Tracking the Leakage of Development Goods Using iBeacon Technology

Jennifer Hamilton

UCLA

​

Muthoni Ng’ang’a

Wilfrid Laurier University

​

Daniel N. Posner

UCLA

​

Africa Evidence Summit

June 2023

•Governments and donors regularly distribute development goods in poor communities to improve people’s welfare

•Farming inputs, food, pharmaceutical products, bed nets

​

​

•Millions of dollars of these goods go missing or are misallocated

•A major challenge to the aid industry

•A leading preoccupation of development practitioners and academic researchers

​

•Tracking “leaked” goods is important

•For devising strategies to reduce misallocation

•For understanding why distributors diverge from program criteria

•For understanding how leaked goods flow through patronage networks

​

Motivation

• “Leakage” of development goods usually measured via
• Perception-based surveys (Transparency International; World Bank; Olken 2009)
• Comparisons between goods allocated and received (Reinikka and Svensson 2004; Golden and Picci 2005; Olken 2006, 2007; Niehaus and Sukhtankar 2010; Banerjee et al 2018)

​

​

​

• These strategies, even when successful, tell us little about where the “leaked” goods go

Measuring “leakage”

We use iBeacon technology to track the distribution, and potential leakage, of solar lanterns distributed in off-grid communities in Western Kenya

​

​

Project objectives

1. Assess efficacy of iBeacon technology
• Hardware and software
• Protocols for deploying it (ethically)
• Interpreting the data it generates
2. Lean where “leaked” lanterns go
3. Better understand why deviations from program guidelines occur (and, in process, rethink what we mean by “leakage”)

​

Our approach

iBeacon technology

Data from a pilot tracking mission

Field officer’s

walking path

​

• Tagged 244 solar lanterns; distributed

them in ~12 off-grid communities in

Western Kenya

​

• Mapped geo-locations of buildings (n = 2,825) in project sites

​

• Distributed 12-34 lanterns to each village elder (VE)
• Instructed to prioritize households w children < 5 yrs old

​

• Sent FOs on tracking missions with iPhone
• 5 rounds of missions over 8 weeks

​

• Followed up with a household survey (n = 589)

Project design

Distributing tagged development goods raises ethical issues (Hamilton, Ng’ang’a and Posner 2021)

​

1. Tags risk violations of privacy, but disclosure may alter behavior
• Inform VEs that 10% of lanterns distributed in the study contain tracking devices (distribution script)
• Include label on tagged lanterns indicating potential tracking

​

• Tracking missions may arouse privacy concerns
• Justified using a real concurrent (but ostensibly separate) malaria project

​

• Risk of data disclosure
• Blind data collection
• Non-identification of project sites
• Reporting only aggregate results

​

​

Ethical issues

• For detecting the beacons, yes!
• Detected 98.8% of the beacons in at least one tracking round
• Tracking success rates similar to “shoe leather” approach
• 2,250 untagged lanterns distributed in auxillary project sites
• FOs traced 98.3% of lanterns across 48 villages

​

• Represents a substantial improvement in performance over previous generations of technology (de Mel et al. 2016)

​

Did the technology work?

• For matching found beacons to households, results are more complex…

​

​

Did the technology work?

Where did the lanterns go?

Based on preliminary analysis, households with a lantern were…

• More likely to have children under 5
• Less likely to have a connection to the electricity grid
• Less likely to have reliable grid electricity
• Less likely to be food secure
• More likely to have spoken with the VE�(see detailed results)

​

​

​

​

​

​

​

55% of beacons were matched to multiple potential households across tracking rounds

​

​

​

​

​

​

​

​

Lantern movement across rounds

• “Technology” is much more than hardware and software

​

• The technology works well for determining whether goods “leak” out of the system/“go missing”
• A potentially powerful tool for designs testing efficacy of various interventions designed to reduce theft (e.g., Jablonski et al 2022)

​

• The technology works less well for figuring out exactly where the “leaked” goods go

What we learned about iBeacon technology

• Much less “leakage” (deviations from program rules) than we anticipated
• Possibly because of tracking warning, but seems unlikely
• Possibly because lanterns were not very valuable
• Possibly because of top-down oversight (sub-chief)

​

• We may want to rethink what we mean by “leakage”
• Assumed to be a bad thing (malfeasance)
• But our evidence may suggest more innocuous (and welfare enhancing) explanations
1. Local knowledge
2. Second-order allocation

What we learned about “leakage”

Thank you

Additional Slides

We are here as part of the Lighting Kenya Project. In this project, we are distributing solar lanterns to several sub-locations in [redacted] County. We chose this sub-location specifically because we know there is a particularly great need here for lighting after dark.

We want to make sure that the lanterns we distribute remain in the villages in which we distribute them. We have placed tracking devices in some of the lanterns we are distributing and plan to use them to confirm that the lanterns remain where they are supposed to be. The tracking devices only reveal the approximate locations of the lanterns but do not collect any other information. For example, they do not record conversations.

For budget reasons, we are not able to put the tracking devices in all of the lanterns we are distributing. Only 10% of the lanterns we are distributing in [redacted] County have tracking devices. This means that for every lantern that contains a tracking device, there are nine that do not. The lanterns we give you may or may not contain tracking devices.

This is what the solar lanterns look like. [Hold up a lantern and demonstrate its function.] As you will notice, the lantern contains a label explaining that it may contain a tracking device. Again, for budget reasons we were not able to put the tracking devices in all the lanterns that we are distributing in [redacted] County. Only one out of every ten lanterns has such a device.

We have prepared boxes of lanterns for each village elder. The box you will receive contains a number of lanterns determined by the size of your village. Larger villages will receive more lanterns, and smaller villages will receive fewer. We ask that you only distribute the solar lanterns you receive within your own village.

​

Distribution text

We are eager for people in this area to benefit from the lanterns we are distributing, so we request that you distribute them in your villages within one week.

While we recognize that many people in your villages would benefit from receiving a solar lantern, we request that the first priority be households with children under 5 years old. These young children may require help during the night, and tin lamps present a danger for their health and safety. The smoke from tin or kerosene lamps are especially harmful to the health of children. In addition, young children may not understand the danger of tin or kerosene lamps, so they are more likely to knock them over. This can result in fire or burns. Solar lanterns can help prevent these accidents.

Thank you in advance for your participation in the Lighting Kenya Project.

[back]

​

Distribution text (cont’d)

Where did the lanterns go?

*Asset index constructed using principal component analysis for all asset variables, except for bank accounts (owing to missing data). First principal component explains 61% of variation.

**As reported by lived food poverty over the preceding month.

• Consistent with findings in other studies? (Alatas et al 2013; Carlson and Seim 2020; Basurto, Dupas and Robinson 2020)

Where did the lanterns go?

Excludes VE households. DK responses treated as no relationship.