Nudging towards micro-savings: A step-wedge experiment on LPG adoption in rural Tanzania

Last registered on May 10, 2022


Trial Information

General Information

Nudging towards micro-savings: A step-wedge experiment on LPG adoption in rural Tanzania
Initial registration date
January 02, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
January 04, 2022, 5:01 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
May 10, 2022, 3:29 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.


Primary Investigator

University of California, Berkeley

Other Primary Investigator(s)

PI Affiliation
University of California, Berkeley
PI Affiliation
University of California, Berkeley
PI Affiliation
University of California, Berkeley
PI Affiliation
University of California, Berkeley

Additional Trial Information

In development
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Sustainable Development Goal 7 calls for “access to affordable, reliable, sustainable, and modern energy for all”. Affordability is one of the largest barriers to providing this access to the 2.6 billion people relying on solid, smoky fuels. This work evaluates the impact of a micro-savings intervention aimed at achieving higher rates of consistent use of clean fuel, in this case, liquified petroleum gas (LPG). We propose to conduct a step-wedge randomized control trial (SW-RCT) in rural Tanzania to evaluate household consumption of liquified petroleum gas before and after enrollment in a savings bank program. We will respectively address the quantitative treatment effect estimates and the qualitative observation and interview results of this SW-RCT.
External Link(s)

Registration Citation

Gill-Wiehl, Annelise et al. 2022. "Nudging towards micro-savings: A step-wedge experiment on LPG adoption in rural Tanzania ." AEA RCT Registry. May 10.
Sponsors & Partners


Experimental Details


Sustainable Development Goal (SDG) 7 calls for the adoption and continued use of clean-burning stoves by the 2.6 billion people relying on solid, smoky fuels. Universal access to clean cooking fuel would help prevent up to 2.3 million annual untimely deaths that are attributed to household air pollution (HAP) —a toll that disproportionately affects women and children—and reduce carbon emissions. However, to date, the clean cooking literature has found low rates of efficient stove adoption. Barriers include: affordability (initial stove and recurring fuel costs), unreliable supply, social acceptability, household education levels, household socio-economic and demographic characteristics, and low total perceived benefits. In Tanzania, the Energy and Water Utilities Regulatory Authority is promoting Liquified Petroleum Gas (LPG), but only 4.4% of the population is using LPG, and even fewer are using it consistently. To meet SDG7, the global community must act quickly to overcome barriers to LPG adoption and sustained use.
In 2018, Gill-Wiehl piloted the use of a locally-trained Community Technology Worker (CTW) to support and encourage behavior change to switch from dirty fuels to LPG in rural Tanzania. That pilot resulted in high rates of LPG adoption (~80%); however, liquidity constraints remained a major barrier to regular refilling. The households in the pilot expressed a desire and need for a place to save for their monthly refill. Additionally, affordability issues and gender norms lead to a few-day gap between clean fuel refills, and thus further unclean fuel use. Microcredit is the most commonly cited instrument to address issues of affordability; however, repaying is often burdensome, especially for women, who need a safe place to save, but do not necessarily need to borrow. To address affordability concerns, the literature has focused mainly on the upfront cost of the LPG stove, not on saving for the recurring fuel costs or reducing the refill gap. The challenge to refill a clean stove (on time) is an opportunity to promote small savings and to increase consistent use. This research asks: How does a microsaving option for the refill fuel cost coupled with local outreach affect LPG adoption? This question addresses the key gap between purchasing an LPG stove (with or without credit) and purchasing the refills that enable regular LPG use.
We propose to conduct a year long step-wedge randomized control trial (SW-RCT) in order to evaluate, understand, and compare rates of adoption of LPG of households with the CTW model before and after enrollment in a savings bank. We will combine quantitative and qualitative methods to analyze (i) the adoption, consistent use, and refilling patterns of households before and after their participation in a savings bank and (ii) behavioral aspects at the individual, communal, and market level that affect LPG adoption. The successful completion of this research will result in the Intent to Treat (ITT) Effects and the observed treatment effects of the savings bank on stove adoption metrics described in detail below, and the revelation of qualitative behaviors and themes that drive LPG adoption.

Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Our key outcome variables relate directly to our four hypotheses:
1) Participating in a savings bank will lead to more purchased LPG refills over the time period.
2) Participating in a savings bank will lead to shorter lag times between cylinder refills.
3) Participating in a savings bank will lead to households using LPG exclusively.
4) Participating in a savings bank will lead to households using less firewood and charcoal
Therefore, our outcome variables are 1) the number of LPG refills over the study period, 2) the refill gap (in days) between when the cylinder runs out and is refilled, (3) the percentage of households exclusively using LPG, and (4) the change in the amount of firewood/charcoal used by each household (as compared to the baseline).
Primary Outcomes (explanation)
Total Annual Refills
This is a continuous variable that will be constructed after all surveys have been collected and will sum the number of LPG cylinder refills a household has purchased throughout the study over each household’s time during control and treatment. We define a refill as exchanging either small cylinder. There is no refilling station in Shirati, but only retail point in which consumers can exchange cylinder. We account for time in the analysis.

