Insolvency Prediction Techniques for Debtors
A summary of my recent article in the Journal of Business Law
The world's companies have never in history held more debt than they do today. Business bankruptcies in the US and Germany have overtaken GFC levels. We know the economic damage that widespread corporate insolvency can do to the economy. In this moment, the economic value of tools for predicting a company's insolvency in advance, and thereby putting it in a position to take preventative steps, is huge.
Predictive techniques, particularly those enhanced by artificial intelligence (AI), are opening up new possibilities for companies to identify and address financial distress early. These tools, once the domain of creditors, could be a game-changer for businesses themselves, enabling them to chart a course to safety rather than sinking under the weight of financial trouble.
This post summarizes an article I have recently had published in the Journal of Business Law. The article explores insolvency prediction techniques and their potential to transform how businesses manage financial distress. In the paper, I suggest that there is potential for these tools, traditionally used by creditors, to be leveraged to help debtors detect early warning signs of insolvency and take preventive action.
Why Focus on Debtors?
Insolvency prediction techniques have actually been around for a long time, but historically they have been used to help lenders assess the risk of extending credit. However, in the paper, I argue that these techniques hold untapped potential for helping businesses themselves. By recognizing the signs of financial distress early, businesses could proactively restructure, cut costs, or otherwise adjust their strategies to prevent insolvency.
For instance, financial ratios like the Altman Z-score have long served as a benchmark for predicting financial trouble. By analyzing variables such as working capital and total assets, the Z-score produces a simple numerical result that indicates the likelihood of insolvency. Yet, in the paper, I suggest that newer AI-based techniques like artificial neural networks (ANNs) offer even more promise, leveraging vast datasets to identify patterns with increased accuracy.
A Look at the Tools
1. Statistical Models: Older models like the Z-score are still widely used due to their simplicity and transparency. A Z-score above 3 indicates strong financial health, while a score below 1.81 signals high insolvency risk. These models are easy to use but may oversimplify complex financial realities.
2. AI-Powered Techniques: Artificial neural networks (ANNs) appear to represent a significant advancement. By analyzing vast amounts of structured and unstructured data—from financial statements to market disclosures and even news reports—ANNs provide deeper insights. However, in the paper, I argue that these tools come with challenges: they require extensive data and their results are less transparent and explain-able.
Challenges and Opportunities
Implementing insolvency prediction tools for debtors is not without its hurdles:
Balancing Accuracy and Simplicity: While ANN-based models are more accurate, their complexity and data requirements may make them impractical, particularly for smaller businesses. In the paper, I argue that a hybrid approach could be a solution: simpler tools like the Z-score for initial assessments and more advanced techniques for deeper analysis.
Addressing Errors: Predictive models must balance two types of errors:
False Positives: Flagging a healthy business as distressed could disrupt operations and scare off creditors.
False Negatives: Failing to identify a distressed company in time could lead to missed opportunities for rescue.
Building Trust in AI: The opaque nature of AI models—their “black box” problem—poses a challenge for adoption. Greater transparency and explainability are essential, especially if these tools are to be used in legal or governmental contexts.
Implications for the law
Like many areas of law, these AI-powered tools have significant potential to disrupt.
If the sorts of predictive techniques I consider in the paper do become more accessible, widespread and widely used, they could reshape the legal landscape.
For example, in many jurisdictions, directors have special obligations when their company nears insolvency. These are sometimes called "pre-insolvency duties". A question that I explore in the paper is that if we have these techniques that arm a director with better information about the likelihood of insolvency, must the director take that information into account? What should be the consequences if they do not? In other words (and this is a question that will arise in many areas of law as artificial intelligence becomes more mainstream in society), does AI, by improving our potential capabilities, raise the legal standards we should hold ourselves too?
A Vision for the Future
Looking ahead, I see immense potential for insolvency prediction techniques to create a more resilient business environment. Realizing this vision will require ongoing refinement of these tools, attention to their legal implications, and efforts to make them accessible to businesses of all sizes. The possibilities are exciting: with the right systems in place, these tools could help countless businesses avoid financial collapse and contribute to a healthier, more dynamic economy.
I hope this summary gives a glimpse into the ideas explored in my article. If you’re interested in reading the full paper, it is in the Journal of Business Law, Issue 8 of 2024. Alternatively, please feel free to reach out to me directly for a copy. You can also listen to a podcast summary of the article here.
Really clever use of the AI podcast format Harry