM) method, which can correct for biases from observed characteristics and contribute to addressing the endogeneity problems, was used. First, the propensity scores were estimated based on livelihood capitals, income shocks, village and province characteristics to match the households between the treatment and control groups using a probit model. The probit model was defined as follows:

P(X) = P(Dij = 1/ Xij, IS ij, Vj, Profe)         (1)

The dependent variable in equation 1 denotes the probability that household i in village j had cattle production. The dummy variable (Dij) is equal to one if household i in village j had cattle production and zero otherwise. The probability depends on observable livelihood capitals ( Xij), the number of income shocks that the household had faced during the last three years (ISij), observable village characteristics (Vj), and unobservable factors that might influence the household decision at the provincial level (Profe).

Three methods of the nearest-neighbor matching (NNM), the Kernel based matching (KBM), and the radius matching (Radius) were used to match households between the treatment and control groups to estimate ATT. The five nearest neighbor method was used with common support and replacement for the NNM and common support and with bandwidth 0.06 for both the KBM estimator and the Radius estimator. For the KBM and the Radius methods, the standard errors are bootstrapped for 1,000 replications to assess the variability of propensity score matching estimators; whereas, for the NNM, the standard errors were not bootstrapped as the standard bootstrap was not valid (Abadie and Imbens 2008).

To test these matching methods, the histograms of the estimated propensity scores and covariate balancing tests were performed. The figure of the histograms of the estimated propensity scores for the treatment and control groups is presented as Figure A1 in the Appendix section. The figure shows that there was a considerable overlap in the common support and the common support conditions was matched. The results of covariate balancing tests before and after matching (Table A5 in the Appendix section) show that: (1) the standardized mean differences (Caliendo and Kopeinig 2008) for overall covariates used in the propensity scores (from 20.1 to 30.3 percent) before matching were sharply reduced (from 1.0 to 2.6 percent) after matching, (2) the percentages of bias reductions were in the range of 76 to 83 percent through matching, (3) the likelihood ratio tests (p-value) show that the joint significance of covariates was always rejected after matching while it was never rejected before matching, and (4) the pseudo-R 2 also decreased significantly from 12.1-25.4 percent before matching to 1.0-2.6 percent after matching. The proposed specification of the propensity scores was successful in balancing the distribution of covariates between the treatment and control groups due to the low mean standardized bias, high percentages of bias reduction, the insignificance of the likelihood ratio test, and low pseudo-R2 after matching.

Based on the propensity scores, the impact of cattle production on household welfare was modeled as follows:

where, C and T denote the treatment and control groups, respectively; Y represents the dependent variable as defined above.

Identification of the determinants of cattle production

To determine the factors affecting cattle production of rural poor households, we use a Poisson regression model (Gujarati 2004) as follows:

Ni = E(Ni) + ui = E(Ni | Xij, ISij, Vj, Profe) + ui                            (3)

where, Ni denotes the number of the cattle of household i. Xij, ISij, Vj, and Profe are as in Equation 1, and ui is the error term. Results of the Poisson regression are reliable if the equidispersion assumption is correct. However, a significant (p <0.05) test statistic from the gof (Stata 15.1) indicated that the Poisson model was inappropriate. Generalized Poisson regression and negative binomial regression models are often more appropriate in cases of overdispersion. Thus, likelihood-ratio test results show that both Generalized Poisson regression and Negative Binomial regression models were better than a Poisson model. Therefore, we used both models to examine factors affecting the number of cattle raised by rural poor households. The robust option was used to control possible heteroscedasticity in both models.

Besides, a binary logit regression model was applied to examine factors affecting the decision of the household on whether or not to raise cattle. The model was as follows:

where, Qij is a binary variable that is equal to 1 household i in the village j who had cattle production and equal to 0 otherwise; Xij, ISij, Vj, and are as in Equation 1, and εi is an error term. The robust option was also used to control possible heteroscedasticity in the model.

