CGIAR Research Program on Policies, Institutions, and Markets Workshop on Rural Transformation in the 21st Century (Vancouver, BC – 28 July 2018, 30th International Conference of Agricultural Economists). Presentation by Solomon Asfaw, GCF-IEU. Co-authors: Antonio Scognamillo, Gloria Di Caprera, Ada Ignaciuk, and Nicholas Sitko.
Heterogeneous Impact of Livelihood Diversification: Cross-Country Evidence from Sub-Saharan Africa
1. Heterogenous Impact of Livelihood
Diversification: Cross-Country Evidence from
Sub-Saharan Africa
Co-authors: Antonio Scognamillo, Gloria Di Caprera, Ada Ignaciuk, and
Nicholas Sitko
PIM Pre-ConferenceWorkshop: RuralTransformation in the 21st century
ICAE 2018 –Vancouver
28 July 2018
Solomon Asfaw, GCF-IEU
2. Motivation and research questions
• Managing risk & income variability
• Adapting to changing weather condition
• Diversification provides safety-net
Push
factors
• Off-farm opportunities
• Higher wage rates & higher returns to
entrepreneurial activities
• Economies of scope
Pull
factors
Lead to lower,
though more
stable, welfare
levels
Impact on welfare
Increase
welfare, but not
necessarily more
stability
Research questions:
What are the linkages of
weather events and
diversification choices?
What are the implications
of crop and income
diversification choices for
household welfare?
Assess heterogeneity of
impact across income
distribution?
3. Crop and income diversification
• We use the Gini Simpson index ranging from 0 to 1 to measure
household-livelihood diversification
- Cropland diversification index: based on number of crop types
planted and the area allocated
- Income diversification index: crop, livestock and fishery
incomes, wage from agricultural and non-agricultural activities and
other incomes received through transfers and remittances
4. Country Year
Type of
data
Data
Sources
Climatic Variables
Malawi
2010 -
2013
Panel
Malawi Integrated Household
Panel Survey (IHPS)
Africa Rainfall Climatology (ARC
2), 1983-2016
Niger
2011 -
2014
Panel
Niger National Survey on Living
Conditions and Agriculture
(ECVM/A)
Africa Rainfall Climatology (ARC
2), 1983-2016
Zambia
2012 -
2015
Panel
Rural Agriculture Livelihood
Survey (RALS)
Africa Rainfall Climatology, (ARC
2), 1983-2016
Data: three countries comparison
6. Empirical strategy
Dit1 = β0 + Xit1βi + ai + uit1
Dit2 = β0 + Xit2βi + ai + uit2
with i=1…N t=1, …, T;
• Dit1, and Dit2 indicate, respectively, crop and income diversification for the household i at time t .
• Xit is vector of explanatory variables affecting the degree of diversification
• ai representing the unobserved individual-level effects.
• uitj represents the observation-specific error in the equation 𝑗.
Estimate the drivers: Seemingly-unrelated regression (SUR) model
7. Malawi Niger Zambia
Crop Income Crop Income Crop Income
Climatic variables
Long-term negative rainfall shocks 0.488** -0.037 0.096 0.156 0.683*** 0.223***
Peer effects
% of crop div. within the EA 0.387*** - 0.590*** - 0.534** -
% of income div. within the EA - 0.366*** - 0.297*** - 0.275***
Human, natural and physical capital
Household head level of education -0.045 -0.502*** 0.016 -0.222** -0.285*** 0.535***
Land size 10.404*** 0.383 0.097*** 0.066 0.134** -0.604***
Wealth Index 12.631*** 9.795*** -3.107 18.073** -4.085*** 13.678***
Institutions and infrastructures
Agriculture Extension officer 2.697** 0.686 3.830*** -1.141 1.337*** 1.245***
Market access (km) -0.132* 0.062 -0.017* -0.004 0.016*** -0.037***
Road access (km) 0.089 -0.022 -0.001 0.003 0.011** 0.003
HH socio demographic YES YES YES YES YES YES
Year and region dummy YES YES YES YES YES YES
Random effects YES YES YES YES YES YES
Simultaneous estimates YES YES YES YES YES YES
Observations 1556 2938 10889
Drivers and constraints for diversification strategy
Results
8. Empirical strategy
Local Average Treatment Effect – Impact of diversification
Dcit = β0 + β1 𝐙kcjt + βiXit + ai + ucit1
Yit = β0 + β1 𝐃cit + βiXit + ai + uit2
• Yit represents the total income of household i at the time 𝑡;
• Dcit represents our endogenous variables
• 𝐙kcjt is a vector of 𝑘 instruments
Identification strategy - IV
• Probability of suffering a negative rainfall shock (Di Falco and Veronesi, 2013)
• Percentage of households in the community adopting the considered diversification
strategy (Townsend 1994)
14. Conclusions
• Drivers are country and strategy-specific.
• Land size, wealth, information availability and the proximity to
diversified farmers are common determinants.
• On average income diversification is a welfare enhancing strategy
in all the countries
• On average, crop diversification increases the household income in
Malawi but the impact turns to be negative in Niger and Zambia.
• The QTE is always positive for the poorest and decreases (or even
turns negative in the case of crop diversification in Niger and
Zambia) moving toward the upper end of the income distribution.
The use of the SPI presents some advantages with respect to other methods. First, in order to identify climate anomalies such as drought or excessive rainfalls, only time-series data on precipitation are required. Moreover, the SPI is an index based on the probability of recording a given amount of precipitation. Since the probabilities are standardized, a value of zero indicates the median precipitation amount, thus the index is negative for drought, and positive for wet conditions. As the dry or wet conditions become more severe, the index becomes more negative or positive, ranging within a commonly-used scale from -2.5 and +2.5 (WMO, 2012). The characteristic of being standardized thus provides a straightforward interpretation and allows for a fully indexed comparison over time and space. In addition, the SPI can be computed for several time scales, ranging from one to 24 months, capturing various scales of both short-term and long-term anomalies. In order to compute our climate shock variables, we first calculate the SPI at 12 months for the reference year 2011. Once the long-run climate anomalies are detected by using the interpretation table provided in WMO (2012), we identify drought and rainfall shocks with dummy variables corresponding to SPI values ranging from less than -2 to more than +2, respectively. Thus, a SPI value of -2.0 or less signals a drought shock while values of +2.0 or more indicates extremely wet conditions.
SPI index calculated using historical series of precipitation during the three months preceding the peak of the rainfall season in each country. Since we excluded the current year 𝑡 from the SPI calculation, the index is expected to be correlated with the diversification index at the household level but to not be correlated with the source of unobserved heterogeneity and to not have any direct effect on current household income. In our framework, we consider only the dry shocks such as droughts. According to McKee et al. (1993, 1995), an extreme dry episode is identified when the SPI index is smaller than -1.5.
The analysis is based on such a restrictive specification to ensure cross-country comparability. Robustness checks using a wider set of controls according to the data availability in each country produce similar results and are available upon request.
Targeting the poorest farmers who do not have access to other instrument to face external shocks is crucial;
A wrong targeting will have no impact or even reduce the welfare of the farmers who would have benefit more from specialization