|
Field
Intervention (Public)
|
Before
We randomize the expansion of new retail mobile money vendors across low-income localities. Our design creates 3 different exogenous variations at the locality / market-level — all guided by theory and practice — (i) some local markets receive the entry of new retail vendors vs not, (ii) for the entry localities, we vary the intensity of entrants across markets (either +33% vs +100% increase in vendorship relative baseline vendorship), and (iii) from an identified pool of “eligible“ local retail shops, we randomly onboard some as new mobile money vendors vs not.
|
After
We randomize the expansion of new retail mobile money vendors across low-income localities. Our design creates 3 different exogenous variations at the locality / market-level — all guided by theory and practice — (i) some local markets receive the entry of new retail vendors vs not, (ii) for the entry localities, we vary the intensity of entrants across markets, and (iii) from an identified pool of “eligible“ local retail shops, we randomly onboard some as new mobile money vendors vs not.
|
|
Field
Primary Outcomes (End Points)
|
Before
We successfully piloted our experiment (key protocols, instruments, proposed design, recruitment and enrollment processes) between January - June 2021.
Measurements:
*For (incumbent) market vendors, we plan to measure impacts on:
(i) Misconduct - Quality: prices charged for retail services; overcharging consumers or illegal markups (incidence and severity).
(ii) Transparency: whether price list posted, visible and clear; including whether vendor attempts to inform consumers about transaction tariffs before conducting transaction.
(iii) Reliability - Quality: whether vendor is present/absent at the retail outlet; whether transaction successful; likelihood of vendors’ declining transactions due to liquidity shortfalls or poor liquidity management; both transaction and service times.
(iv) Innovation: bundling M-Money with other non-M-Money services (or expansions to other lines of business); including total exits; changing vendor locations; hours of operations; vendor's accessibility to previously unbanked / unserved households.
(v) Service - Quality: consumers' views and experiences about safety, privacy, invasion, harassment, discrimination, and respect after visiting vendor points.
We will adapt Annan (2020, 2021) audit study protocols (and surveys) to measure vendor outcomes objectively (and subjectively respectively). We will combine the outcomes to derive an index of overall service quality.
*For consumers, we plan to examine the impacts on usage and trust outcomes.
(i) We will field questions that will provide subjective measures of trust in M-Money providers, market vendors, carrying out vendor-involved transactions on M-Money, consumers’ family and friends, commercial and rural banks, and the regulator of financial services in Ghana (Bank of Ghana) if they have heard enough about them to say.
(ii) We will supplement this with objective measures of trust measured from trust games between consumers and their local vendors. In its simple form: this entails a real monetary payoff game between the consumers (i.e., trustors) and one randomly selected (anonymous) local vendor (i.e., trustee). We impose vendor anonymity to mitigate against issues of social and individual preferences. Consumers will be endowed with 50GHS and will decide how much to send to another person (i.e., M-Money vendor). We will triple it, so that the vendor receives three times the amount of money the consumer sent. The vendor will then decide how much he/she wants to send back to the consumer. The total payoffs depend on the choices made between the consumers and vendor. Our objective measure of consumer trust will reflect the amount they sent and expected the vendor to send back to them.
*For (all market participants: vendors, consumers, nearby stores), we plan to measure their beliefs and expectations about potential competition effects if competition is introduced on: (i) misconduct / overcharging, (ii) reliability / illiquidity (transaction declines due to liquidity shortfalls), (iii) transparency / tariff posting behavior of vendors, (iv) location of vendor's retail store, (v) innovation / bundling of M-Money with other non-M-Money services at agent points, (vi) consumer trust, and (vii) consumer experiences and usage.
*We will combine baseline and endline (i) market census, (ii) surveys, and (iv) audit studies to track vendor, customer, and business outcomes (including: revenues, profits, number of customers, business expenses, employment, wagebill, household expenses), supplemented with administrative data (on sales revenue, vendor entry / exit, consumers' usage) from the commercial providers to examine broader impacts, including persistence and entry of new vendors across the localities.
|
After
We successfully piloted our experiment (key protocols, instruments, proposed design, recruitment and enrollment processes) between January - June 2021.
Measurements:
*For (incumbent) market vendors, we plan to measure impacts on:
(i) Misconduct - Quality: prices charged for retail services; overcharging consumers or illegal markups (incidence and severity).
(ii) Transparency: whether price list posted, visible and clear; including whether vendor attempts to inform consumers about transaction tariffs before conducting transaction (verbal disclosure of price).
(iii) Reliability - Quality: whether vendor is present/absent at the retail outlet; whether transaction successful; likelihood of vendors’ declining transactions due to liquidity shortfalls or poor liquidity management; both transaction and service times.
(iv) Innovation: bundling M-Money with other non-M-Money services (or expansions to other lines of business); including total exits; changing vendor locations; hours of operations; vendor's accessibility to previously unbanked / unserved households.