At the village level, we used four variables, viz. the distance from the village to the nearest town, a dummy variable of whether the village was physically accessible during the whole year, the share of farmland that an irrigation system could provide water, and a dummy variable of whether the household accessed to the national electricity grids. The dependent and independent variables are summarized in Table A2 in the Appendix section. As the number of independent variables was large, we used the variance inflation factor (VIF) test to detect potential multicollinearity. The results of the VIF test rejected the null hypothesis of the problem (see Table A4 in the Appendix section).


Results and discussion

Description of household groups

Table 1 presents assets of both households with cattle (with cattle group) and households without cattle (without cattle group) together with their village-linked characteristics.

Table 1. Basic household assets and village-linked characteristics

Variable

With cattle group
(n = 220)

Without cattle group
(n = 280)

Statistic
test

p

Mean

SD

Mean

SD

Human capital

Ethnic (very small minority = 1; otherwise = 0)

0.11

0.31

0.14

0.35

1.37c

0.24

Gender (male=1, female=0)

0.85

0.36

0.79

0.41

3.36c

0.07

Age (year)

43.3

13.9

41.7

15.2

1.80b

0.07

Education (level)

2.89

1.39

2.64

1.41

1.87b

0.06

Household laborers (person)

2.54

1.12

2.13

1.14

4.41b

<0.001

Household size (person)

4.32

1.38

3.70

1.51

5.02b

<0.001

Natural capital

Farm land size (ha)

0.27

0.65

0.16

0.47

4.14b

<0.001

Perennial land size (ha)

0.58

1.11

0.38

0.70

1.69b

0.09

Social capital

SPO (Yes=1, No=0)

0.26

0.44

0.21

0.41

1.88c

0.17

Subsidy (Yes=1, No=0)

0.66

0.48

0.58

0.50

2.99c

0.08

Mobile phones (number)

1.40

1.07

0.86

0.91

5.87b

<0.001

Physical capital

Motorbike (number)

0.84

0.59

0.59

0.57

4.84a

<0.001

Financial capital

Access to credit (Yes=1, No=0)

0.66

0.48

0.53

0.50

8.65c

0.003

Shocks

Weather shocks (number)

1.13

1.26

0.95

1.26

2.19b

0.03

Health shocks (number)

0.29

0.54

0.30

0.60

0.13b

0.89

Other shocks (number)

0.78

0.92

0.62

0.86

2.14b

0.03

Village variables

Road type (Yes=1, No=0)

0.91

0.30

0.92

0.28

0.27c

0.60

Distance to town ( km)

32.7

10.5

31.6

11.3

1.22b

0.22

Water system (Yes=1, No=0)

0.82

0.17

0.79

0.18

1.74b

0.08

Access to e lectricity (Yes=1, No=0)

0.90

0.31

0.94

0.23

3.85c

0.05

Note: a T-test, b nonparametric two-sample test: Mann–Whitney U test, Chi-square test; SD is standard deviation, SPO is social-political organizations

The results of the T-test, Mann-Whitney U, and Chi-square tests showed that the human capital of the with cattle group was higher than that of the without cattle group. In details, the average education level of the household head of the with cattle group was higher than that of the without cattle group. The households with cattle had higher numbers of laborers and members than households without cattle. For natural capital, the households with cattle owned larger farmland size and perennial land size compared to the households without cattle. For social capital, there were no differences in participation of household members in social-political organizations between the two groups. However, the households with cattle received higher subsidies from the national government and the local government compared to the households without cattle. The social network through mobile phone of the households with cattle was better than that of the households without cattle. For physical and financial capitals, the households with cattle had a higher number of motorbikes and better access to credit than the households without cattle. For income shocks, the households with cattle faced a higher number of shocks for the last three years than the others. Regarding village characteristics, the households with cattle had better access to irrigation but worse access to electricity compared to the households without cattle.

In Central Vietnam, the number of headcounts and liveweight of cattle increased throughout the period 2014-2018 (see details in Table A3). Table 2 reports that the number of cattle the households raised ranged from one to nine. Most of the households raised from 1 to 3 cattle. The number of households with a larger number of cattle decreased. The percentage of households with one cattle was the highest (45.5%); whereas, the percentage of households with eight or nine cattle was the lowest (0.45%).

