The effects of information sharing on moral hazard in credit markets - Evidence from a randomized evaluation in the Philippines

We design an information campaign to increase knowledge about a private credit registry that shares information on borrowers among microfinance institutions in the Philippines. The information campaign is conducted in lending centers of a partner microfinance institution that offers group loans in lending centers of 5 to 40 borrowers. The intervention is randomly assigned at the center level.

We study the introduction of a private credit registry for information sharing. Our Partner Microfinance Institution (MFI) does not really use the credit registry for borrower selection and screening. Borrowers are not aware of potential information sharing with other MFIs even though they sign a waiver giving consent to information sharing knowledge is non-existent.

We conduct an information campaign to exogenously vary the knowledge of the credit registry and possible consequences for borrowers. We use the exogenous variation to study the effects of information sharing on borrower’s investment and repayment behavior. We can clearly identify the incentive effect of information sharing on borrower’s since a) we exclude selection effects and potential changes in the borrower pool since MFI does not use credit registry yet and b) identify effect on ex ante moral hazard (project and effort choice) and ex post moral hazard (repayment performance).

Three treatment arms are designed:
1) Standard campaign: information on credit registry
2) Detailed campaign: additional information that voluntary default will be especially noted in the credit registry
3) Control campaign: no information on credit registry; information on importance of repaying loan
4) Control group: no information campaign

The information campaign essentially contains information on the consequences of non-repayment with and without the credit registry. In paticular, during the 15-20 minutes lasting ‘information campaign’, interactive visualizations are used to explain
a) the functioning of the loan cycle & the importance of repayment for loan renewal as the clients know it
b) that no new loan will be received when the old one has not been paid back
c) that in the presence of another MFI, it might be possible to get a loan elsewhere despite of default
d) the functioning of the credit registry and the type of information collected and shared (name, address, loan outstanding, client assessment (good vs default) and reporting date)
e) the implications of the registry for loan default (no access to other loans)
This is followed by a short discussion of which factors influence loan approval (1) within the partner MFI and 2) within other MFIs). It is highlighted that with the registry in place, it is even more important to repay the loan.
Three different types of interventions take place
1. standard campaign: includes all elements a)-e) mentioned above
2. detailed campaign: as standard campaign, but in d), it is explained that loan default is classified into different categories that allows the MFI to distinguish between voluntary and involuntary default
3. control campaign: only parts a)-b) and discussion of factors influencing loan approval within the partner MFI.

2016-08-01

2016-08-31

Effort provision (in work/ household business),
Repayment,
Monitoring,
Peer pressure,
Usage of alternative credit sources,
Investments

The following variables will be constructed: effort, use of alternative credit sources, monitoring, and peer pressure.

For each variable, we have a set of questions from household surveys that adress different aspects.
In our analysis, we will use the individual questions as well as an index combining different questions to get a singular measure for 1) effort, 2) use of alternative credit sources, 3) monitoring, and 4) peer pressure.
For the index calculation, we use the ordinal alpha since we have 3-point Likert Scales (the commonly used Cronbach's Alpha is not applicable for Likert Scales with fewer than 5 points).
We also construct an index for competition as an explanatory variable.

We first elicit baseline values of variables of interest in a random subset of our study population via an individual survey. We then administer the above described intervention (i.e. information campaigns) in randomly selected centers. Shortly after the intervention, we re-survey the same individuals in the midline and approximately one year after the intervention, we administer the endline survey.

We conduct individual phone surveys with approximately 10 randomly selected clients in each of 200 different centers to elicit baseline values of variables of interest. Shortly after the end of the baseline survey, 8 research assistants visit 178 randomly chosen centers to conduct the information campaign on the credit registry.

Three different types of interventions take place
1) Standard campaign: information on credit registry
2) Detailed campaign: additional information that voluntary default will be especially noted in the credit registry
3) Control campaign: no information on credit registry; information on importance of repaying loan

The treatments, i.e. intervention types + control group, are randomly assigned to borrowing centers. First, centers were randomly assigned to the treatment of receiving information on the credit registry (i.e. standard or detailed campaign) with 147 centers in this group vs. 156 in the control group. Then from the control group centers 31 were randomly selected to receive the control campaign.

Once all information campaigns have been conducted, we administer the midline survey (and approx. 1 year later the endline survey) to collect panel data on the individual level.
Note that not all centers that are surveyed receive an intervention, these are the ‘pure control’ centers. Together with the control campaign, this allows us to estimate the effect of visiting a center and merely highlighting the importance of repayment. Comparing standard and control intervention thus provides the causal effect of information regarding the credit registry, holding constant the information on importance of repayment. Based on differences in individual behavior in standard and detailed campaign, we can classify voluntary and involuntary defaulters.

In addition, note that not all centers that receive an intervention are surveyed, such that we can separately estimate the effect of surveying based on administrative repayment data.

Randomization done in office by a computer

Borrowing centers consisting of 1-8 borrowing groups with 5 borrowers each

Yes

Number of borrowing centers included in study: 303
o Treatment 1: Standard campaign: 72
o Treatment 2: Detailed campaign: 75
o Treatment 3: Control campaign: 31
o Control: 125

We have two data sources: 1) Household phone survey with randomly selected clients from the participating centers. The planned number of borrowers being interviewed is 2000. Actual number of surveyed borrowers: in Baseline: 1941 in Midline: 1898 in Endline: currently ongoing 2) Administrative data on all clients in 303 participating centers from 2012 to 2017

Treatment randomly assigned to borrowing centers with the following number of centers in each treatment
o Treatment 1: Standard campaign: 72
o Treatment 2: Detailed campaign: 75
o Treatment 3: Control campaign: 31
o Control: 125