- 1. ETHIOPIAN DEVELOPMENT RESEARCH INSTITUTERoads, Agriculture
and Welfare:Evidence from a Quasi-ExperimentalSetting in Rural
EthiopiaDavid Stifel Lafayette College & IFPRI ESSP-IIBart
Minten IFPRI ESSP-IIBethlehem Koro EDRI & IFPRI
ESSP-IIEthiopian Development Research InstituteMarch 15, 2012Addis
Ababa 1
2. The Question: What are the benefits of rural feeder roads? 3.
What do we know?Improved rural infrastructure affects...Transport
costsInput costsTimely input availabilityAgricultural productivity
(Minten & Stifel, 2008)Nonfarm production (Binswanger et al.,
1993)Poverty (Lokshin & Yemtsov, 2005; Khandker et al.,
2009)Also, Ethiopian Road Sector Development Program (RSDP) 4.
Measuring Benefits Two Issues1. The measure of benefits Impacts
(accessibility, quality, mobility) Savings in transport costs
Income / Consumption / Poverty impacts2. Reverse causality
Non-random road placementHigh productivity Road constructed ?Road
constructed High productivity ? 5. 1. How to handle causation?
Panel data Dercon et al., 2009 Difference-in-differences Mu and van
de Walle, 2007 Propensity score matching Lokshin & Yemtsov,
2005Rely on estimators to do the work 6. 1. How we handle causation
Quasi-Experiment Sample area selected purposefullyoHomogeneous
regionoExcept for transport costs Households circumstances differ
because ofdifferent transport costs... ...not because of land
characteristics, etc.Let the data to the work 7. Transport Costs
Donkey costs (Birr/kg)oCost of renting donkeyoWeight donkey can
carry Economic transport costsoInclude the opportunity cost of time
8. Average Travel Times andTransport Costs to the Market TownTravel
Time Transport Cost(hours) (Birr/Quintal)Transport Cost
QuintileLeast Remote1.5 18.2Quintile 23.6 40.2Quintile 35.2
52.5Quintile 46.0 60.4Most Remote 6.5 73.4Total 4.5 48.4 9. Is this
a Quasi-Experiment? Is the primary difference between
communitiesdue to transport costs? Compare...o Land
characteristicso Land productivity 10. Characteristics of
Agricultural LandPercent of Land Holding AreaMedian Land Median
Plot HoldingsTanDifficult SteepSize (HA)(HA) Color to Plow
SlopeTravel Cost Quintile Least Remote 0.3 2.09.517.6 6.3 Quintile
2 0.3 1.87.427.816.4 Quintile 3 0.3 1.48.425.812.8 Quintile 4 0.3
1.13.133.115.3 Most Remote0.3 1.33.537.915.0Total 0.3
1.56.428.113.0 11. Altitude of sample households250020001500Meters
1000 5000 020406080 100 Transport Costs (Birr/Quintal) bandwidth =
.8 12. Land Productivity What crops?o Sorghumo Milleto Maizeo
Black/mixed teff Counfounding factors?o Weather and pest shockso
Inputs labor, fertilizer, herbicides, Improved seeds 13. Modern
Input Use Percent of households usingChemical Fertilizer Improved
SeedsAnyDapUrea(maize only)Transport Cost QuintileLeast Remote94.2
94.2 83.075.6Quintile 286.2 86.2 61.431.2Quintile 379.9 78.5
46.515.0Quintile 473.2 73.5 49.312.4Most Remote 71.1 71.7 37.5
9.4Total 81.2 81.1 56.333.3 14. Cereal Yields by Transport
CostSorghum Millet 202015 15Quintals / hectare Quintals / hectare
1010 55 0002040 60801000204060 80100Transport Costs (Birr/Quintal)
Transport Costs (Birr/Quintal)Unadjusted Adjusted for
weatherUnadjustedAdjusted for weatherAdjusted for weather and
inputsAdjusted for weather and inputsMaizeBlack/Mixed Teff 20
2015Quintals / hectare Quintals / hectare1510 10550
0020406080100Transport Costs (Birr/Quintal) 0 20406080100Transport
Costs (Birr/Quintal)Unadjusted Adjusted for weather
UnadjustedAdjusted for weatherAdjusted for weather and inputs
Adjusted for weather and inputs 15. The Setting Stylized
FactsAnnual Household Per Capita Consumption60004000 Birr per
person 2000 0 0 2040 6080 100Transport Costs (Birr/quintal) Total
Food Non-Food 16. The Setting Stylized FactsSchooling by
Transaction Costs Adults (age 15-30)Enrollment ratePercent
withAverage years of schooling(ages 5-15) some schooling Full
sample Those with schoolingTransport Cost Quintiles 41.9 Least
Remote41.91.84.4 39.1 Quintile 232.81.64.8 46.5 Quintile
337.01.74.7 33.2 Quintile 440.71.64.0 32.4 Most Remote 36.81.64.5
Total 36.538.01.74.5 17. The Setting Stylized FactsFood Insecurity
18. 2. Measuring Benefits Previous outcomeso Indicators of cost of
remotenesso Indicators of benefits of reduced transport costs Our
measure Households willingness-to-pay forreduced transport costs
(Jacoby and Minten, 2009) 19. 2. Measuring Benefits Thought
experiment...Compensate a remote household just enoughsuch that
indifferent betweeno Remote ( = 0)o Situation in market town ( = 0)
Estimate this compensation Equivalent variation Willingness-to-pay
20. 2. Measuring Benefits Let household income be defined as...
