NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
The impact of prepaid electricity meters on low income households in Cape Town South Africa
Initial registration date
December 16, 2014
December 11, 2016 7:55 AM EST
UC Santa Barbara
Other Primary Investigator(s)
University of Cape Town
Additional Trial Information
Prepaid water and electricity meters offer a promising solution to lumpy and unpredictable bills by allowing customers to choose the timing and quantity of purchases. At the same time, prepayment circumvents debt accumulation, allowing utility companies to serve otherwise high-cost customers. Researchers will collaborate with the municipal utility in Cape Town, South Africa to evaluate customer responses and revenue recovery for prepaid versus credit electricity metering. The study generates new evidence on the impacts of prepaid meters on energy use and revenue recovery using a randomized phase-in of new prepaid customers. Researchers will explore how prepaid meters change the customer experience and the municipality’s revenue recovery.
Jack, Kelsey and Grant Smith. 2016. "The impact of prepaid electricity meters on low income households in Cape Town South Africa." AEA RCT Registry. December 11.
The City of Cape Town switched approximately 4000 households from credit (post paid) meters to prepaid meters. These households have been identified based on property value. Household meter conversion was involuntary. The meter installation is not accompanied by any additional information about electricity tariffs or consumption.
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
We will use two data sources to construct outcomes. First, we will use administrative data on electricity consumption (credit households) and electricity expenditures (prepaid households). These are official records from the City of Cape Town’s Electricity Department and provide outcomes at the customer level. We will also construct revenue and revenue recovery measures to analyze the impact of the meter on the City’s budget.
Second, we will gather survey data from a sub-sample of households that are part of the City’s prepaid installation project. The survey will be conducted in two waves, one after roughly half of the households have been targeted for conversion and a second after the installation project is complete. This will generate three groups of households for comparison: credit households (wave 1 only), households recently switched to a prepaid meter (waves 1 and 2) and households less recently switched to a prepaid meter (wave 2 only). The survey data will be used to analyze outcomes associated with four groups of hypotheses for why and how households are affected by prepaid electricity meters: (1) the implicit cost of electricity, (2) greater flexibility in expenditures, (3) salience of consumption, and (4) other psychological channels.
Primary Outcomes (explanation)
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
In collaboration with the researchers, the City will phase in the prepaid meters according to a randomized phase in design. The randomization unit is a block of approximately 200 geographically contiguous households. The resulting blocks are assigned a random order and meter installation proceeds accordingly. Not all meters are installed according to the schedule. “Hold out” households receive meters late, and other households request installation before their randomization block is reached. Thus, compliance with treatment order is imperfect in both directions.
The study timeline is compressed by the demands of the City, which requires all prepaid conversions in Mitchells Plain to be complete by February. Consequently, the lag between the initial phase in block and the final phase in block will be short (~3 months), allowing for only short run impacts to be analyzed with the phase in design.
Analysis of outcomes from the administrative data will be conducted both with treatment assignment on the righthand side (ITT) and with the assigned treatment order as a predictor of actual treatment date (TOT). Using the panel nature of the data to introduce household fixed effects, a difference in difference regression will compare electricity usage outcomes (consumption or expenditure) between credit, newly prepaid and less newly prepaid households before and after the treatment date. Analysis of the survey outcomes will follow a similar design but rather than a continuous treatment timing variable, the survey will assign households to treatment (prepaid) and control (credit): their status at the time of the first survey wave.
Experimental Design Details
The randomization blocks are assigned a random order using a random number generator in Excel.
The randomization unit is a group of approximately 200 geographically contiguous households identified by the City for prepaid meter installation. Blocks were identified by the researchers using GIS to map the project households and identifying groups of approximately 200 households based on spatial coordinates. Other features of the map, such as roads, were not taken into consideration, making block boundaries largely arbitrary.
Was the treatment clustered?
Sample size: planned number of clusters
Sample size: planned number of observations
Sample size (or number of clusters) by treatment arms
Phase in design: all clusters have both treated and control status at some point in time.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
INSTITUTIONAL REVIEW BOARDS (IRBs)
University of Cape Town
IRB Approval Date
Massachusetts Institute of Technology
IRB Approval Date
IRB Approval Number
IRB Approval Date
IRB Approval Number
Post Trial Information
Is the intervention completed?
Intervention Completion Date
April 30, 2016, 12:00 AM +00:00
Is data collection complete?
Data Collection Completion Date
September 30, 2016, 12:00 AM +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
Phase in design over 27 clusters
Reports, Papers & Other Materials
The standard approach to recovering the cost of electricity provision is to bill customers monthly for past consumption. If unable to pay, customers face disconnection, the utility loses revenue, and the service provision model is undermined. A possible solution to this problem is prepaid metering, in which customers buy electricity upfront and use it until the prepaid amount is consumed. We use data from Cape Town, South Africa to examine the effects of prepaid electricity metering on residential consumption and returns to the electric utility. Over 4,000 customers on monthly billing were involuntarily assigned to receive a prepaid electricity meter, with exogenous variation in the timing of the meter replacement. Electricity use falls by about 13 percent as a result of the switch, a decrease that persists for the following year. This creates a tradeoff for the utility: revenue from consumption falls but more of it is recovered on time and at a lower cost. The benefits to the electric utility outweigh the costs, on average, though results are very heterogeneous. Poorer customers and those with a history of delinquent payment behavior show the greatest improvement in profitability when switched to a prepaid meter. These findings point to an important role for metering technologies in expanding energy access for the poor.
Jack, B.K. and G. Smith, (2016) "CHARGING AHEAD: PREPAID ELECTRICITY METERING IN SOUTH AFRICA," NBER Working Paper w22895.
REPORTS & OTHER MATERIALS