Machine learning is transforming the financial sector more than anybody could have ever predicted. This technology might be more important than ever during the pandemic, as financial institutions discover that many traditional protocols aren’t nearly as effective.
One of the most significant changes brought by advances in machine learning is with the loan underwriting process. It indirectly affects loan underwriting by changing credit scoring models, but some of these changes go to the heart of the loan underwriting process itself.
Towards Data Science analyzed several dozen papers on the use of machine learning in loan scoring. They found that it will play a huge role in the future of the profession.
Companies like Blue Line Loan are likely to use more cutting-edge loan underwriting tactics. This could include machine learning.
How has machine learning shaped the loan underwriting process?
Some experts have raised a variety of concerns about the conventional approach to credit scoring and its role in loan underwriting. In order to appreciate the criticisms, it is important to understand credit scoring in the traditional loan underwriting context.
FICO scores are determined by analyzing credit history, credit utilization, length of credit history, types of credit and recency. Each of these factors is assigned a different weight, with credit history and utilization accounting for two-thirds of the overall score.
Lenders take a hard look at the FICO score, as was well as a few other variables. They must also pay close attention to debt to income ratios, total amount of outstanding debt, certain lifestyle choices and other actuarial factors.
Although some of these factors make a lot of sense, others are archaic principles that don’t reflect modern actuarial standards. Even some underwriting standards that makes sense in many instances might be inappropriate in others. Loan underwriters need to recognize the nuances of different applications and account for the dynamic nature of the lending market.
This is where machine learning can be beneficial. There are a number of shortcomings in modern underwriting processes that machine learning can account for.
What are some potential benefits of using machine learning in loan underwriting?
Many experts have stated that machine learning can help overcome a variety of limitations with loan underwriting. The AI Research and Advisory Company has addressed the biggest benefits, a couple of them are listed below.
Accounting for “credit invisible” applicants
Lenders struggle to make appropriate determinations on lending with customers that have no credit. Millennials that didn’t go to college tend to be most likely to fall into this category, since many of them didn’t take out credit cards or student loans.
Many lenders have been tempted to reject applicants without any credit. However, this has reduced their ability to maintain adequate loan volume.
Instead, some of them are using machine learning to analyze pools of borrowers that receive loans without any prior credit. They can look for commonalities between these borrowers to determine the likelihood that an individual applicant will default. Although these machine learning algorithms are still in their infancy, they have proven to be highly effective so far.
Evaluate concerns with borrowers with variable income
The rise of the gig economy has led to a growing number of consumers with variable income. This has created some complications for lenders. Borrowers with regular 9-to-5 jobs can often get approved just by showing a paystub, while independent contractors and business owners need to show a couple of years of tax returns. This is a huge issue for unconventional members of the workforce that have recently started a new gig.
Lenders are also using machine learning to account for these shortcomings. They will probably start to use new algorithms to evaluate various factors, such as the type of business, the applicant’s work history and thoroughness of their business plans.
Are these machine learning algorithms going to prove to be even more important during and after the pandemic?
There is no question that the pandemic has created a lot of complications. Many people are having a harder time getting access to capital due to the economic implications it has created.
This means that machine learning might be even more important for the loan underwriting process than ever. Here are some ways that machine learning might prove most useful:
- Predictive analytics should probably be able to accurately forecast which unemployed people will bounce back after the pandemic ends. Rather than looking solely at their current employment status, it will evaluate their long-term income prospects, which could be helpful if they have enough of a cash reserve to cover a months’ worth of payments.
- Machine learning will identify the evolving nature of the economy. These algorithms are able to propose changes to risk scoring calculations to account for this new reality.
- Algorithms might be able to account for some COVID-19-related health risks, such as whether or not an applicant has been vaccinated and will be able to actively resume work. However, it is important to make sure that they don’t utilize these in a discriminatory way that might violate the Civil Rights Act.
Machine learning has introduced a number of beneficial changes for the loan underwriting profession. These changes might be even more welcome as challenges from the COVID-19 pandemic continue to mount.