Ex Post Impacts of Improved Maize Varieties on the Poor in Rural Ethiopia Di Zeng, Jeffrey Alwang, George Norton, Bekele Shiferaw, Moti Jaleta, Chilot.

Ex Post Impacts of Improved Maize Varieties on the Poor in Rural Ethiopia

Di Zeng, Jeffrey Alwang, George Norton, Bekele Shiferaw, Moti Jaleta, Chilot Yirga

Ex Post Impacts of Improved Maize Varieties on the Poor in Rural Ethiopia

Di Zeng, Jeffrey Alwang, George Norton, Bekele Shiferaw, Moti Jaleta, Chilot Yirga

Ex ante

Poverty Impact Assessment: Ex Ante vs. Ex PostPoverty Impact Assessment: Ex Ante vs. Ex Post

Poor RichPoverty Line

Predicted incomedistribution

Observed income distribution

Predicted poverty impact

Poor RichPoverty Line

Observed incomedistribution

Counterfactual income distribution

Estimated poverty impact

Ex post

Maize Production in EthiopiaMaize Production in Ethiopia

• A major maize producer in Sub-Saharan Africa

• 19% daily energy contribution (Smale, Byerlee and Jayne, 2011)

• Mainly cropped in central highlands (>93% total yield, Schneider and Anderson, 2010)

• Over 40 improved varieties released since 1970s (hybrid and OPV)

Data DescriptionData Description

• Four regions surveyed in 2010

• 1,359 households with 2,443 maize plots• 564 adopters, 535 non-adopters, and 260 partial adopters

• 43.3% of maize area under improved varieties

• Woreda-level monthly precipitation datal for the past 5-10 years from National Meteorology Agency of Ethiopia

Tigray

Amhara

OromiaSNNPR

Kernel Density of YieldsKernel Density of Yields

• Normalize the utility from local varieties to zero, and denote the utility from improved varieties as

• The decision rule of adoption

• The potential outcomes (Rubin, 1974) in logarithm form are

or

• The generalized Roy model (Heckman et al., 2006)

Empirical SpecificationEmpirical Specification

• Endogenous adoption decision: IV methods

• Homogeneity• Probit-2SLS (Wooldridge, 2002)

• Selection model (Heckman, 1979)

• Heterogeneity• Marginal treatment effect via semiparametric local IV estimation (Björklund and

Moffitt,1987; Heckman et al., 2006)

• Obtain estimates of percentage yield increase (treatment effect)

Treatment Effect EstimationTreatment Effect Estimation

Welfare Changes: the Economic Surplus ModelWelfare Changes: the Economic Surplus Model

Welfare Changes: Small Open EconomyWelfare Changes: Small Open Economy

• Directly estimated at household level

• Plot level income change:

• Aggregated to household:

• ΔCik — IV cost function estimation

• Counterfactual income distribution computed

ikikikikobsik

obsikik CYPCPYCPYI **

k ikiki CYPI

STEP 1: Estimate market-level economic surplus changes

• The k-shift (Alston et al., 1995)

• The counterfactual price level (elasticities synthesized from literature)

• The aggregate surplus changes

Welfare Changes: Closed EconomyWelfare Changes: Closed Economy

STEP 2: Allocate market level surplus changes to households

• Decomposition of ΔPS

• ΔPSprice — allocated to all maize sellers by market shares

• ΔPSyield — allocated to all adopters by the yield increases' shares

• ΔCS — allocated to all maize buyers by purchase shares among total supply

• Counterfactual income distribution computed

Welfare Changes: Closed EconomyWelfare Changes: Closed Economy

)5.01( where ** ZZQPPSPSPSPS pricepriceyield

• Foster-Greer-Thorbecke (FGT, 1984) poverty indices calculated for both observed and counterfactual income distributions

• The differences are poverty impacts

Poverty ImpactsPoverty Impacts

Instrumental VariablesInstrumental Variables

• Production

• Rainfall intensity of the sowing month

• Local population density

• Distance to the nearest agricultural extension office

• Temporary seed supply shortage (yes / no)

