Estimating Impact or rural electrification programs in Ethiopia Maximo Torero International Food Policy Research Institute ([email protected]) With the.

Estimating Impact or rural electrification programs in Ethiopia

Maximo ToreroInternational Food Policy Research Institute([email protected])

With the collaboration of: Sebastian Galiani (UDESA), and Tanguy Bernard (IFPRI)

Impact evaluation in EthiopiaFriday, July 20th, 2006

Background: Accelerated Electricity Access Expansion Ethiopia is implementing an ambitious plan to expand the electric

system to both support the high demand growth and to extend access to the rural population at a very accelerated pace.

The objective of the UEAP is to electrify rural towns, as well as villages, through extension of the grid. These rural towns/villages range in size from 70 to 4000 households.

Beneficiaries include residential, commercial and industrial customers (such as flour mills, irrigation centers, water pumping, and telecommunications).

The initial objective of the UEAP is to electrify approximately 50% of rural towns over five years, with a long term view to connecting virtually all towns and villages to the grid in a 10-year horizon.

Background: Accelerated Electricity Access Expansion The Access Expansion component would:

Connect about 382 towns to the grid (through substation expansions and installation of sub-transmission lines (typically 33kV)), and installation of distribution systems in the towns (through transformers and LV lines);

Connect customers to the distribution systems in the towns electrified under component 1a (including installation of meters and drop down wires to homes);

Install efficient public street lighting in towns electrified; and Distribute CFLs (compact fluorescent lamps) to rural

customers to foster energy efficiency and therefore affordability.

How we can design the impact evaluation for the electricity access expansion program

Objective of impact evaluation The objective of the impact evaluation is to

measure the impact that the AEAE Program has on the welfare of the population, and

to investigate whether specific interventions are more effective than others in achieving positive outcomes.

Main questions to answer

Questions aimed at understanding the overall impact of AEAE on socioeconomic development, namely: What is the impact of expanded access and use of electricity on

household welfare and local economic development? What is the impact of introducing energy efficient technology on

uses of electricity?

Questions aimed at understanding the relative effectiveness of alternative program features in achieving the objectives of the project. What is the impact of alternative financing mechanisms on

household connection rates? What is the optimal subsidy to CFLs to maximize CFL usage per

dollar spent? What is the impact of alternative modes of delivery of CFLs

(public distribution, private, vouchers)?

Distribution of observations by

control and treatment groups

• Identification of 17 substations

• Identification of the towns that fall in the 100k radius

• Identification of towns that do not have access to electricity

• feasibility study (?)

Distribution of observations by

control and treatment groups

• Identification of 17 substations

• Identification of the towns that fall in the 100k radius

• Identification of towns that do not have access to electricity

• Non parametric matching to control for any additional selection bias as a result of the feasibility study

Statistical and Econometric Methodology: Two step matching – Base Scenario

1. TOWN MATCHNG:Among towns selected with and without electricity, find comparable towns regarding the following criteria: (1) Comply same selection criteria (2) Access to public phone; (3) type of road; (4) secondary school in town; (5) quintile of population; (6) region.

2. HOUSEHOLD MATCHING (PROPENSITY SCORE MATCHING):Within comparable towns, households with electricity are matched to those without electricity. Propensity score matching (kernel) is performed using the following variables: HH size, demographic composition, ethnicity of HH head (self-reported), age of HH head, education of HH head, sex of HH head, access to water public network, and other possible controls.

Town A (with access)

Town B (no access)

Town C (no access)

Town D (no access)

Town E (no access)

Town F (no access)

Town A (access)

Town B (no access)Propensity score matching of households among comparable towns

Town F (no access)

Proposed Randomized experiments:1. Randomization on the last mile connection within identified treated and control towns Objective: to randomize through incentive mechanisms the connection from

the mainline of the town to each of the households and business Instrument to be use: a discount voucher (could be a discount in the loan)

to pay the last mile connection. Procedure:

Between 10-50% of eligible survey respondents will be randomly assign vouchers for a discount charge to cover the connection of the hh or business

Varying amounts of the offer will be distributed in equal proportions among voucher recipients (we are proposing 2 to 3 sizes of vouchers)

Vouchers will have a period of duration (4 months) In a subsample of towns randomly selected from the pool of the 382 intervened

towns no vouchers will be distributed. Methodologically:

Vouchers will serve as an IV for intensity of electricity access. We need to assure treatment and control groups are demographically

representative within each of the possible treatment amounts. Difference in difference estimators.