Table 2. Histogram of the number of cattle by households

Number
of cattle

Number of
households

Percentage

Cumulative
percentage

1

100

45.5

45.5

2

56

25.5

70.9

3

35

15.9

86.8

4

13

5.91

92.7

5

6

2.73

95.5

6

6

2.73

98.2

7

2

0.91

99.1

8

1

0.45

99.6

9

1

0.45

100

Total

220

100

-

In respect to income, the households had several income sources such as crop production income, livestock production income, forestry income, government salary, labor rented income, and other incomes (Table 3). Income from cattle production was the main income source of the households with cattle. This finding is consistent with that from the study of Beranger and Vissac (1993) showing that on mountainous areas development of livestock production is more favorable than crop production. Comparing incomes between the with cattle group and the without cattle group, the former had higher total income and income per capita than the latter. The with cattle group also had both higher crop production income and livestock production income than the other group.

Table 3. Annual household income by income source in 2018 (in 100 USD)

Income source

With cattle
group (n = 220)

Without cattle
group (n = 280)

Statistic
test

Prob.

Mean

SD

Mean

SD

Crop production income

3.77

5.07

2.64

4.70

4.02b

<0.001

Livestock without cattle income

0.04

0.08

0.03

0.08

4.39b

<0.001

Cattle production income

5.40

3.73

0.00

0.00

24.2a

<0.001

Forestry income

1.04

2.67

1.21

3.18

-0.17b

0.86

Government salary

0.76

3.08

0.49

2.75

1.78b

0.08

Labor rented income

4.75

6.93

5.08

6.13

-1.64b

0.10

Other incomes

1.04

3.15

1.15

4.04

-0.35b

0.73

Total income

16.8

10.4

10.6

9.25

7.69b

<0.001

Income per capita

4.48

3.13

3.31

3.00

5.59b

<0.001

SD is standard deviation; b nonparametric two-sample test: Mann–Whitney U test

Impact of cattle production on the welfare of poor households of ethnic minorities

Table 4 reports the ATT with diverse matching algorithms that measured the impact of cattle production on the poor household’s welfare. The results show that the with cattle group had higher income and lower poverty rate than the other group. In details, for the multidimensional poverty rate, cattle production contributed to reducing 17-18% of the multidimensional poverty rate at 1% statistical significance. Cattle production also decreased income poverty rate for both poor households that had daily income per capita less than or equal to USD1.90 and ultra-poor households which had daily income per capita less than or equal to USD1.25 at 10% or 5% statistical significance.

The results in the last column of Table 4 show that the households with more than one cattle had higher welfare than the households with one or no cattle. This implies that a higher number of cattle had a higher contribution to increased income and thus reduced poverty of poor households.

Determinants of cattle production of the ethnic minority households

Table 5 shows that the household heads with higher education levels or older ages were more likely to raise a larger number of cattle and to participate in cattle production than the others. This is probably because the young household heads were more likely to find other jobs such as non-farm jobs in enterprises, coffee harvest at the big coffee farms, etc. The numbers of motorbikes and mobile phones had positive effects on both the number of cattle raised and the participation in cattle production of the households. This is because motorbikes and mobile phones were important means to support cattle production. Farmers usually used motorbikes to transport grass or go to buy or sell young cattle. Farmers also used mobile phones to get and share information in terms of inputs, disease, and output market for cattle production. Governmental supports did not have any significant effects on the number of cattle, but had positive effects on the household’s participation in cattle production. In other words, governmental supports played an important role in starting cattle production by households.

Table 4. Propensity score matching estimates of the impact of cattle production on welfare of poor households (ATT)

Variable

Matching
algorithm

Without cattle

With more than one cattle

ATT

Prob.

ATT

Prob.