• Cost• Rainfall intensity of the sowing month

• Distance to the nearest agricultural extension office

Yield Impact: Mean EstimatesYield Impact: Mean Estimates

ATT EsimatesRobustness

checkProbit-2SLS

Selection LIV

C-D .474** .551*** .662***

Translog .552*** .584*** .514***

PSM-NN .419***

PSM-Radius .442***

PSM-Kernel .454***

Partial Adopter FD: C-D .386***

Partial Adopter FD: Translog .409***

Yield Impact: MTE EstimatesYield Impact: MTE Estimates

C-D technology Translog technology

Other Parameter EstimatesOther Parameter Estimates

• Cost increase due to adoption — 32.5%

• The k-shift — 39.1% cost reduction per kilogram

• Elasticities• ε — 0.5• η — -1

• Aggregate impacts• ΔPS in small open economy — 135.9 thousand USD• ΔPS in closed economy — 101.3 thousand USD• ΔCS in closed economy — 50.7 thousand million USD• Only 6.37% sold maize is consumed by surveyed households

Poverty Impacts: Small Open EconomyPoverty Impacts: Small Open Economy

Poverty Line FGT IndexPoverty impact

under Homogeneity

Poverty impact

under Heterogeneity

$1

Headcount .0095 .0088

Depth .0029 .0032

Severity .0015 .0017

$1.25

Headcount .0103 .0089

Depth .0042 .0045

Severity .0023 .0025

$1.45

Headcount .0103 .0118

Depth .0049 .00453

Severity .0029 .0031

Poverty Impacts: Closed EconomyPoverty Impacts: Closed Economy

Poverty Line FGT IndexPoverty impact

under Homogeneity

Poverty impact

under Heterogeneity

$1

Headcount .0110 .0066

Depth .0048 .0031

Severity .0027 .0019

$1.25

Headcount .0162 .0089

Depth .0064 .0040

Severity .0038 .0025

$1.45

Headcount .0147 .0081

Depth .0073 .0047

Severity .0046 .0030

Further InterpretationFurther Interpretation

• Individual level• A typical adopter with average maize area (0.39 ha) observe 440.5 kg yield

increase

• Such an adopter observe an income increase of 45.6 - 72.4 USD (evaluated using average per-capita maize consumption)

• Population level• Sensitivity analyses lend credence to previous estimates

• 0.7 - 1.2 percentage headcount poverty reduction means 0.48 - 0.83 million rural people have escaped poverty

• A major achievement

Further Interpretation: Producer BenefitsFurther Interpretation: Producer Benefits

Concluding RemarksConcluding Remarks

• Maize research and variety diffusion has had a substantial effect on poverty in rural Ethiopia

• The poor benefit the least from maize technologies due to resource constraints: still much room for micro-level policies to work

• Methodological remarks

ReferencesReferences

• Smale, M., D. Byerlee, and T. Jayne. 2011. Maize revolutions in Sub-Saharan Africa. World Bank Policy Research working paper. No. WPS 5659.

• Schneider, K., and L. Anderson. 2010. Yield Gap and Productivity Potential in Ethiopian Agriculture: Staple Grains & Pulses. Evans School Policy Analysis and Research (EPAR) Brief No. 98.

• Rubin, D. 1974. Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies. Journal of Educational Psychology 66: 688-701

• Wooldridge, J. 2002. Econometric Analysis of Cross Section and Panel Data, MIT Press.• Heckman, J.J., S. Urzua, and E. Vytlacil. 2006. Understanding Instrumental Variables in Models with

Essential Heterogeneity. The Review of Economics and Statistics 88: 389-432.• Heckman, J.J. 1979. Sample Selection Bias as a Specification Error. Econometrica 47: 153-61.• Björklund, A., and R. Moffitt. 1987. The Estimation of Wage and Welfare Gains in SelfSelection Models.

Review of Economics and Statistics 69: 42-49.• Alston, J.M., G.W. Norton, and P.G. Pardey. 1995. Science under Scarcity: Principles and Practice for

Agricultural Research Evaluation and Priority Setting. Ithaca, NY: Cornell University Press.• Foster, J., J. Greer, and E. Thorbecke. 1984. A Class of Decomposable Poverty Measures. Econometrica

52: 761-766.

Thank you.