The Future of Algorithmic Decision Making in Big Data Lending
Big data sets are providing financial institutions with a wealth of new insights that can be used to improve upon the traditional model of determining credit-worthiness, but at the same time these insights present a new set of challenges in ensuring that automated decision-making isn't used to indirectly express bias against factors like race, gender, or political affiliation.
Since William Fair and Earl Isaac began marketing the first data-driven credit rating system in the mid-1950s, their "FICO" score has become the standard for determining consumers' credit-worthiness in the United States. Today, Fair Isaac says FICO scores are used in more than 90 percent of loan-making decisions. But for all its ubiquity with consumer credit decisions, the FICO score still relies predominantly on a single data point—payment history, which produces only a narrow view of a consumer's financial situation and leaves millions of so-called "credit invisibles" with little or no history on which a credit decision can be based. This is what's driving a growing number of financial service providers to explore less traditional consumer data.
What Big Data Means for Financial Service Providers
Big data can be succinctly defined as very large data sets that require a significant amount of computing power to analyze effectively. Data can be collected from a wide array of sources and compiled to create extremely large data sets that can then be mined for information that fits certain parameters. In the financial services field, the array of information available from a rapidly growing number of data brokers can provide a big-picture look at individual consumers that can help determine whether these individuals are good risks for credit arrangements. This approach also enables investment advisors to determine the best options for their clients and their funds. Firms utilizing big data analysis services can combine data sets from a multitude of sources to extract actionable insights for in-house use or for their clients, which can streamline the integration of big data into the everyday workflows of financial institutions and lending companies.
Some nascent credit solutions that are heavily reliant on big data include:
- Peer-to-peer lending platforms
- Microlending (both small business and personal)
- Venture capital firms
- Small dollar loan markets (predominantly payday loans)
Increasing Access to Credit Through Peer-to-Peer Lending, Microloans, and Small Dollar Lending Platforms
Peer-based lending platforms are churning big financial data through machine learning algorithms to estimate return on investment and calculate risk exposure. The same idea has been used in the now $5 billion-a-year payday loan space. Without access to the big three credit reporting agencies, alternative lending platforms now rely almost entirely on automated big data analysis to make credit decisions. Decisions that can be based on virtually anything—from where consumers attended high school, to what types of magazines they read.
Big financial data consists of an extremely large number of data points that relate to individual consumers and their families. These data points include information from public records, personal internet activity, and privately held data sold or shared between companies. Perhaps some of the most notable factors being utilized by modern credit assessment algorithms are those that pertain to consumer behavior and begin to press ethical lines. With a consumer profile that consists of things like social media activity, product reviews, spending habits, or even insights gleaned from friends and family, financial institutions can deliver a surprisingly complete picture of overall character and the probability of repaying a loan as agreed in the future.
Big Data and Small Towns: Making Loans Personal Again
One way to understand the impact of big data as it applies to financial decisions is to look at the early days of financial services in the United States, especially in smaller towns and rural areas. Before there were formalized systems of credit rating, banks and lending institutions made decisions based on a few basic factors:
- The duration of residence in the area
- The character references available for the individual
- The amount of collateral for the loan
- The history of reliable repayment of debts in the past
The same factors are used by many modern lenders to determine whether an individual is eligible to receive a loan or credit arrangement. The primary differences between the earlier processes and modern banking practices are in the types of technology used to collect and analyze this information.
Just as the rise of the credit bureaus and automated information revolutionized the way in which banks assessed investments and loan opportunities in the past, big data-driven algorithms are transforming the way financial institutions eliminate missed opportunities and spot risks before they show up on credit reports. The information derived from big data can paint a picture of an individual consumer almost as clear as the personal impression of the small-town banker considering a lending arrangement for a member of the local community.
The Motivations for Consumers and Businesses
Modern lenders are looking at more than just credit histories and ratings to determine whether individual borrowers constitute acceptable risks and to fine-tune the rates offered to those approved for lending arrangements. Social media postings and other publicly available data can round out the picture for lenders, allowing them to make more informed decisions when deciding to extend or deny credit to applicants. Finding ways to leverage available data to increase the information available to financial institutions can be an effective strategy for expanding their lending portfolios. By taking all relevant factors into consideration, companies can put the power of big data on their side when making decisions to lend or invest in the consumer marketplace.
Encouraging consumers to opt in to big data sets and to provide personal information voluntarily in exchange for lower rates, improved terms, and better offers can benefit lenders with increased accuracy for risk calculations. This can create a win-win scenario in which customers who choose to opt in can count on better rates and fees, while financial institutions can reduce their exposure to risk in the future. Customer referrals are another way in which businesses can bolster their data and create a more comprehensive picture of current customers and their acquaintances. This in turn enables banks to make more accurate assessments of the ability of their customers to take on more credit or to qualify for loans and other financial arrangements.
The Future of Big Data for Financial Institutions: Mitigating the Potential Risks for Consumers
The most important virtue of big data mining and analysis is the ability to drill down to individuals and groups to assess specific risk profiles. By combining social media posts, credit histories, behavioral habits, and other personal data, financial institutions can make the most accurate decisions to approve or deny applications from consumers and to determine how much credit a consumer can reasonably maintain. In some cases this can be used to increase access to credit, but when used irresponsibly—or even malevolently, this access to data can have chilling effects.
One grim picture of the potential for abuse in big financial data can be seen in China's social credit system. Instituted in 2014, the system uses big data analysis to produce ratings for individuals based on the activities they engage in. With a nationwide rollout expected to be complete in 2020, the social credit system will derive data from a wide range of sources online and in the consumer credit system. There have already been reports of people being denied access to dating websites and having their ability to travel restricted. Those who engage in approved behaviors, by contrast, receive discounts on utility bills, preferential employment status, and other state benefits.
While these circumstances are unlikely to be mirrored in the U.S., banks and lending institutions already have access to the same types of data used to create China's social credit system. This could have a particularly profound impact in the subprime lending industry where consumers already face a greater risk of abusive lending practices and higher interest rates. Data-driven systems can predict that consumers with certain racial affinities or political inclinations are more likely to accept a more expensive loan, and while it's not legal to base credit decisions on these factors, correlative findings could arrive at this conclusion without any explicit knowledge of legally protected characteristics.
Better Understanding Big Data Sets to Create a Positive Path Forward
As more information about individual consumers becomes readily accessible through big data applications, it's imperative that fairness and transparency are made priorities in our decision-making algorithms, both to prevent bias and to avoid casting consumers into a class with predetermined outcomes.