Multidimensional
poverty rate

NNMa

-0.18

<0.001

-0.35

<0.001

KBMb

-0.18

<0.001

-0.38

<0.001

Radiusc

-0.17

<0.001

-0.38

<0.001

Headcount index with the
poverty line of USD1.25

NNMa

-0.10

0.06

-0.17

0.01

KBMb

-0.11

0.04

-0.20

0.004

Radiusc

-0.10

0.05

-0.19

0.01

Headcount index with the
poverty line of USD1.90

NNMa

-0.09

0.01

-0.15

0.002

KBMb

-0.09

0.02

-0.16

<0.001

Radiusc

-0.09

0.01

-0.17

<0.001

The daily income
per capita in USD

NNMa

0.26

0.01

0.39

0.001

KBMb

0.27

0.002

0.44

<0.001

Radiusc

0.27

0.001

0.43

<0.001

Total household income
in thousand USD

NNMa

0.33

0.004

0.43

0.01

KBMb

0.32

0.01

0.51

<0.001

Radiusc

0.31

0.01

0.50

<0.001

Note: Standard errors bootstrapped 1,000 replications only for Kernel matching and Radius matching,
a
NNM = five nearest neighbor matching with common support and replacement,
b
KBM = Kernel matching with common support and band width 0.06,
c
Radius matching with common support and band width 0.06

Regarding shocks and village variables, households with health shocks reduced the number of cattle and the probability that they participated in cattle production. The share of households with access to national electricity had negative effects on the number of cattle. This is because the households with better access to national electricity might move to raise other livestock that needed electricity than cattle such as poultry or to develop off-farm jobs. This implies that cattle production was an important livelihood strategy for ethnic households in the mountain areas without the national electricity system. These findings are in line with previous studies (Ngigi and Birner 2013; Do et al 2019).


Conclusions

Cattle production improved poor ethnic minority households’ welfare in Central Vietnam. A higher number of cattle was associated with the higher welfare of the poor household. Cattle production was positively influenced by the education level of household head, number of mobile phones, and number of motorbikes. Governmental supports were important for a household to participate in cattle production. Health shocks and access to electricity had negative effects on cattle production of the households.

The findings can have several important implications for improving welfare of ethnic minority households in Central Vietnam, especially for policymakers to consider in order to encourage households of ethnic minorities to develop cattle production for their better livelihood. First, poor ethnic minority households, especially extremely poor households, should be supported to raise cattle as this could have a positive impact on their income and poverty reduction. Second, poor ethnic minority households should be supported in coping with health shocks to avoid cattle and welfare losses. This is because cattle are both an asset and the most important livelihood strategy of the households. If the households lose their cattle, they would find it difficult to improve their welfare. Final, it is essential to educate ethnic minority people to improve their cattle production, subsequently increasing their welfare.


Acknowledgments

The authors gratefully acknowledge the financial support from the Vietnam Committee for Ethnic Minority Affairs under the research project entitled “Major solutions for sustainable poverty reduction in ethnic minority and mountainous regions of Vietnam by 2030”, with the research code number of CTDT.43.18/16-20, which is under the CTDT/16-20 program.

Table 5. Determinants of household cattle production

Variable

Generalized Poisson
Regression

Negative Binomial
Regression

Binary Logistic
Regression

Coef.

Prob.

Coef.

Prob.

Coef.

Prob.

Human capital

Ethnic (very small minority = 1; otherwise = 0)

-0.19

0.55

-0.22

0.46

0.12

0.79

Gender (male=1, female=0)

0.07

0.73

0.10

0.60

-0.07

0.81

Age (year)

0.02

0.004

0.02

0.001

0.02

0.02

Education(level)

0.17

0.001

0.18

0.001

0.19

0.03

Household laborers (persons)

0.06

0.38

0.08

0.29

-0.03

0.82

Household size(persons)

0.04

0.47

0.04

0.43

0.21

0.05

Natural capital

Farm land size (ha)

0.03

0.62

0.01

0.92

0.17

0.41

Perennial land size (ha)

0.04

0.42

0.08

0.22

0.19

0.10

Social capital

SPO (Yes=1, No=0)

-0.15

0.37

-0.09

0.58

-0.36

0.24

Subsidy (Yes=1, No=0)

0.11

0.48

0.16

0.26

0.39

0.09

Mobile phones (number)

0.26

<0.001

0.28

<0.001

0.35

0.01

Physical capital

Motorbikes (number)

