Technology recommendations and social concerns: Evidence from savings technologies in Pakistan

In the experiment, we will measure rural Pakistanis willingness-to-recommend a familiar savings technology, a lockbox, and an unfamiliar technology, a digital savings account. Our interventions attempt to de-risk recommendations. The first intervention makes the recommendation anonymous. In the second intervention, the recommender is made aware that their partner will receive consumer financial protection training prior to using the technology.

This experiment also includes a village-level intervention. In two thirds of the villages, we will seed socially-central individuals with information about digital savings accounts. In half of these "seeding" villages, we will also include consumer financial protection information and offer an anonymous communication channel for discussion of digital savings accounts.

Seeding village-level

Both treatment groups (100 villages) will receive the Seeding intervention. In the Seeding intervention, field officers will visit the seeds at their homes, inform the seeds they have been selected to be a `Gallup Pakistan Responsible Financial Inclusion Ambassador’ provide the `core digital savings information’ pamphlet, a digital banking-themed calendar and offer to open a digital savings WhatsApp/SMS message group with the seed and their contacts. All seeds will receive weekly `digital savings tips’ via SMS for eight weeks. All sampled respondents will receive an SMS informing them of the identity of their village's `Ambassador'.

Anonymous-CFP village-level

In the 50 Anonymous-CFP villages, we give villagers the option to post in the digital banking WhatsApp/SMS group anonymously. Villagers will be able to send messages to the research team phone number and then, the message will be posted by the research team in the group after the removal of any personally-identifiable information. Alternatively, users will be allowed to create pseudonyms when using the group if this is easier to implement. In addition, there will be additional consumer financial protection information in the pamphlet, calendar and SMS tips.



Consumer Financial Protection Information individual-level

The CFP Info respondents will be told how to avoid risks associated with digital savings accounts and will be told that we will relay this same information to their peer. This information will be relayed prior the willingness-to-adopt or willingness-to-recommend measurements.

Anonymous individual-level

In the Anonymous treatment, we will not tell the peer who recommended the digital savings account or lockbox account. In other arms, respondents will know that we will tell the peer that they were the ones who recommended the product. This information will be relayed prior the willingness-to-adopt or willingness-to-recommend measurements.

2023-02-15

2023-04-15

Individual-level: willingness-to-recommend digital savings account to co-villager; willingness-to-recommend lockbox to co-villager
Village-level: experiences with digital savings accounts; knowledge of digital savings accounts;

Willingness-to-recommend is measured through a Becker-DeGroot-Marschak mechanism.
Experiences with digital savings accounts and knowledge of digital savings accounts will be indices constructed from endline survey responses following the Anderson (2008) method.

Individual-level: use of digital savings account, Uptake of digital savings account/lockbox by co-villager
Village-level: communication about digital savings accounts, amount saved, opinions about digital savings, advanced use of digital banking, resilience to shocks.

Use of digital savings account will be measured via the endline survey.

Uptake of digital savings account/lockbox by co-villager is a dummy variable equaling 1 if the co-villager accepts our offer and 0 otherwise.
Communication about digital savings accounts is an index of endline responses relating to communication about digital savings accounts.
Amount saved is the total amount saved measured in PKR at endline.

Opinions about digital savings is an index of how positive the respondent views digital savings at endline.

Advanced use of digital banking is an index of how much the respondent uses digital banking products such as insurance, loans etc. at endline.

Resilience to shocks is measured by questions on how the respondent would deal with a negative shock at endline.

I propose to run this field experiment in 150 villages in Pakistan. Villages will be randomized into one of five experimental groups stratifying by subdistrict, digital banking knowledge, social proximity of respondents to seeds, and population size: Control, Seeding, Seeding + Anonymous-Consumer Financial Protection (CFP). In the baseline survey, we will randomize respondents into one of four groups: Control, Consumer Financial Protection (CFP) Info, Anonymous recommendation, and CFP Info + Anon. recommendation. After eight weeks, the field team will conduct phone surveys with the baseline respondents.

I propose to run this field experiment in 150 villages in Pakistan in partnership with Gallup Pakistan. We will randomly sample subdistricts (tehsils) from the chosen eligible area. We will stratify by mouza/deh (an administrative unit in Pakistan encompassing a few villages) to ensure villages are sufficiently far apart to rule out one treatment affecting the another village through spillovers. In each village, the Gallup Pakistan field officers will conduct circular random sampling of the households. In the baseline survey these households will answer the “gossip” questionnaire of Banerjee et al (2019). Through the “gossip” questionnaire, we will identify a household in each village to be our seeds.

Village-level randomization

Villages will be randomized into one of five experimental groups stratifying by subdistrict, digital banking knowledge, social proximity of respondents to seeds, and population size: Control, Seeding, Seeding + Anonymous-Consumer Financial Protection (CFP).

Individual-level randomization

In the baseline survey, we will randomize respondents into one of four groups: Control, Consumer Financial Protection (CFP) Info, Anonymous recommendation, and CFP Info + Anon. recommendation.


Individual-level randomization stratification

To ensure heterogeneity by risk-sharing, we will stratify the peer of the respondent by whether the respondent shares risk (someone listed in ``Who do you give/receive money from in times of need?") with them. We will also record whether the peer ends up taking the recommendation of the respondent or not.

Endline survey and measurement of outcomes

After eight weeks, the field team will conduct phone surveys with the baseline respondents.

randomization done in office by a computer

Household-level for information treatments; village-level for seeding treatments

Yes

150 villages; 2100 households

2100 adult household members

50 villages per village-level treatment arm; 525 adults per individual-level treatment arm

For the individual-level, we will have a MDE of .17 SDs between subgroups with .80 power at a .05 significance level for our main outcomes. At the village-level, we will have a MDE of .30-.35 SDs between subgroups with .80 power at a .05 significance level assuming an ICC between .03-.20 for our main outcome.