Credit to SMEs? Robust Lending Decisions with Exploratory System Dynamics Modelling and Analysis

Abstract

This project’s commissioner, the head of ING’s emerging markets credit trading desk developed a loan structure called Revenue Participation Loan to fit the needs of SMEs in emerging markets. He now wants to combine this new product with innovation in his approach to uncertainty. With these types of loans SMEs agree to pay a fixed percentage of their revenues (a ‘cut’) to the lender until an agreed multiple of the principal is reached. The advantage of this structure is that it tolerates the sales fluctuations, but the implication is that it is unknown how long it will take to repay the loan. In the face of this uncertainty the bank needs analytical tools to select the best candidates and make robust lending decisions. No analytics can exactly predict the future. However, a good model-based analysis usually leads to better decisions than flipping a coin. But that too depends on good quality models and on how they are treated. A good addition to the well-established accounting models seems to be System Dynamics (SD) modelling. SD’s dynamic approach allows building and simulating stock-flow models that better represent the non-linear systems of companies. To make sense of the uncertainties in these models the Exploratory Modelling and Analysis (EMA) tools are proposed. The objective of this project is to investigate in what ways SD modelling combined with an exploratory approach could support lending decisions and monitoring of an SME credit portfolio. Traditional business plans aim to create the model that most accurately predicts the future performance of the company. Lending decisions based on such best-guess models are problematic given the deep uncertainties inherent in the mid- to long-term future. Opposed to this ‘consolidative’ approach, an exploratory study aims to systematically analyse a wide range of plausible future scenarios. The aim becomes to better understand the nature and implications of uncertainties and to devise robust measures that perform well over a wide range of possible futures. Such analyses can be supported by existing EMA tools also developed at TU Delft. Two case studies based on existing loans are presented to explore how SD modelling of SMEs could be performed and to illustrate the capabilities offered by the EMA workbench. Although these cases featured two very different companies a relatively similar framing of the problem can be recognized: the availability of cash balance is a critical enabler of these companies’ growth. Based on the SD models and their uncertainties considered the EMA tools can plot the range of scenarios that can occur. The most important indicator from the point of view of the bank is the time it takes to repay the loan. Using plausible uncertainty ranges for each input parameter the possible outcomes range from around 4 years to infinite repayment time. Feature selection algorithms then can be used to understand what are the most influential model parameters determining this outcome. Further analysis can reveal what combinations of uncertainty ranges will likely lead to the most (un)desirable scenarios. Finally, a robust optimization tool is introduced that can determine the optimal loan size and cut taking into account the uncertainty surrounding key exogenous variables. The study performed on the two cases could be performed during the assessment of the companies that are willing to receive a loan. However, this has some implications on the nature of the assessment. It became clear that more information needs to be asked and the nature of questions has to be broadened to what is commonly considered ‘soft information’. Although SME managers might not be used to it, explicit, clear and quantitative expression of causal relationships is at the core of SD modelling. Historical data records and a skilled modeller can make the inquiry manageable. The detailed modelling might be demanding for both the bank and SMEs, but the effort can pay off when more company models are analysed together. A combined analysis is envisioned in which the company models are embedded into the relevant macroeconomic environment, therefore allowing a consistent portfolio-level analysis. A higher level of detail allows for broader detection of weak signals of change and can also lead to more efficient monitoring through shared understanding between bank and company. It is recommended that the steps presented in the case studies are performed for a real company assessment to gain real-time experience with the EMA tools. The insights gained should be used to continually develop and improve the process of application. Despite its added value, the EMA approach and analytic tools probably cannot replace the need for good human judgement. But as a supporting companion for decision-making, they can give a sharper and more colourful picture of the uncertain future.