0.49

<0.001

0.46

<0.001

0.43

0.04

Financial capital

Access to credit (Yes=1, No=0)

0.03

0.86

0.01

0.96

0.05

0.82

Shocks

Weather shocks (number)

0.03

0.57

0.03

0.54

0.00

0.98

Health shocks (number)

-0.21

0.05

-0.24

0.03

-0.45

0.02

Other shocks (number)

0.06

0.46

0.08

0.31

0.05

0.73

Village variables

Road type (Yes=1, No=0)

-0.47

0.22

-0.50

0.20

-0.46

0.44

Distance to town ( km)

0.01

0.46

0.003

0.66

0.02

0.11

Water system (Yes=1, No=0)

-0.01

0.17

-0.01

0.22

-0.02

0.14

Access to electricity (Yes=1, No=0)

-0.36

0.07

-0.36

0.08

0.07

0.87

Province variables

Nghe An

1.71

<0.001

1.73

<0.001

2.35

0.001

Quang Binh

1.40

0.004

1.53

0.001

1.22

0.08

Quang Ngai

1.26

<0.001

1.35

<0.001

1.23

0.02

Kon Tum

1.30

0.001

1.27

0.002

1.31

0.04

Constant

-2.17

<0.001

-2.55

<0.001

-3.30

<0.001

Number of observations

500

500

500

Wald chi2(24)

228

200

69.1

Prob > chi2

<0.001

<0.001

<0.001

Pseudo R2

0.11

0.11

0.12

Note: SPO is social-political organizations; robust standard errors are clustered at the household level; very small minority is an ethnic minority with its total population of less than 10,000


References

Abadie A and Imbens G 2008 On the Failure of the Bootstrap for Matching Estimators, Econometrics 76, 1537-1557.

Ashley C and Carney D 1999 Sustainable livelihoods: Lessons from early experience, London: Department for International Development.

Baulch B, Chuyen T K, Haughton D and Haughton J 2007 Ethnic minority development in Vietnam, The Journal of Development Studies, 43(7), 1151-1176.

Baulch B, Hoa N T M, Phuong N T and Hung P T 2011 Ethnic minority poverty in Vietnam, In Thang (Ed), Poverty vulnerability and social protection in Vietnam: Selected issues (pp. 101-165). Hanoi, Vietnam: Vietnam Academy of Social Sciences.

Baulch B, Pham H T and Reilly B 2012 Decomposing the ethnic gap in rural Vietnam, 1993-2004, Oxford Development Studies, 40(1), 87-117, DOI: 10.1080/13600818.2011.646441.

Beranger C and Vissac B 1993 A holistic approach to livestock farming systems: Theoretical and methodological aspects. In Brossier, J., De Bonneval, L., and Landais, E. Eds, Systems studies in agriculture and rural development, A selection of papers published by researchers in the Agrarian Systems and Development Department of INRA (1990-1994). INRA. Paris.

Caliendo M and Kopeinig S 2008 Some Practical Guidance for the Implementation of Propensity Score Matching, Journal of Economic Surveys, 22, 31-72.

Carney D 1998 Implementing the sustainable rural livelihoods approach. Pages 3–23 in Sustainable Rural Livelihoods: What Contribution Can We Make? D. Carney, ed. Department for International Development, London, UK.

Do T L, Nguyen T T and Grote U 2019 Livestock Production, Rural Poverty, and Perceived Shocks: Evidence from Panel Data for Vietnam, The Journal of Development Studies, 55:1, 99-119, DOI: 10.1080/00220388.2017.1408795.

Dong H S, Vinh N T and Lan C N 2005 Report on main challenges in growth and poverty reduction in the Northern mountain region of Vietnam. Evaluation report of the Development Strategy Institute, Ministry of Planning and Investment. Hanoi, December 2005.

Duc L T 2019 Multidimensional Poverty in Viet Nam: Sustainable Poverty Reduction 2016–2020. Multidimensional Poverty Peer Network, Dimensions, Number 8, December 2019.

