Does Basic Energy Access Generate Socio-Economic Benefits? … · ContextExperimentFindings Does...
Transcript of Does Basic Energy Access Generate Socio-Economic Benefits? … · ContextExperimentFindings Does...
Context Experiment Findings
Does Basic Energy Access GenerateSocio-Economic Benefits?
A Field Experiment with Distributed SolarPower in India
Michaël Aklin Patrick Bayer S.P. Harish Johannes Urpelainen
IGC Energy and GrowthNovember 13, 2015
Context Experiment Findings
Motivation
300-400 million people without basic householdelectricity in India2001-2011: household electrification increased from55% to 67%Kerosene still used by 31% the population in 2011 asprimary lighting fuelSolar remains below 1%
Context Experiment Findings
Rural Electrification: Literature
Rural electrification as a force of socio-economicdevelopment (World Bank 2008; Bernard 2010;Dinkelman 2011; Lipscomb et al. 2013)Impact evaluation of off-grid solar power missingFeasibility and effects of off-grid power (Brass et al.2012; Palit et al. 2014; Urpelainen 2014)
⇒ No reason to believe comparable to grid connection
⇒ Existing randomized controlled trials focus on lanterns,exclude commercial consideration
Context Experiment Findings
Roadmap
Mera Gao Power’s business modelResearch designProgress reportPreliminary results
Context Experiment Findings
Mera Gao Power: Village by Village
Context Experiment Findings
Mera Gao Power: A Home Run?
Context Experiment Findings
Mera Gao Power: Business Model
2× LED with mobile charging, price INR 100 / month(USD 1.5)Simple, robust, inexpensive DC system without bells orwhistlesInstallation cost of $800-1,000, covers ∼20 households(10 required for installation)More expensive but more reliable than grid electricityin Uttar PradeshMost households in a typical rural village can affordthe payment2,000 systems in 1,500 habitations
Context Experiment Findings
Benefits of Solar Household Lighting
Improvement over kerosene: quality, cost of lighting,health/safetyEnables small household business: roadside vendors,saree decoration, livestock handlingNo financing needed, subsidies unnecessary, easy toscale
Context Experiment Findings
Limitations
Productive loads unavailable: cost can be 10-20times higherNot useful for irrigation, industry, et cetera
Poor villagers pay more than wealthy urbanites
Context Experiment Findings
Experimental Design
Sample of 101 unelectrified habitationsSuratganj block in Barabanki district of Uttar Pradesh1,597 householdsDeployment of microgrids at habitation level
50 treatment villages31 control villages nearby20 control villages outside Suratganj
Baseline (0), customer (3), midline (6), endline surveys(12)
Context Experiment Findings
Study Setting
35% literacy rateMedian monthly household expenditure: INR 4,000Median monthly kerosene expenditure: INR 100
Context Experiment Findings
Measurement for Impact Evaluation
1st Order: kerosene expenditures (PDS, privatemarket)1st Order: availability of electricity (connection,hours)
2nd Order: income, savings, home business2nd Order: children’s study habits and educationalattainment
3rd Order: social capital, female empowerment,public opinion
Context Experiment Findings
Treatment Assignment
Treatment AssignmentRemote Control: 20/101Local Control: 31/101Assigned to Treatment: 50/101
RealizationInstalled as per Treatment: 21/101Installation by Error: 4/101Flooding: 5/101
CustomersBaseline: 0Midline: 132Endline: 93
Context Experiment Findings
Geographical Representation
#*")
#*
") #*
")
!(
!(!(")
#*")
!(!(
!(")
#*!(
#*
")
")#*
")
")
")
#*
")")")
!(
#*
#*
#*
!(")
#*!(
#*
!(#*
")
") !(!(!(!(
!(
!(
!(
!(!(
!(!( #*
!(")#*
!(!(
#*
#*#*!(#*
#*#*
!(!(
#*")")
!(
#*
")
")!(
!(
!(
!(
!(
!(
!(
!(!(
!( !(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!( ^
!( Control") Installed on Purpose") Installed Accidently#* Lack of Demand!( Flooding!( Remote Habitation/
Context Experiment Findings
Estimation and Diagnostics
ITTYi,h,t = αi + ωt + τ ITT Treatmenth,t + εi,h,t
LATE
Installedh,t = βi + θt + δTreatmenth,t + µi,h,t
Yi,h,t = αi + ωt + τ LATE ̂Installedh,t + εi,h,t
Baseline balance: XPower analysis: X
Context Experiment Findings
Results: PDS Kerosene Expenditures
Intent to Treat Local Average Treatment Effect
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)OLS FE FE FE FE IV IV IV IV IV
Treat 0.42 -0.20 4.72 1.33 2.91 0.91 -0.44 2.51 2.88 6.26(1.66) (1.58) (2.78) (2.16) (2.65) (3.57) (3.41) (4.87) (4.64) (5.71)
HH FE X X X X X XWave FE X X X X X XRemote X X X
N 4761Standard errors in parentheses (clustered by habitations).∗ p < 0.05, ∗∗ p < 0.01
Context Experiment Findings
Results: Private Kerosene Expenditures
Intent to Treat Local Average Treatment Effect
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)OLS FE FE FE FE IV IV IV IV IV
Treat -22.2∗∗ -33.2∗∗ -18.6∗∗ -6.6 -17.5∗ -47.9∗∗ -71.6∗∗ -19.5∗ -14.2 -37.6∗
(3.2) (3.3) (5.3) (5.7) (7.3) (8.7) (10.3) (9.5) (12.0) (15.8)HH FE X X X X X XWave FE X X X X X XRemote X X X
N 4761Standard errors in parentheses (clustered by habitations).∗ p < 0.05, ∗∗ p < 0.01
Context Experiment Findings
Private Kerosene Spending among Compliers0
.005
.01
.015
.02
.025
Den
sity
0 100 200 300Rupees per Month (Private Market)
Pre-Treatment Midline Endline
Context Experiment Findings
Results: 1st Order Effects
LATE
Electricity access (%): 23-78 percentage points
Hours of electricity (0-24): 1-4 hours
Context Experiment Findings
Results: 2nd Order Effects
Null Results
Savings (INR / month)Expenses (INR / month)Home Business (Yes/No)Work Time (hours)Lighting for Study (Yes/No)Phone Charging (INR / month)
Context Experiment Findings
Conclusion
Adoption of MGP service: strong and robust 1st-ordereffects
No socio-economic transformation
No subsidies are needed
Modest cost reductions could significantly improveimpact
Maximizing impact: larger systems, grid extension,complementary interventions