(v) Service - Quality: consumers' views and experiences about safety, privacy, invasion, harassment, discrimination, and respect after visiting vendor points.
We will adapt Annan (2020, 2021) audit study protocols (and surveys) to measure vendor outcomes objectively (and subjectively respectively). We will combine the outcomes to derive an index of overall service quality.
*For consumers, we plan to examine the impacts on usage and trust outcomes.
(i) We will field questions that will provide subjective measures of trust in M-Money providers, market vendors, carrying out vendor-involved transactions on M-Money, consumers’ family and friends, commercial and rural banks, and the regulator of financial services in Ghana (Bank of Ghana) if they have heard enough about them to say.
(ii) We will supplement this with objective measures of trust measured from trust games between consumers and their local vendors. In its simple form: this entails a real monetary payoff game between the consumers (i.e., trustors) and one randomly selected (anonymous) local vendor (i.e., trustee). We impose vendor anonymity to mitigate against issues of social and individual preferences. Consumers will be endowed with 40GHS and will decide how much to send to another person (i.e., M-Money vendor). We will triple it, so that the vendor receives three times the amount of money the consumer sent. The vendor will then decide how much he/she wants to send back to the consumer. The total payoffs depend on the choices made between the consumers and vendor. Our objective measure of consumer trust will reflect the amount they sent and expected the vendor to send back to them.
*For (all market participants: vendors, consumers, nearby stores), we plan to measure their beliefs and expectations about potential competition effects if competition is introduced on: (i) misconduct / overcharging, (ii) reliability / illiquidity (transaction declines due to liquidity shortfalls), (iii) transparency / tariff posting behavior of vendors, (iv) location of vendor's retail store, (v) innovation / bundling of M-Money with other non-M-Money services at agent points, (vi) consumer trust, and (vii) consumer experiences and usage.
*We will combine baseline and endline (i) market census, (ii) surveys, (iii) field trust games, and (iv) audit studies to track vendor, customer, and business outcomes (including: revenues, profits, number of customers, business expenses, employment, wagebill, household expenses), supplemented with administrative data (on sales revenue, vendor entry / exit, consumers' usage) from the commercial providers to examine broader impacts, including persistence and entry of new vendors across the localities.
|
|
Field
Experimental Design (Public)
|
Before
Design: ~1/3rd of localities will randomly receive “no-entry” of new entrant vendors (Control program), while the remaining ~2/3rd of localities will receive entry of new entrant vendors (Treatment program). For the treatment localities, we will randomly vary the density of entering vendorships (i.e., new vendors who are not presently doing M-Money business but are local microentrepreneurs) from 1 to 3: half will receive 1 additional vendor each (Treatment I) and the other half will receive 3 additional vendors each (Treatment II). We refer to existing vendors as “incumbents” (averaging at 3-4 vendors per locality). Treatment I represents 33% increase, while Treatment II represents 100% increase in vendorship. We will use this to trace out impacts of the various exogenous vendor entry levels. In a locality, we will enroll only a random subset of eligible and existing microenterprises as new entrant vendors. Thus, overall, we create three experimental variations: (i) only a random subset of localities receive entry, so we can compare impacts of entered versus not; (ii) we vary the density of entry (1 vendor each=33% increase in vendorship versus 3 vendors each=100% increase in vendorship relative baseline vendorship), so we can trace out the equilibrium impacts of competition; and (iiii) we enroll only a random subset of eligible microenterprises, so we can compare business impacts on enrolled enterprises versus not enrolled.
|
After
Design: ~1/3rd of localities will randomly receive “no-entry” of new entrant vendors (Control program), while the remaining ~2/3rd of localities will receive entry of new entrant vendors (Treatment program). For the treatment localities, we will randomly vary the density of entering vendorships (i.e., new vendors who are not presently doing M-Money business but are local microentrepreneurs) from 1 to 3: half will receive 1 additional vendor each (Treatment I) and the other half will receive 3 additional vendors each (Treatment II). We refer to existing vendors as “incumbents” (averaging at 4-5 vendors per locality). Treatment I represents ~25% increase, while Treatment II represents ~67% increase in vendorship. We will use this to trace out impacts of the various exogenous vendor entry levels. In a locality, we will enroll only a random subset of eligible and existing microenterprises as new entrant vendors. Thus, overall, we create three experimental variations: (i) only a random subset of localities receive entry, so we can compare impacts of entered versus not; (ii) we vary the density of entry (1 vendor each=25% increase in vendorship versus 3 vendors each=67% increase in vendorship relative baseline vendorship), so we can trace out the equilibrium impacts of competition; and (iiii) we enroll only a random subset of eligible microenterprises, so we can compare business impacts on enrolled enterprises versus not enrolled.
|