Government of Vietnam 2019 Report on “Implementation of policies, laws about the National Targeted Program on Stainable Poverty reduction in ethnic minority and mountainous areas of Vietnam for period of 2012-2018”. Report number 255/BC-CP, issued on June 12, 2019.

GSO 2019 Statistical year book of Vietnam in 2019. General Statistical Office of Vietnam.

Gujarati D N 2004 ‘Basic Econometrics’, Basic Econometrics, Fourth Edition, pp. 620–621.

Ha D N, Hung P V, Huyen N T T, Boney L and Ives S 2014 Impacts of Socio-Cultural Factors on Beef Cattle Value Chain: A Case Study of Producers in the Northwest Region of Vietnam, Proceedings of the 16th AAAP Animal Science Congress Vol. II 10-14 November 2014, Gadjah Mada University, Yogyakarta, Indonesia.

Hoddinott J 2006 Shocks and Their Consequences across and within Households in Rural Zimbabwe, Journal of Development Studies 42, 301–321.

Huyen L T, Herold P, Markemann A and Zárate A V 2011 Resource use, cattle performance and output patterns on different farm types in a mountainous province of Northern Vietnam. Animal Production Science, 2011, 51, 650–661.

Huyen L T, Van D T, Markemann A, Herold P and Zárate A V 2013 Beef cattle keeping by smallholders in a mountainous province of northern Vietnam in relation to poverty status, community remoteness and ethnicity, Animal Production Science, 2013, 53, 163–172.

Imai K S, Gaihe R and Kang W 2011 Poverty, inequality and ethnic minorities in Vietnam. International Review of Applied Economics, 25 (3), 249-282.

Kazianga H and Udry C 2006 Consumption Smoothing? Livestock, Insurance and Drought in Rural Burkina Faso. Journal of Development Economics 79, 413-446.

McMichael A J, Powles J W, Butler C D and Uauy R 2007 Food, Livestock Production, Energy, Climate Change, and Health. The Lancet 370, 1253-1263.

Mogues T 2011 Shocks and Asset Dynamics in Ethiopia. Economic Development and Cultural Change 60, 91-120.

MOLISA 2019 Results of multidimensional poverty rate of households in 2018, the Decision No. 1052/QD-LDTBXH issued by the Ministry of Labour- Invalids and Social Affairs of Vietnam in July 29, 2019.

Moll H A J 2005 Costs and Benefits of Livestock Systems and the Role of Market and Nonmarket Relationships. Agricultural Economics 32, 181-193.

Ness B, Urbel-Piirsalu E, Anderberg S and Olsson L 2007 Categorizing tools for sustainability assessment. Ecological Economics, 60, 498-508.

Ngigi M W and Birner R 2013 Shocks, Livestock Assets and Climate Change Adaptation in Kenya. AAAE Fourth International Conference, September 22-25, 2013, Hammamet, Tunisia. African Association of Agricultural Economists (AAAE)

Nguyen C V 2012 Ethnic minorities in Northern Mountains of Vietnam: poverty, income and assets. MPRA Working Paper 40769.

Nguyen C V, Tran T Q and Van Vu H 2017 Ethnic Minorities in Northern Mountains of Vietnam: Employment, Poverty and Income. Soc Indic Res 134, 93–115.

Nguyen T, Bauer S and Grote U 2016 Does Land Tenure Security Promote Manure Use by Farm Households in Vietnam? Sustainability 8, 178.

Nguyen T, Do T L and Grote U 2018 Natural resource extraction and household welfare in rural Laos. Land Degradation and Development, 29, 3029–3038.

Parsons D, Lane P A, Ngoan L D, Ba N X, Tuan D T, Van N H, Dung D V and Phung L D 2013 Systems of cattle production in South Central Coastal Vietnam. Livestock Research for Rural Development 25 (2) 2013.

Pham H, Le T and Nguyen C 2011 Poverty of the ethnic minorities in Vietnam: Situation and challenges from the P135-II communes. Research report for State Committee for Ethnic Minority Affairs of Vietnam and United Nations Development Program, Hanoi, Vietnam.

