Does phone-based heuristic financial training alter entrepreneurial behavior? Experimental Evidence from Ethiopia

Last registered on May 19, 2022


Trial Information

General Information

Does phone-based heuristic financial training alter entrepreneurial behavior? Experimental Evidence from Ethiopia
Initial registration date
September 23, 2021

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
September 28, 2021, 3:46 PM EDT

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

Last updated
May 19, 2022, 3:11 AM EDT

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



Primary Investigator

The World Bank

Other Primary Investigator(s)

PI Affiliation
The World Bank

Additional Trial Information

On going
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Classroom-based financial literacy training programs have been popular interventions implemented to promote financial inclusion, bridge the observed gap in financial management and improve business practices among microentrepreneurs. Their effectiveness, however, have been largely limited for a multitude of reasons. Is there a more innovative and less expensive means of delivering concise and actionable financial lessons to microentrepreneurs? Heuristics-based financial lessons delivered through mobile phone-based Interactive Voice Response (IVR) platform might hold a promise to improve financial knowledge and practices. We conduct an RCT to test the effectiveness of a heuristic-based financial literacy training program in the context of female owned businesses in five large cities in Ethiopia. We specifically focus on: 1) training engagement and likelihood of phone pick up; 2) changes in financial behavior, 3) changes in business practices, and 4) downstream effects on firm performance.
External Link(s)

Registration Citation

Abebe, Girum and Adiam Hagos Hailemicheal. 2022. "Does phone-based heuristic financial training alter entrepreneurial behavior? Experimental Evidence from Ethiopia." AEA RCT Registry. May 19.
Sponsors & Partners


Experimental Details


The study has four treatment arms: a heuristic-based financial training (T1), a heuristic-based financial training plus post-training summary lessons for eight weeks (T2), a placebo treatment arm (P) and a pure control (C).
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
5.2.1. Primary outcomes: business knowledge
We estimate the impact of the treatment on business knowledge using equation 2. A measure of aggregate business knowledge would be constructed using a set of questions on business knowledge as indicated in Family 5.2.1.

Outcome Family 5.2.1: business knowledge
Variable Definition Source (Question number)
business_knowledge Test score for business knowledge z-score of averages of all variables listed in the following 10 rows
test_1 Answers correctly the question on how profit is determined Dummy: bk_1==1
test_2 Answers correctly the question on why market research is important Dummy: bk_2==3
test_3 Answers correctly the question on the best method of checking business progress Dummy: bk_3==1
test_4 Answers correctly the question on why advertising is important Dummy: bk_4==2
test_5 Answers correctly the question on business discounts Dummy: bk_5==1
test_6 Answers correctly the question on why collateral for a loan is required Dummy: bk_6==2
test_7 Answers correctly the question on what happens to loans when business is not performing well Dummy: bk_7==2
test_8 Answers correctly the question on customer relationship Dummy: bk_8==1
test_9 Answers correctly the question on interest rate on a loan Dummy: bk_9==3
test_10 Answers correctly the question on combining personal and household resources Dummy: bk_10==2

