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.
Income Timing, Temptation, and Expenditures: Field Experimental Evidence from Malawi
Initial registration date
July 15, 2014
November 29, 2016 1:29 PM EST
University of Minnesota
Other Primary Investigator(s)
University of Michigan
Additional Trial Information
Developing economies are characterized not only by low incomes but also by low savings and apparent failures to take advantage of profitable opportunities for investment (e.g. de Mel at al. 2008). One common explanation is that the poor face temptations to misspend their income which they are unable to overcome given limited financial markets, and that aggregating steady streams of small amounts of income into larger lump sums would help mitigate these problems. This assumption has helped drive the research on commitment savings accounts (Ashraf and Karlan 2006, Brune et al. 2013). However, converting smooth income streams into larger, deferred sums may also lead to increased temptation and potentially poor choices - the proverbial effect of "money burning a hole in one's pocket". Banerjee et al. (2010) find evidence of this behavior among some users of microfinance in Hyderabad, and increasing temptation is consistent with anecdotal reports of behavior around payday in Malawi as well as research from the developed world (e.g. Stephens 2003). It is not currently understood what factors lead certain lump sums to be beneficial to achieving savings goals, while others lead people to give in to temptation, but the specific timing of the lump sum may play a role. This project uses an RCT to explore how the timing of wage payments changes the relative benefits of lump sums vs. small installments.
We partner with a local NGO to expand a public works program by 350 workers who are employed for 16 days over the course of three months (Oct through Dec 2013), subdivided into two rounds. Workers are cross-randomized into two sets of experimental conditions, separately for each round of the study. The first study arm varies whether workers receive their pay in a single, lump-sum payment or in weekly installments. The second study arm varies how tempting the environment is in which workers receive their payments: All workers receive their pay at the same location, which is the site of the major local market, but some receive their pay on the weekly "market day", while others receive it the day before. To ensure that all workers face equal transaction costs, everyone will come to the market for brief surveys on both days, irrespective of whether they are being paid. The main outcomes of interest are survey measures of saving level, expenditure composition, loans, and transfers, as well as uptake of a high-interest, zero-risk "bond" offered to all the workers in the study that provides an objective measure of savings behavior. Registration Citation
Brune, Lasse and Jason Kerwin. 2016. "Income Timing, Temptation, and Expenditures: Field Experimental Evidence from Malawi." AEA RCT Registry. November 29.
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
Our key outcome variables are:
1. Savings, expenditure, loans and transfers as measured via household surveys
2. Spending behavior captured through high-frequency surveys conducted while respondents are picking up their wage payments
3. Uptake of a high-interest "bond" offered to all respondents, that must be bought in large installments, as an objective measure of savings behavior
Primary Outcomes (explanation)
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
The experiment was organized into two rounds that occurred over a period of three months from November 2013 to January 2014, with subjects randomized into treatment conditions separately by round. During each round, subjects worked for two weeks and then received their pay either a) in weekly installments beginning at the end of the second week of work; or b) in a single lump sum, about three weeks after the last day of work.
In addition to variation in payment frequency, workers received their pay either c) on Fridays or d) on Saturdays. The two variations on the timing of pay -- weekly vs. monthly and Friday vs. Saturday -- were cross-randomized, creating four study arms in each round. The payments were made at the site of a major local market that occurs on Saturdays, with the intention of inducing variation in people's temptation to overspend. During the week after the last payday in each round, all workers were visited for a detailed survey about their expenditure and income.
The variation in the day of the week of the payment was designed to shed light on the mechanisms behind the savings constraints people face. If money is received in a tempting environment, like the local market day, then arguably costs to resisting that temptation increase and workers would decide to spend and consume more right at the market when receiving their pay.
We picked Saturdays at the local trading center -- so that payroll for this group happened during the major market in the local area -- as a tempting context for the receipt of income. This choice was based on extensive qualitative and descriptive work with people in the local area. Anecdotally, people in Mulanje District often describe market days as tempting situations, in which excitement can cause them to purchase things they would rather not.
Experimental Design Details
Randomization done in office by a computer. Randomization was stratified on workers' gender and village in round 1. The randomization for round 2 was then stratified on the round 1 assignment and village.
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
350 worker-rounds monthly lump-sum, 350 worker-rounds weekly installment
350 worker-rounds paid on Fridays, 350 worker-rounds paid on Mondays
Cross-randomization yields 4 cells of 175 worker-rounds apiece
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Our MDES for each study arm, for each round, is 0.30 SDs of the outcome variable. For detecting differences between each cell of the 2x2 cross-randomization, our MDES is 0.43 SDs. Pooling the data across rounds will increase our power (and decrease our MDES); exact calculations will depend on the ICC of the outcome variable in question.
INSTITUTIONAL REVIEW BOARDS (IRBs)
University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board
IRB Approval Date
IRB Approval Number
Innovations for Poverty Action IRB
IRB Approval Date
IRB Approval Number
Post Trial Information
Is the intervention completed?
Intervention Completion Date
January 24, 2014, 12:00 AM +00:00
Is data collection complete?
Data Collection Completion Date
January 31, 2014, 12:00 AM +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
715 worker-rounds (365 workers, workers re-randomized to different treatments across rounds)
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
Lump sum treatment:
359 workers treatment, 356 workers control
Saturday payday treatment:
358 workers treatment, 357 workers control