Tra H T H, Lebailly P, Cuong V C and Duqesne B 2010 Beef cattle feeding systems and measuring their sustainability in Bac Kan province, the Northern Mountainous Region, Vietnam. Contributed Paper prepared for presentation at the international symposium on Sustainable Land Use and Rural Development in Mountainous Regions of Southeast Asia, Hanoi, 21-23 July 2010.

Tuyen T Q 2014 Determinants of nonfarm participation among ethnic minorities in the Northwest Mountains, Vietnam. MPRA Working Paper 59185. Retrieved from http://mpra.ub.uni-muenchen.de/59158/

UNDP 2018 Multidimensional poverty in Vietnam: Reducing poverty in all its dimensions to ensure a good quality life for all.

Van de Walle D and Gunewardena D 2001 Sources of ethnic inequality in Vietnam. Journal of Development Economics, 65(1), 177-207.

WB 2012 Vietnam poverty assessment - Well begun, not yet done: Vietnam's remarkable progress on poverty reduction and the emerging challenges. Washington DC: The World Bank.

Werner Stür, Khanh T and Duncan A 2013 Transformation of smallholder beef cattle production in Vietnam, International Journal of Agricultural Sustainability, 11:4, 363-381, DOI: 10.1080/14735903.2013.779074.



Appendices


Table A1. Basic socio-economic characteristics of the study areas in 2018

Indicators

Unit

Nghe
An

Quang
Binh

Quang
Ngai

Kon
Tum

Dak
Nong

Total natural land

mills. ha

1.65

0.80

0.52

0.97

0.65

Agricultural land

thous. ha

300

90.1

151

266

360

Forestry land

mills. ha

1.15

0.63

0.30

0.61

0.24

Average population

mills. persons

3.16

0.89

1.27

0.54

0.65

Total number of households

thous. households

937

248

355

132

156

Total number of ethnic minority households

thous. households

109

6.15

50.7

69.2

102

Multi-dimensional poverty rate

%

5.54

6.98

9.39

17.3

13.6

Percentage of the poor ethnic minority householdsa

%

66.2

24.8

54.6

93.6

63.5

GDP per capita

USD1000/year

1.62

1.61

2.60

1.59

1.74

Source: MOLISA (2019); GSO (2019); Government of Vietnam (2019). aComparison of total number of the poor ethnic minority households over total poor households



Table A2. Name and definition of the variables in the regression models

Variable

Definition

Scale

Dependent variables

Cattle production

1 if a household has cattle production, 0 otherwise

Binomial

Number of cattle

No. of cattle of household

Metric

Independent variables

Household level

Human capital

Ethnic

The ethnicity of household head (1 = very small minority; 0 = otherwise)

Binomial

Gender

Gender of household head (1 = male; 0 = otherwise)

Binomial

Age

Age of household head

Metric, in years

Education

Years in school of the household head

Metric, in level

Household laborers

Household laborers

Metric

Household size

The household size (persons)

Metric

Natural capital

Farm land size

Farm land area of household in ha

Metric, in ha

Perennial land size

The perennial land area of household in ha

Metric, in ha

Social capital

SPO

No. of social/political groups

Metric

Subsidy

If household receipt subsidy (1 = yes; 0 = otherwise)

Binomial

Mobile phones

No. of mobile phones of household

Metric

Physical capital

Motorbike

No. of motorbikes of household

Metric

Financial capital

Access to credit

If households with access to credit (1 = yes; 0 = otherwise)

Binomial

Shocks

Weather shocks

No. of weather shocks during the last three years

Metric

Health shocks

No. of health shocks during the last three years

Metric

Other shocks

No. of other shocks during the last three years

Metric

Village level

Road type

Accessible to the village at all times (1 = yes; 0 = otherwise)

Binomial

Distance to town

Distance from home to the nearest town

Metric, in km

Water system

Share of farmland area with available irrigation

Metric, in %

Access to electricity

If households with access to electricity (1 = yes; 0 = otherwise)

Binomial

Province level

Nghe An

If household in Nghe An province (1 = yes; 0 = otherwise)

