Research in progress.
Project last updated on: 29 Jun 2017.
Understanding the efficiency effects of a credit bureau: A cluster randomised controlled trial
Over the past 50 years MFIs have worked on a model to provide credit without the usual infrastructure required by traditional banks. Increasingly, however, MFIs are looking for new ways of doing business, and as their scale and income increases, they are able to incorporate some of the features of modern banking, for example the use of a credit bureau for sharing information. These new features comes with possible benefits – for example improved incentives and clients selection – but also costs – for example exclusion of poor clients – that were not present in traditional micro-credit models. This project aims to shed light on the costs and benefits of a credit bureau system, studying the impacts of a credit market information system in Pakistan.
The objective of this research project is to understand how a microcredit credit bureau changes the behaviour of borrowers and lenders, and how these responses depend on the structure of the market. In theory, a credit bureau has two main effects. First, it enables lenders to screen out bad borrowers and to lend more to good borrowers. Second, it can cause behaviour change among borrowers, who may change the loans they apply for, or their repayment of existing loans. The functioning of both of these mechanisms depends on the equilibrium responses of firms and therefore on market structure. Further, while these mechanisms are theoretically relevant, there is a dearth of empirical work providing concrete evidence.
To study our questions, we exploit the roll-out of a national credit bureau for microfinance lenders in Pakistan. This bureau already holds credit records for essentially all borrowers, which were provided by MFIs upon its inception. However, not all lenders currently have access to the credit reports generated by the system.
Our study uses a two-part randomisation. The first part randomises the roll-out of access to the credit bureau information at the branch level with a small number of large partner microfinance lenders. We use detailed geographic data on the location of lending branches to ensure that we treat branches in both more and less competitive areas, enabling us to study the role of market structure. The second part randomises a telephone information campaign to borrowers, informing them about the credit bureau and how it works. By randomly informing some borrowers who already have loans we can study changes in repayment behaviour, by randomly informing other “potential” borrowers we can study loan application behaviour.
By using both administrative and credit bureau data over two years we will track borrowers’ loan applications, lenders’ lending decisions, and borrowers’ repayment behaviour, to analyse how each of these is affected by the presence or knowledge of the bureau, and the role of market structure in changing behaviour. Do borrowers become more or less likely to apply for loans, and do they apply for bigger or smaller loans? Do borrowers become less likely to default? Do lenders approve bigger loans and what borrowers do they choose to lend to?
Our hope is to inform theory, both qualitatively and quantitatively, of the relative importance of key behavioural effects of the credit bureau, and thus to inform policymakers about how to better implement credit market information systems in future. For example, if it turns out that the primary role of the bureau is deterring moral hazard (default by borrowers) then there may be a rationale for subsidising the bureau since the deterrent depends on take-up by lenders. If the primary role is screening, subsidies are less important.