Using Big Data to Manage Credit Risk: An Industry in Transition

Rahul Sood and Paresh Banka

It's no secret that the past decade has been a rocky one, financially speaking. For companies in the finance and banking sector, this means dealing with everything from irate consumers to stricter regulations. Managing credit risk — individually in determining an entity’s creditworthiness and organisationally by understanding how much risk a company can afford, given their current capital — has become more important than ever before. It has also become harder than ever.

Below are four contributing factors:

1. Emerging types of risk (cyber, model, contagion, etc.) will require new tools and techniques.


2. The public has less tolerance for high risk, so institutions need to avoid even the appearance of making preventable errors.


3. There is a need to reduce operating expenses, so costs involved in credit risk management must be kept manageable.


4. Regulations have started or will start affecting more operations while also becoming more detailed and stringent.

Current State of Credit Risk Assessment

There are several ways that financial organisations are navigating today’s stressed and demanding market conditions. One of the most common is risk modeling, which can be performed using several methods:

• Credit portfolio modeling, which assesses creditworthiness based on industry, geography, and credit grade (among other factors). These can be run on various scenarios, which allows businesses to simulate economic upturns and downturns and the impact of these conditions on credit portfolio values.


• Credit ratings, which estimate how likely an entity is to repay (or default on) debt.


• Limiting exposure by limiting the number of trades with an entity (counter-parties, bond issuers, product types etc.). By blocking trades once entities have reached a predefined limit, the organisation’s financial exposure is not overly influenced by any one entity.


• Applying stress testing to models based on historical data — e.g. modeling the effects of shocks and other scenarios on historical data. This mitigates some of the drawbacks of using past data and allows businesses to test specific risk scenarios.

Clearly, this is not a case of neglecting to use any safeguards. But there are problems with the current process. Inefficient data management means that the right data isn't always available when it's needed. There's little support for modeling group-wide risk, leading to problems generating complex risk measurements. The available risk management tools don't change parameters easily. There's a lot of time and effort wasted in manually creating reports and re-grading portfolios (which may not get re-graded as much as they should).