Binomial

Quang Binh

If household in Quang Binh province (1 = yes; 0 = otherwise)

Binomial

Quang Ngai

If household in Quang Ngai province (1 = yes; 0 = otherwise)

Binomial

Kon Tum

If household in Kon Tum province (1 = yes; 0 = otherwise)

Binomial



Table A3. Cattle production in the studied provinces in period of 2014-2018

Study area

Indicators

Unit

2014

2015

2016

2017

2018

Average
annual
growth
rate (%)

Nghe An

Head count

thous. heads

391

413

426

455

470

4.72

Liveweight

thous. tones

13.7

14.6

15.3

16.4

17.6

6.37

Quang Binh

Head count

thous. heads

89.2

96.1

104

107

105

4.08

Liveweight

thous. tones

5.82

6.15

5.98

6.07

7.17

5.35

Quang Ngai

Head count

thous. heads

274

279

277

277

278

0.32

Liveweight

thous. tones

-

18.0

18.3

18.2

19.0

1.81

Kon Tum

Head count

thous. heads

60.0

62.3

68.2

73.9

77.7

6.68

Liveweight

thous. tones

4.07

4.08

4.19

4.42

4.53

2.71

Dak Nong

Head count

thous. heads

18.1

18.7

27.9

33.8

33.3

16.4

Liveweight

thous. tones

0.93

0.90

1.10

1.45

1.73

16.8

Source: GSO, 2019; Provincial Statistic Offices



Figure A1. Propensity score distribution and common support for propensity score estimation by groups
Note: “Treated: on support’’ presents the households with cattle production that have a suitable match, while ‘‘Treated: off support’’ presents the households with cattle production that do not have a suitable match, and “Untreated” presents the households without cattle production. The basic case is that the group of households with cattle production is Treatment and the group of households with no cattle production is Control. Case 1 is that the group of households with more than one cattle is Treatment and the other households group is Control


Table A4. Multicollinearity test

Variable

VIF

1/VIF

Human capital

Ethnic

2.47

0.40

Gender

1.20

0.83

Age

1.48

0.68

Education

1.56

0.64

Household laborers

2.19

0.46

Household size

2.17

0.46

Natural capital

Farm land size

1.12

0.89

Perennial land size

1.20

0.83

Social capital

SPO

1.60

0.63

Subsidy

1.36

0.74

Mobile phones

1.79

0.56

Physical capital

Motorbike

1.58

0.63

Financial capital

Access to credit

1.44

0.69

Shocks

Weather shocks

1.46

0.68

Health shocks

1.20

0.83

Other shocks

1.56

0.64

Village variables

Road type share

2.40

0.42

Distance to town

1.86

0.54

Water system share

4.76

0.21

Electricity share

1.49

0.67

Province variables

Nghe An

7.07

0.14

Quang Binh

8.14

0.12

Quang Ngai

3.80

0.26

Kon Tum

4.64

0.22

Mean

2.48

Note: VIF is variance inflation factor



Table A5. Quality test for propensity score matching

Group and
matching
algorithm

Pseudo R2
Before
matching

Pseudo R2
After
matching

LR test
(p-value)
Before matching

LR test
(p-value)
After matching

Mean
standardized bias
before matching

Mean
standardized bias
after matching

Percent
bias
reduction

Basic case: Group of households with cattle is Treatment, another group is Control

NNM

0.12

0.01

<0.001

1.00

20.15

3.56

82.3

KBM

0.12

0.01

<0.001

1.00

20.15

4.14

79.5

Radius

0.12

0.01

<0.001

1.00

20.15

3.65

81.9

Case 1: Group of households with more than one cattle is Treatment, another group is Control

NNM

0.25

0.03

<0.001

1.00

30.31

7.16

76.4

KBM

0.25

0.02

<0.001

1.00

30.31

6.58

78.3

Radius

0.25

0.02

<0.001

1.00

30.31

5.11

83.1

Note: a NNM = five nearest neighbor matching with common support and replacement,
b
KBM = Kernel matching with common support and band width 0.06, cRadius matching with common support and band width 0.06