5.2.2. Primary outcome: Financial behavior
We estimate the impact of the treatment on financial behavior using outcomes indicated in Family 5.2.2. We mainly plan to estimate the aggregate outcomes indicated in the first row. We will also look at individual outcomes when further insight is desired. There are three family of outcomes that we intend to explore under the rubric of financial behavior: financial practice, credit management and inventory management.
Outcome Family 5.2.2: Financial Behavior
Variable Definition Source (Question number)
Family 1: Financial Practice
financial_score Overall score for financial practice z-score of averages of all variables listed in the following 7 rows
written_plan have a written business plan Dummy: mr_4=1
separate_expense Separate business and household expenses Dummy: mr_9=1
separate_account Separate bank account for the business Dummy: mr_10=1
Separate_location Keep business money in a separate location than the money for the household Dummy: sa_2=1
caculate_profit Calculates profit either the entrepreneur herself or uses professional accounts Dummy: mr_11=1 or mr_11=2
correct_profit_calcu Enumerator’s assessment of how correctly the entrepreneur calculates her profit Dummy: mr_12=2
know_items_selling knows which items are making more money and which items are incurring losses Dummy: mr_14=1
Family 2: Credit Management
credit_managment Overall credit management score z-score of averages of all variables listed in the following 7 rows
credit_limit have a credit limit Dummy: cc_3 is different from -22 if cc_1==1
payment_write Keep payment date in writing Dummy: cc_6_1b =1 if cc_1=1
amount_write Keep payment amount in writing Dummy: cc_6_2b =1 if cc_1=1
penality_write Keep penalty in writing Dummy: cc_6_3b =1 if cc_1=1
track_credit Track the credit issued by keeping written record or keeping formal agreement Dummy: cc_7 =2 or cc_7=3 if cc_1=1
incentive_repayment Provide incentive to those who pay early Dummy: cc_9 =1 if cc_1=1
credit_procedure Established procedure to follow-up with customers who miss payments Dummy: cc_10 =1 if cc_1=1
Family 3: Inventory Management
inventory_managment Overall inventory management score z-score of averages of all variables listed in the following 5 rows
categorize_items Categorize items in the inventory systematically based on type or priority Dummy: im_3 =1
inventory_record Decides to buy inventory based on the records maintained Dummy: im_5=2
always_stocked Did not run items out of stock Dummy: im_6=1
safety_stock Carry safety stock to guard against running out of important items Dummy: im_7=1
no_bad_stock Does not carry inventory of items if they were not selling for a long time Dummy: im_8=2

5.2.3. Primary outcome: Business practice
We intend to construct a measure of business practice using the outcome families indicated in Family 5.2.3. We estimate equation 2 to explore whether the intervention impacts the business practice score of the treatment group entrepreneurs.
Outcome Family 5.2.3: Business practice
Variable Definition Source (Question number)
business_practise Overall business practice score z-score of averages of all variables listed in the following 15 rows
visit_competitor visited one of your competitors to learn about the products he/she offers Dummy: s4f_1a=1
consult_client Checked with clients on products and services that they would like to buy Dummy: s4f_1b=1
compared_suppliers Compared the prices or quality of supplier’s product/service with other suppliers Dummy: s4f_1c=1
new_market Looked for new markets Dummy: s4f_1d=1
analyzed_sales analyzed the sales of your most important product/services Dummy: s4f_1e=1
ways_mark_adver Looked for new ways of improving marketing and advertising Dummy: s4f_1f=1
add_finance Looked for additional financial resources for business Dummy: s4f_1g=1
network discussed with other entrepreneurs about production techniques, suppliers or new product Dummy: s4f_1h=1
record Record every transaction the business makes Dummy: mr_3=1
proper_recording Record transactions using either a dedicated notebook or a computer Dummy: mr_5=2 or mr_5=3
preserve_purchase Preserve documents when making purchase Dummy: mr_7=1
preserve_sales Preserve documents when making sales Dummy: mr_8=1
know_cash_record Can use records to see how much cash the business has on hand at any point in time Dummy: mr_13=1
archive keep archives of important documents Dummy: mr_17=1
historical_data analyze historical data (information compiled in the past) for business decisions Dummy: mr_18=1

We will generate indices of these family of outcome along the lines of Drexler et al. (2014) to ensure comparability of our results with the existing literature. In addition, the use of these indices reduces the probability of rejecting the null hypothesis of no effect of at least one outcome which would be the case if indicators are tested individually (Karlan and Valdivia, 2011). We will standardize these measures by converting them to z-scores.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
5.2.4. Secondary outcome: Savings
We also plan to examine the impact of the treatment on the savings behavior of entrepreneurs.
Outcome Family 5.2.4: savings
Variable Definition Source (Question number)
Twelve_month_saving Savings amount in the last 12 months sa_4_1
six_month_saving Savings amount in the last six months sa_4_2
one_month_saving Savings amount in the last month sa_4_3
intent_saving Given monthly business earnings, how much the entrepreneur would you like to save per month sa_6
twelve_month_invest Investment amount in the last 12 months sa_4_1
six_month_invest Investment amount in the last six months sa_4_2
one_month_invest Investment amount in the last month sa_4_3

5.2.5. Secondary outcome: Borrowing
We also plan to examine the impact of the treatment on the credit worthiness of entrepreneurs.
Outcome Family 5.2.5: Borrowing
Variable Definition Source (Question number)
mfi_borrow Borrowed from MFI in in the last 12 months Dummy: s4e_9=1
bank_borrow Borrowed from a bank in the last 12 months Dummy: s4e_11=1
total_borrow Total amount of money borrowed from all sources s4e_23 (s4e_23=0 if did not borrow from any source)
max_borrowing maximum amount of money the entrepreneur is able to borrow in two weeks sa_6

5.2.6. Secondary outcome: firm performance
We also plan to examine the impact of the treatment on the performance of the entrepreneurs—namely, these family of outcomes include number of workers, sales and profitability.
Outcome Family 5.2.6. firm performance
Variable Definition Source (Question number)
empt Number of workers s4b_1
Log_sales Log of sales in the last completed month Log of s4d_11
Log_profit Log of profit in the last completed month Log of s4d_9

Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The main objective of the heuristic-based financial training is to alter people’s financial knowledge and behavior by offering basic (rules of thumb) training programs that are easy to understand and recall. The training will be designed by an international NGO that has the experience of implementing similar training programs in India, the Philippines, and the Dominican Republic. The phone-based audio messages will be delivered through an Interactive Voice Recording (IVR) platform.
We propose four treatment arms for this study: a heuristic-based financial training (T1), a heuristic-based financial training plus post-training summary lessons for eight weeks (T2), a placebo treatment arm (P) and a pure control (C).

Table A1. Experimental design
Receives Financial Heuristic training Receives post-training SMS reminder Receives health-related training
T1 Yes No No
T2 Yes Yes No
Placebo No No Yes
Control No No No

Experimental Design Details
Not available
Randomization Method
We assign 3000 entrepreneurs to receive the treatment and 3000 to be either the placebo (1000) or the pure control (2000). From those who receive the treatment half would be offered the post-treatment SMS follow-ups, which provides them with the summary of the training modules for eight weeks. We follow a rerandomization rule where we continue to randomize up to 500 times or until the balance p-value exceeds 0.2. Randomization is done by a computer using STATA.
Randomization Unit
Randomization is carried out at the individual level with stratification at city levels; i.e., individual entrepreneurs grouped into five strata made up of cities and are randomly assigned to each of the four arms randomly.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
We have samples drawn from five cities.
Sample size: planned number of observations
We conducted a baseline data from 6000 observations. And we aim to track all 6000 of them in subsequent follow-ups
Sample size (or number of clusters) by treatment arms
Of the 6000 surveyed in the baseline, 3000 entrepreneurs receives the financial literacy training and 3000 would be control. Of the former, 1500 would receive eight weeks of summary lessons. And among the control, 1000 would be considered as a placebo group and would receive health related messages on the same frequency as the treatment group.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We conduct power analysis for one of our key outcome variables where an accurately measured survey is available, namely test results on business knowledge. A standardized business knowledge measure with mean zero and Sd. 1 is obtained from a previous surveys conducted on similar sample pool. The test result data shows that the mean score was 13.295 and Sd. was 1.791, the highest score was 18 and the lowest was 0. A total of 19 business knowledge questions were asked. We transform this test results into z-score with mean zero and Sd.1. We assume power of 0.8 and a significance level of 0.05. With a sample size of 6000 entrepreneurs, we find that at the 95 percent confidence levels, we have sufficient statistical power to rule out a change in more than 0.075 standard deviation in test results, which is also our minimum detectable effect size. Further, imposing an 60 percent training take-up increases the minimum detectable effect size to 0.12. Allowing for a 20 percent attrition increases the MDE to 0.0808. Combining a 20% attrition with a 60 percent take up increases the MDE to 0.135, which is a highly punitive assumption. Overall, we estimate that we will be able to fairly detect a 7.5 percent increase in business knowledge.

Institutional Review Boards (IRBs)

IRB Name
HML IRB Research and Ethics
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information