Lag Time Between Refills
This dependent variable will be a continuous variable regarding the interval in days between when a household has run out of both LPG cylinders and when the household purchases a refill of at least one cylinder.

Exclusive LPG Use
This dependent variable will be a binary variable in which 1 indicates exclusive LPG use for all cooking activity (excluding water boiling for drinking water). We will also track if cooking is performed indoors or outdoors. We will also track categorically the level of LPG use. Pilot data has shown that the majority of households prepare tea and two meals a day. Beyond exclusive users as defined above, we define habitual users as households that continually refill their LPG cylinder and utilize LPG for at least at half of their meals. We define an occasional LPG user as a household that uses LPG just for tea or less than half of their cooking demand (e.g., only to make vegetables, but not the ugali, a staple dish). We will track these levels of LPG use through binary variables for each category (No LPG Use, Occasional LPG Use, Habitual LPG Use, and Exclusive LPG Use). These variables will be self-reported, but also cross checked with the household’s responses to survey questions on how much they cook with LPG each day and fuel expenditures. Regardless of the household’s reported usage, the CTW will photograph biomass fuel inventory (e.g., take pictures of any stored biomass in home), look for a traditional stove, and potentially use white paint or a piece of paper to truly measure “zero usage of the biomass stove.” We will also feel the firewood/charcoal stoves in order to access if they have been recently lit. We will also track the percentage of biomass being used for other activities such as water boiling or ironing, which is not considered in the scope of this outcome variable. The primary variable will be the binary indicator of exclusive LPG use; however, a secondary analysis will be conducted to see if the savings bank had an effect on the other levels of LGP use.

Firewood and Charcoal Use
This dependent variable will be a continuous variable regarding the amount of firewood and charcoal the household uses. We will weigh typical bundles of firewood and the three types of charcoal quantities (black sacks, buckets, and large sacks). Additionally, we will triangulate this with the cost data the household provides to determine the average kilogram weight for firewood/charcoal of that monetary amount.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The SW-RCT is a cross over design in which households (N=~500) will cross over from control to intervention with a year-long follow-up period to allow consistent use to be established. Each household will receive two LPG stove and burner sets, along with education from their CTW and weekly follow up visits. The CTW’s role is to provide education and maintenance support, but also to nudge the household weekly to save money towards their refill. The treatment in this study is the option to deposit money in a micro-savings program for their LPG refill. We piloted whether the microsavings program should be through mobile money or a physical lockbox. Households in Shirati preferred the lockbox to a mobile money program. Households will have the option to have the CTW keep the key or care for it themselves. The micro-savings program mirrors how households are accustomed to purchasing firewood or charcoal—small sums on a day-to-day basis. There is no other micro-savings program for cooking fuel in Shirati available to the households while in the control (before they cross over) and access to formal financial services is limited, expensive, and inconvenient.
Experimental Design Details
Randomization Method
We will randomize both the order in which the households receive the intervention and the households who will participate in the program. We will utilize a spatial sampling method to randomly select households from the villages. Unfortunately, the 2012 Tanzanian Census was not updated from the 2002 Census, and therefore, is not reliable for the sampling strategy. We will utilize Google Maps and Python to randomly place ~500 points within Shirati and the closest household will be invited to participate.
Randomization Unit
Our unit of randomization is the household.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
~500 Households (510 enrolled to account for attrition, refusal, etc.)
Sample size: planned number of observations
~500 Households (510 enrolled to account for attrition, refusal, etc.)
Sample size (or number of clusters) by treatment arms
~500 control households that will cross over into treatment (510 enrolled to account for attrition, refusal, etc.)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Our unit for sample size is the household. We powered for the third hypothesis (outcome: households using LPG exclusively) as exclusive use of LPG is very rare and thus requires the largest sample size to achieve an α of .05 and a β of .2 for 80% power. For the power calculation, we summarize this single outcome pre/post treatment as the proportion of observations that are exclusive users. The treatment impact is the difference between these two measures. From the pilot data, we calculated the variability of this outcome across households and the co-variability within households to be 0.48 and 0.20 respectively . In the pilot study, we found that 40% exclusively used LPG. To be conservative, we want to power for at least a 10% difference (50% exclusive use in treatment). Utilizing these assumptions and prior pilot data, we can calculate the necessary sample size for 80% power through a paired t-test sample size calculator in STATA. These results reveal that 292 households are necessary to detect a 10% effect. This sample size reduces to 75 households to detect a 20% effect. To overpower (and to account for possible dropout) we will enroll ~500 households.

Institutional Review Boards (IRBs)

IRB Name
University of California, Berkeley Institutional Review Board
IRB Approval Date
IRB Approval Number
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials