Experimental Design
The study is conducted in multiple rounds.
We will be working with potential and actual SME and microentrepreneur clients of a Colombian bank. To date, we have conducted three rounds of intervention, in efforts to increase the study sample.
First round of intervention (May 17-August 13, 2021):
Firms are divided in the following four eligible groups, with 1,004 firms total:
i. Type 1: Clients from a previous credit campaign that received loans but may need loans again now.
ii. Type 2: Clients from a previous credit campaign who were not interested in loans then but may need loans now.
iii. Type 3: Firms which are at the credit score margin and thus were not considered for contact in any previous credit campaign.
iv. Type 4: Firms that are above the credit score margin, but have not been contacted in any previous credit campaign.
These firm types are used as a stratification variable in the randomization procedure. To avoid sparse strata, we merge together Group 2 and Group 4, thus leaving us with 3 strata for this variable.
We also stratify by the number of months where clients made at least one deposit in 2020 (12 strata). Clients that did not make at least one deposit in 2020 were excluded from the experiment.
There are 59 SMEs that made their first deposits in 2021. For these clients, we only stratify by the firm type, as we have no available transaction data to randomize in a similar fashion to firms who were clients since 2020 or earlier.
Second round of intervention (September 13-November 30, 2021):
To increase the sample of the first round of intervention, we have added an additional 2,424 SME firms to the experiment. These firms made at least one deposit between January and August 2020 (no updated data was available before the second intervention round launched). These firms were stratified into the client type as defined above:
i. Type 1: Clients from a previous credit campaign that received loans but may need loans again now.
ii. Type 2: Clients from a previous credit campaign who were not interested in loans then but may need loans now.
iii. Type 3: Firms which are at the credit score margin and thus were not considered for contact in any previous credit campaign.
iv. Type 4: Firms that are above the credit score margin, but have not been contacted in any previous credit campaign.
As in the case of the first intervention round, we merge together Group 2 and Group 4, thus leaving us with 3 strata for this variable. Since the transaction data was incomplete for this new additional SME sample, we only stratified on firm type.
For the firms that are randomized into the treatment group, these will be offered the credit product by the bank via phone call, at which point they can express interest in receiving the loan. After a bank evaluation, these loans are approved and disbursed to the client. We will then receive administrative bank data for these clients to evaluate the outcomes of interest in the short and medium term.
Third intervention round with microentrepreneurs
In parallel with the SMEs added to the second intervention round for the experiment, we planned to launch an additional intervention round with 19,615 self-employed microentrepreneur clients of Davivienda. However, a rule change at the Colombian National Guarantee Fund (Fondo Nacional de Garantías), the government institution which provides the loan backing that is part of our experiment's intervention, now required a handwritten certification of income from loan applicants. This made compliance by Davivienda's microentrepreneur team impossible, because they offer loans 100% digitally through the bank’s mobile app, have little to no physical contact with the clients, and thus cannot obtain this handwritten certification. As a result, none of the microentrepreneur clients that we had randomized into treatment have been pre-approved for loan guarantees from the government, and because of that, this intervention round was discarded.