Search

We study repayment and delinquency in the context of an alternative financial product that enables the purchase of a large asset---a solar panel home system---while offering complete repayment flexibility. Using a large administrative data set on daily repayment of 38,400 borrowers in Tanzania over 5.5 years, we perform unsupervised pattern analysis to classify repayment behavior. We show that borrowers with fluctuating incomes use the loan's flexibility more and that farmers in particular adjust their repayment to cash flow. We further find that use of flexibility is linked to repayment difficulties, yet does not automatically lead to default. Our results indicate that low-income households can finance large assets through innovative financial approaches that allow aligning payments to financial circumstances.

We study repayment and delinquency in an innovative loan contract that offers borrowers a wide range of flexibility. Using a large administrative dataset, we perform unsupervised pattern analysis to study how borrowers repay within the framework of this loan. We identify eight clusters that can be grouped into three distinct repayment types. We show that borrowers with fluctuating incomes and limited consumption smoothing resources use the loan’s flexibility more and that farmers in particular adjust their repayment to cash flow. Finally, we show that high use of the loan’s flexibility is associated with repayment difficulties, yet typically does not result in eventual default; whereas borrowers who face repayment difficulties that are likely driven by economic shocks face a high probability of default.

Data extracted from software repositories is used intensively in Software Engineering research, for example, to predict defects in source code. In our research in this area, with data from open source projects as well as an industrial partner, we noticed several shortcomings of conventional data mining approaches for classification problems: (1) Domain experts’ acceptance is of critical importance, and domain experts can provide valuable input, but it is hard to use this feedback. (2) Evaluating the quality of the model is not a matter of calculating AUC or accuracy. Instead, there are multiple objectives of varying importance with hard to quantify trade-offs. Furthermore, the performance of the model cannot be evaluated on a per-instance level in our case, because it shares aspects with the set cover problem. To overcome these problems, we take a holistic approach and develop a rule mining system that simplifies iterative feedback from domain experts and can incorporate the domain-specific evaluation needs. A central part of the system is a novel multi-objective anytime rule mining algorithm. The algorithm is based on the GRASP-PR meta-heuristic but extends it with ideas from several other approaches. We successfully applied the system in the industrial context. In the current article, we focus on the description of the algorithm and the concepts of the system. We make an implementation of the system available. © 2020 The Authors