Link (Network) Analysis for Strengthening Credit Risk Assessments

Introduction

Credit delivery has become sophisticated. At the click of a button, one can avail a loan, while behind-the-scenes algorithms work on determining the #creditrisk profile of the borrower, and the #creditdecision is typically made in a very short period of time.

Traditional Credit Risk Assessment approaches usually focus on the standalone #creditrisk profile of the borrower. In reality, there may be other factors impacting an individual's #creditrisk profile. Such factors could be a result of other banking products that he may be consuming or his relationship with other borrowers, which may have a bearing on his standalone #creditrisk profile; such factors may go unnoticed but eventually may prove to be the tipping point.

Let's look at this aspect in slightly more detail.

Credit Risk Assessments

#Credit Risk Assessments can be broadly categorized into two kinds:

• Standalone Assessment: A standalone #credit risk assessment is usually performed based on the borrower's individual characteristics such as demography, income, and past behavior. For a wholesale (corporate) customer, the factors under consideration could include financials, group/management credentials, external rating, and collateral (Popularly called as Five C's). Credit Scores/Ratings are examples of such stand-alone assessments.
• Collective Assessment: Collective Assessment involves assessing a borrower on the basis of not just standalone data but also by using collaborative (combined) information that may be available for a borrower. For example, while assessing the creditworthiness of a retail borrower, collective assessment could consider factors such as the following: (i) Other banking products which he is consuming and the risk factors emanating out of such products (ii) Relationship with other borrowers such as Spouse who has a risky behavior (iii) Joint actions between two or more borrowers to game the system through fraud or collision.

#Link Analysis (#Network Analysis) could help unearth relationships or linkages with other factors in a borrower's credit ecosystem.

What is Link Analysis?

#Link Analysis or #Network Analysis is a visual approach to establishing the nature of a relationship between different variables in a data set. In a typical financial institution example, relationships could be identified between variables such as customers, demographic records, products, and collateral.

An illustrative link analysis for loan data is provided below:

Usage

#Link Analysis (#Network Analysis) could help augment #credit risk measurement and management in the following ways:

Identifying Shared Risk Characteristics

On a standalone basis, a borrower may not exhibit certain risk characteristics, but on a shared basis the risk perception may be high. For example, a spending analysis of the spouse of the borrower may reveal that he/she has a high propensity to spend on gaming websites, which increases the risk of default of the borrower in question as well.

Collision

Evidence of collision can be observed through link analysis. For example, two related parties may hypothecate the same vehicle with two different banks. which can be spotted through a link analysis of the borrowers and vehicle data.

Other use cases for a collision could be identifying borrowers with the same details such as phone numbers, addresses, identification numbers, and even biometric details.

Fund Diversion

Identifying fund diversion could be a classic use case for link analysis. When funds are layered (routed) through several interconnected parties, link analysis could prove to the extremely useful in observing the trail of funds across multiple parties.

Fraud

Link analysis could be used for fraud detection as well. For example, it could be used in cases where the same smartphone was used to purchase from an e-commerce website from geographically separate locations, within a very short period of time.

Analyzing Social Media Data

The social standing of an individual could be analyzed by studying the network of friends/followers on social media. Likewise, for a company, link analysis could be useful in analyzing the network of group companies and related parties.

Conclusion

#Link analysis could provide a very significant add-on to the credit scores and risk ratings. For example, there could be a modifier to the credit score, considering his/her nearness (closeness) to other individuals (for example: Spouse, Members of the same family, or related Parties) who may be exhibiting abnormal credit behavior.

The enhancement of computing power and evolution of graph databases such as Neo4j have made #Link Analysis more easier and accessible. Of course, the generation of powerful link analysis requires reasonably clean data and expertise in the use of applications such as Python.

Nevertheless, #Link Analysis (#Network Analysis) would prove to be a very useful tool to be factored in #credit decisions.