Writing this blog with an intent to share learnings from a research conducted on Micro, Small, and Medium Enterprises (MSMEs) lending space. I am open to a dialogue on this subject. Happy reading!
From the sources of “Office of the US Trade Representative”, Micro, Small, and Medium Enterprises (MSMEs) are the backbone of the American, European, and widely global economies. The United States’ 30 million MSMEs account for nearly two-thirds of net new private sector jobs in recent decades. MSMEs that export tends to grow even faster, create more jobs, and pay higher wages than similar businesses that do not. But MSMEs face a tough challenge when it comes to liquidity and lending. An outstanding MSME loan of > $650 billion in the USA alone signifies the lending challenges.
Basel Capital Accords and MSME Lending:
The Basel Capital Accords and the recent financial crisis have provided renewed impetus for lenders to research and develop adequate default/failure prediction models for all of the corporate and retail sectors of their lending portfolios. In view of this, MSMEs face key challenges to capital access from lack of historical financial information, the track record of creditworthiness, and uncertain future cash-flows to name a few. Under these uncertainties, liquidity needs are getting aggravated as small business failure prediction is a hand-folding task. The new Basel Capital Accords guidance also states that the financial sector has to recognize MSME as a distinct client from the large corporate client. One argument is treating MSMEs like a retail client from a risk management point of view in order to lower capital requirements and realize efficiency and profitability gains. As building credit risk models for MSMEs are necessarily limited by data availability, the blending financial and non-financial information is the right approach.
- MSME financial information: typically five categories of variables – liquidity, profitability, leverage, solvency, and activity – to predict insolvency
- MSME non-financial information: Reflects company characteristics and aspects of operational risk such as financial reporting compliance, internal audit and trade credit relationships, as predictive variables of company distress.
SME Lending Risk Assessment Models:
In the USA, while the MSME bankruptcy is on gradual reduction, ~50% of businesses in a financially distressed state. As per the latest Statista SME survey, only 52% of respondents evaluated the overall state of their business positively. Another important statistic at a global level is on MSME startup success rate and longevity in the business. Only 9% of businesses have a chance to succeed beyond 10 years of existence. 80% of new businesses fail within their first year.
In the above context, it is increasingly important to develop appropriate risk models for MSMEs. The models available for large organizations have to be greatly contextualized for MSMEs with an appropriate criterion to boost the confidence of banks, institutional lenders, P2P lenders and private investors. Let us dive into building this foundational framework.
1) Expert Models: Risk and credit rating is derived based on the expert opinion, could be a simple yes or no responses analyzing the characteristics of the MSME loan applicants. A qualitative analysis is conducted by scoring the main factors of the credit, such as moral quality, repayment ability and the collateral of the applicants, the purpose and deadline of the loans. However, this method is highly dependent on the experience of experts and their tacit knowledge, which makes it a time-consuming task and brings fatigue and classification error.
2) Statistical Models: If the financial data of MSMEs is available and easily accessible, sophisticated statistical tools ranging from MDA, LOGIT, and PROBIT models can be deployed in failure prediction. One such popular method is Altman Z-Score where the probabilities of bankruptcy ranges are 95% for one year and 70% within two years.
i) The Original Z-Score formula for public manufacturing companies:
Original Z-Score = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 0.999X5
Bankruptcy probabilities with Z-score above 3.0 are not likely; 1.8 to 3.0 cannot be predicted-Gray area; Below 1.8 likely failure
ii) Model A Z-Score for private manufacturing companies: this model substitutes the book values of equity for the Market value in X4 compared to the original model.
Model A Z-Score = 0.717X1 + 0.847X2 + 3.107X3 +0.420X4 +0.998X5
Bankruptcy probabilities with Z-score above 2.9 are not likely; 1.23 to 2.9 cannot be predicted-Gray area; Below 1.23 likely failure
iii)Model B Z-Score for private general companies: this model analyzed the characteristics and accuracy of a model without X5 – sales/total assets.
Model B Z-Score = 6.56X1 + 3.26X2 +6.72X3 +1.05X4
Bankruptcy probabilities with Z-score above 2.6 are not likely; 1.1 to 2.6 cannot be predicted-Gray area; Below 1.1 likely failure
X1 = working capital/Total Assets. Measures net liquid asset of a company relative to total assets
X2 = retained earnings/total Assets. Measures the financial leverage level of a company
X3 = earnings before interests and taxes/total Assets. Measures productivity of a company’s total assets.
X4 = market value of equity/book value of total liabilities. Measures what portion of a company’s assets can decline in value before the liabilities exceed the assets.
X5 = sales/total Assets. Measures revenue-generating ability of a company’s assets.
The other models include Springate’s modified Z score, Fulmer’s index in 1984, and CA-Score method developed by Jean Legault. But the direct applicability of these models to MSMEs is limited. The right statistical approach for MSME failure prediction is factoring custom variables (example, MSMEs/their buyer’s defaults & delays in payments) on top of above models and increase the weighted relevance of small business financial performance influencers.
3) Behavioral Models: While analyzing the financial distress and failure, it is important to analyze human – managerial-decision making side of MSMEs alongside the commerce or financial dimensions (loan default and credit scoring) with the advances in cognitive psychology and neuroscience bundling with predictive behaviors and heuristics.
Even though the loan application variables result in a risk, there might be reasons the decision makers develop an inclination resorting to a limited set of mental shortcuts (or heuristics) to simplify things and move forward with lending money. Heuristic models include The Affected Heuristics (feeling the state of goodness or badness of lending), The Representative (Similarity) Heuristic for producing rapid probability decisions rather than consciously adopted procedures, The Availability or Recency Heuristic, and Anchoring & Adjustment. A few of the dynamic behavioral models which have seen the light include survival model (allow us to model not just if a borrower will default, but when – by analyzing survival data with hazard function) and Markov transition models (especially useful for modeling revolving credit with highly variable credit usage). Predominantly logistic functions (Sigmoid) is used to perform binary classification in predicting behavior enabling bilk financing of small loans.
4) New Age Models: With the rapid development of machine learning, more and more Artificial Intelligence methods are applied for risk assessment and credit scoring, such as artificial neural networks (ANN), genetic algorithm (GA) and support vector machine (SVM).
5) Hybrid Models: For the practical purposes, not one model fits to predicting precisely the MSME failures. The logical way out is developing hybrid models combine two or more of above models to tailor closely to the context of the MSMEs and categories under focus while evaluating lending risks and credit scoring.
In summary, developing a robust risk assessment and credit rating framework and integrating it with evolving alternative online lending platforms not only help to reinforce the lending confidence but also constantly learn the human patterns of MSME lending ecosystems’ key stakeholder augmenting the cognitive lending intelligence.