Households maximize income & utility 21. 2. Measuring Benefits
Benefit is defined by... How can we estimate ??? 22. 2. Measuring
Benefits Differentiating the identity gives us... 23. 2. Measuring
Benefits These are just marginal changes,but if we sum them up
(i.e. integrate)... 24. 2. Measuring Benefits We are interested in
the average benefits...This is just the area under the demand
fortransport tonnage curve. 25. Measuring Willingness to Pay
forTransport Cost Reduction 26. Demand for Transport
Tonnage12501000750 kg5002500 0 20406080 100Transport Cost (Birr/kg)
Total Freight Imported Consumption Agricultural SurplusInput
Purchases 27. Demand for Transport Tonnage Controlling for
Transport Cost Simple Model landholdings DifferenceCoeff
t-statCoeff t-statDiffz-statTotal Freight Transport cost per
quintal -7.9-9.52-6.9-8.51 -1.0 -0.86 Log of HH landholdings (HA)
190.610.46Agricultural Surplus Transport cost per quintal
-3.5-6.06-2.4-4.22 -1.1 -1.31 Log of HH landholdings (HA)
143.711.02Imported Consumption Transport cost per quintal
-2.1-4.39-2.1-4.060.00.00 Log of HH landholdings (HA)16.01.41Input
Purchases-20.7 Transport cost per quintal -2.62-2.5 -20.79 -0.1
-0.80 Log of HH landholdings (HA)33.512.43 28. Non-Farm
EarningsPct. of HH Median NF Percent difference in HHwith earnings*
expenditures between those NF earnings (Birr)w/ and w/o NF
earningsLeast Remote71,00020.0Quintile 2 121,30026.1Quintile 3
131,20022.8Quintile 4 141,18022.2Most
Remote171,10218.4Total121,10222.1* Among those with non-farm
earnings 29. Benefits Estimate Most remote households as accessible
as theleast remote transport costs by 75 Birr / quintal Benefit
3,300 Birr per yearo48% due to agric surplus priceso42% due to
consumption prices 30. Benefit EstimatesFor households inBenefit as
percent ofeach of the followinghousehold consumptionevenly spaced
gridpointsUncorrectedAdjusted*2nd 2.0 2.03rd 5.4 5.34th 6.5 6.55th
6.7 6.76th 7.4 7.27th17.216.98th23.523.09th53.051.8Most
remote60.557.6Average for all households9.39.1* Adjusted for
landholdings 31. Benefits vs. Costs Cost 28 million Birr800,000
Birr / km of gravel road35 km Benefits 10 million Birr per
year1,930 Birr benefit on average5,180 households in survey
areaThree years for accrued benefits to exceed cost 32. Concluding
RemarksEstimate benefits of a rural feeder roadIssues:1. Causality
(endogenous road placement) o Quasi-experimental data set2. What
benefit measure to use o Willingness to pay for transport cost 33.
Concluding Remarks Benefit to most remote HH 60% of HH consumption
Costs of construction recovered in 3 years Final commentsoOnly
rural feeder roadsoPotential non-farm earningsoTransport services
are necessaryoOnly this study area, but informative nonetheless 34.
Thank you