Predicting bankruptcy
A complexity science-based approach might prove useful for making predictions across all types of firms
The parliament passed India’s first national bankruptcy law in May. Why should it be news-worthy? Bankruptcy law provides an exit option to entrepreneurs/ventures that have failed, and venture failure in India’s start-up ecosystem has been estimated to be as high as 80 per cent considering start-ups that have survived 18 months since inception.
Bankruptcy law also provides for better debt recovery from larger enterprises that have been making losses for a while, that would allow banks and creditors to recover a fair portion of their credit. It is estimated that around 25 per cent of creditors in India get their money back after businesses they have lent to have gone bust, similar figures in Europe and the US indicate closer to 77 per cent recovery of credit extended.
Bankruptcy law is important because it ensures the efficient allocation or re-allocation of resources within an economy, and if India’s bankruptcy law is implemented well it should improve innovation and productivity in the economy. Hence, if bankruptcy law is likely to add fillip to the economy, it seems appropriate to find ways of predicting bankruptcy or to at least see it coming rather than wait for it to occur.
Many approaches have been used by academics and analysts for tracking insolvency and bankruptcy. A metaanalysis conducted a decade ago classified all these models into theoretical and statistical models, and artificially intelligent expert systems. The statistical and artificially intelligent models use accounting data while the theoretical models use market data. A British study found the theoretical models implemented using contingent claims were more accurate in bankruptcy prediction than the others. Models dependent on accounting data are likely reflecting historical trends while contingent claim models incorporating upto-date market information fare better.
All the above models consider either accounting data or market data for predicting bankruptcy. The big question is are these sufficient for the purpose? Are factors such as vision — strategy — business model alignment, types of leadership at work, inherent competencies developed by various firms not important, amongst others, or could all of these be gathered by a combination of accounting data and market data? The most successful contingent claim models used for bankruptcy prediction do not directly consider any of the above factors nor is it apparent that they even do it in an indirect manner.
The other question to ask would be — don’t we need separate models to understand the evolution of start-ups, SMEs, and larger enterprises or could a single bandwagon model predict bankruptcies across all types of firms? The answers to both are understandably complex which is where a complexity science-based approach might prove useful to find a way out.
Let us examine the venture exit phenomena first. There are around 19,000 technology-enabled start-ups in India as per the Economic Survey 2015-16. VCs have invested around $12 billion in the years 2014 and 15 in some of these start-ups, of which $7.3 billion alone was invested in 2015, as per Tracxn. This raises some interesting questions. Is VC funding important for venture success, if so at what stage might VC funding actually help a start-up? Has VC funding improved the survival rate or reduced the bankruptcy rate of startups? How many VC-funded ventures have successfully reached an IPO stage? To expand the scope of this line of thinking another question could be — Are starting resources including financial and intellectual important for a venture success or could ventures work their way through bricolage?
The above questions provide a logical basis for developing a model that might track venture success or failure. If we did an impact-influence map of the questions there would be complex nested layers of influences in the model which might require the use of say Deep Boltzmann Machines, among other approaches. Will such a model, even if it is reasonably successful in predicting venture failures, be able to predict SME or large enterprise bankruptcies? It does not seem likely at all. Thus bankruptcy prediction is a long way from a reasonably perfected science; however it is certainly worth investing in.