Do transparent moneyboxes increase savings? A note on visual feedback and savings behavior.

Last registered on June 21, 2015

View Trial History

Pre-Trial

Trial Information

General Information

Title

Do transparent moneyboxes increase savings? A note on visual feedback and savings behavior.

RCT ID

AEARCTR-0000618

Initial registration date

February 04, 2015

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

February 04, 2015, 9:14 AM EST

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

Last updated

June 21, 2015, 4:44 PM EDT

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

Locations

Country

Ethiopia

Region

Primary Investigator

Name

Niels Kemper

Affiliation

University of Mannheim

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Markus Froelich

PI Affiliation

University of Mannheim

PI Name

Albrecht Bohne

PI Affiliation

University of Mannheim

PI Name

Alexandra Avdeenko

PI Affiliation

University of Mannheim

Additional Trial Information

Status

On going

Start date

2013-11-15

End date

2015-10-31

Keywords

Finance & Microfinance

Additional Keywords

Limited attention, visual feedback, savings

JEL code(s)

Secondary IDs

Abstract

Limited attention may play a role in explaining the inability to reach savings goals. Presuming that attention is a scarce resource and important for the pursuit of saving plans, limits to attention in inter-temporal consumption and saving decisions may cause savers not to reach their savings goals. In turn, constant feedback on the savings performance may help to reach savings goals in the presence of limited attention. In this project we draw on a simple experiment to test whether the visual representation of savings balances stimulates savings behavior and helps individuals reach their savings goals. Doing so, we randomly offer transparent and non-transparent moneyboxes as a complementary savings device to current and former clients of a microfinance organization in Ethiopia. We hypothesize that the visual representation of saving balances via transparent money boxes increases savings and helps microfinance clients reach their savings goals. While this is our working hypothesis, we may also observe the opposite, i.e. that the visual representation of saving balances via transparent money boxes decreases savings (e.g. because of a higher temptation to spend money that can be seen, or because it is less likely to hide savings from others).

External Link(s)

Registration Citation

Citation

Avdeenko, Alexandra et al. 2015. "Do transparent moneyboxes increase savings? A note on visual feedback and savings behavior. ." AEA RCT Registry. June 21. https://doi.org/10.1257/rct.618-2.0

Former Citation

Avdeenko, Alexandra et al. 2015. "Do transparent moneyboxes increase savings? A note on visual feedback and savings behavior. ." AEA RCT Registry. June 21. https://www.socialscienceregistry.org/trials/618/history/4542

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

In partnership with a local microfinance organization and a local university, we distribute a simple savings technology - a moneybox - to smallholder farmers in northern Ethiopia. All money boxes are cylindrical with a slot on top, to chip in money, and an opening at the bottom, to get money out of the box. Each household receives two keys to open the box, one given to the head and one to the spouse. Apart from that, moneyboxes differ along one important dimension:

Transparent: Transparent money boxes allow seeing the bank notes and bills held inside of the box at all times.

Non-transparent: Non-transparent money boxes do not allow seeing the inside of the box.

All households who receive a money box formulate a savings goal and amount for which they would like to save. In addition to transparent and non-transparent boxes, we cross-randomize the recommendation to save in a daily or weekly frequency towards the savings goal. Savings goal, savings period, and saving installment are written on the label of the box.

The treatments for this study were cross-randomized with the treatments of another study in the same sample of current and former clients of the local microfinance institution:

https://www.socialscienceregistry.org/trials/613

Intervention Start Date

2014-12-15

Intervention End Date

2015-06-15

Primary Outcomes

Primary Outcomes (end points)

Monetary savings, Savings goal attainment

Primary Outcomes (explanation)

Monetary savings: total cash savings, bank account savings as well as savings held with Iddir, a funeral society, and Equb, a savings club. For treated households this indicator also includes the savings held in the moneybox.

Savings goal attainment: difference between goal amount and actual savings.

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

Our study covers households from 94 village clusters in remote areas in northern Ethiopia. This basically creates a census of the zone in which the RCT is conducted. Zone is third highest level of administration of the country, subordinate to region and national level. In each cluster 10 households were randomly sampled from client lists held at the branches of the local microfinance institution. The client list comprises current and former clients from the local microfinance institution.

Treatments are randomly assigned based on balanced baseline covariates with respect to the Mahalanobis distances. The 940 households were randomly assigned to either control or to one of twenty different treatment arms, such that 300 observations were assigned to control and 32 observations to each treatment arm. Thereafter the Mahalanobis distance with respect to the baseline covariates given below was calculated. Mahalanobis distance was calculated pairwise between any two treatment arms and also for each treatment arm relative to control. The random assignment was only accepted if the maximum distance to control was smaller than 0.605 and the maximum distance between any of the treatment arms was smaller than 1.322. These numbers corresponded to the 0.01 percentile of the empirical distribution of these distances.

The baseline covariates used to calculate Mahalanobis balance were current savings in cash (amount in Birr), current savings with the local microfinance institution (amount in Birr), whether household is a member of an Iddir, a funeral insurance, or an Equb, a savings group (both binary indicators), the demographic composition of the household (measured by the number of household members between 0-5, 6-11, 11-17, 18-65 and more than 65 years of age), whether the household head is female or married (both binary indicators), the years of education of the head, the total land area under cultivation (in hectare), the total revenues from crop sales (in Birr) and the per-capita consumption expenditures (in Birr). The treatments in the study were assigned according to the first random assignment vector that passed the balance thresholds.

Experimental Design Details

Randomization Method

Randomization was done with a computer using Gauss and Stata.

Randomization Unit

Household

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

940 households

Sample size: planned number of observations

940 households

Sample size (or number of clusters) by treatment arms

A
Transparent moneybox: 320 households
Non-transparent moneybox: 320 households

B
Recommendation to save daily: 320 households
Recommendation to save weekly: 320 households

C
Transparent moneybox *daily: 160 households

CONTROL
Pure control group: 300 households

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

MDE (A versus CONTROL): Minimum detectable effect size in ETB: 871.69 Birr, MDE/Std.dev. in percent: 22,55 percent, MDE/mean in percent: 29,60 percent; MDE (B versus CONTROL): Minimum detectable effect size in ETB: 871.69 Birr, MDE/Std.dev. in percent: 22,55 percent, MDE/mean in percent: 29,60 percent; MDE (C versus CONTROL): Minimum detectable effect size in ETB: 1062.42 Birr, MDE/Std.dev. in percent: 27,48 percent, MDE/mean in percent: 36,08 percent.

Supporting Documents and Materials

IRB