Proposed Randomized experiments:1. Randomization on the last mile connection within identified treated and control towns Potential problems:

How to define the size of the voucher – could be a WTP battery of questions in the baseline survey.

Local providers must comply with protocol established for the distribution of the vouchers.

How to avoid contagious effect – i.e. households that did not receive vouchers will be expecting them in the future and therefore will not buy the connection – possibility to announce the lottery publicly and clearly point out there is X number of vouchers.

Competitive underground market of vouchers – try to include identity in vouchers and train providers to match id of individuals to vouchers. We will also include random checks by undergraduate research assistants from the local universities.

Proposed Randomized experiments:1. Randomization on the last mile connection within identified treated and control towns Potential problems:

Problems of non compliance – we will check this by examining whether the level of vouchers distributed within the village predict village-level electricity access but not household use. Alternatively we can use random assignment as an IV for actual assignment

(Duflo and Kremer 2003; Angrist and Krueger, 2001). Although initial assignment does not guarantee that someone actually receives treatment, in most circumstance assignment is nonetheless correlated with participation.

Also refer to local treatment effects paper of Imbens and Angrist (1994), i.e. it is possible to obtain the local average treatment effect estimate by comparing outcomes in the initially assigned group and scaling up the difference by dividing it by the difference in the probability of receiving the treatment in those two groups.

Political feasibility of limited number of vouchers – implement a sequencing in the distribution.

Impact Indicators: Impacts of access and better quality of electric service

n

i i

ii

n

i i

ii

n

i i

ii l

ySlL

l

ySlLL

l

ySlY

111

Change in the proportion of time for activity “i”

Change in the total number of hours worked by the

householdInteraction effect

Impact Indicators: Impacts of access and better quality of electric service

Total hours of work – household Hours of work – household (inc. chores) Hours of work – individual Hours of work – individual (inc. chores) % hours of non-ag work household % hours of non-ag work household (inc. chores) % hours of non-ag work individual % hours of non-ag work individual (inc. chores) All above by gender

Impact Indicators: Changes in the quality of the electricity service

Expenditure in electricity (proportion of total energy sources)

Expenditure in electricity (proportion of total expenditure)

Number of failures Price Sources of energy

Proposed Randomized experiments:2. Randomization of the provision and distribution of CFLS (Compact fluorescent lamps) Objective: implement randomized trials to try to answer four sets of

questions: What is the willingness to pay of CFLs by households? What is the optimal subsidy to CFLs to maximize CFL usage per dollar

spent? What is the impact of alternative modes of delivery of CFLs (public

distribution, private, vouchers)? How arbitration of replaced traditional bulbs by CFLs can be avoided?

Major advantage: this is a pilot so we can even randomized the towns in where to test.

Proposed Randomized experiments:2. Randomization of the provision and distribution of CFLS (Compact fluorescent lamps) Instrument to be use: a discount voucher (could be a discount in

the loan) to pay the last mile connection. Procedure:

Between 10 to 50% of eligible survey respondents will be randomly assign vouchers for a discount charge to cover the cost of the CFL

Varying amounts of the offer will be distributed in equal proportions among voucher recipients. The quantity of vouchers will vary between towns.

Vouchers will have a period of duration (1 month) In a subsample of towns we will also include in addition to the vouchers a

promotion to buy the old bulbs (the amount will be established on the baseline through a willingness to receive battery of questions)

In a subsample of towns randomly selected for the pilot no vouchers will be distributed.

Thanks!