How AI will prevent another housing market crash

In the film Minority Report, a supercomputer predicts when someone is going to commit a crime, so they can be arrested before the crime is committed. It’s the stuff of science fiction, and yet like the best science fiction, it’s getting closer to reality. Data scientists can now predict where and when crimes will take place – and it’s accurate 90% of the time. But this type of forecasting isn't just useful for predicting crime - with enough data, it's possible to predict massive socioeconomic changes. This leads to the question: what if we could use the behaviours of individual people to predict and prevent economic disasters before they happen?

You may have noticed there has been some rather negative media coverage of the housing market lately. Everything that’s supposed to be up is down; everything that’s supposed to be down is up. The housing market is one of the cornerstones of the economy. When the housing market collapses, everything collapses. That was what came out of the subprime mortgage fiasco in 2008. Millions of people had been quietly heading towards default, and the banks hadn’t paid any attention until it was too late. Nobody saw it coming.

But what if we had? What if we had predicted that millions of people were going to default on their mortgages, and could have intervened to help them before it got out of hand? That’s the possibility that is arising out of the new generation of AI.

Here’s the truth: the bank doesn’t want to repossess your house. The bank wants you to pay your mortgage. The rate of mortgage repossession in the past decade has consistently declined; in 2011, 25,463 private individuals had their homes repossessed, while last year it was less than 3,000. Repossession of one house is a headache for a bank; repossession of a thousand houses is a nightmare. The bank would much rather step in to extend mortgage terms, offer payment holidays, switch appropriate customers to interest-only deals, or even help homeowners downsize. Repossession is now a last resort.

This is where AI comes in. By leveraging data from an individual’s lifestyle habits, AI can predict when someone may be heading into financial trouble. The indications can be subtle, but numerous. For example, if someone who frequents a high-end supermarket like Waitrose suddenly switches their weekly shop to Aldi, that may indicate their purse strings are tightening a little. However, it could also be that the Waitrose they used to go to shut down and was replaced with an Aldi, and it’s just their nearest store. One piece of data doesn’t reveal much. Maybe that same person begins splitting purchases on instalments more frequently – they could be broke, or they could be managing their money more efficiently. What if that person cancels their Netflix, Spotify, and other streaming services? Money troubles, or have they just decided to reduce their screentime? Individually, all of these things are innocuous. But what if they all happen at the same time? What if there are indications of a change in behaviour? Are they suddenly spending money on online gambling? Are they withdrawing lots of cash with the possible intention of concealing its use? Are their wider socioeconomic circumstances changing?

AI can be trained to monitor events and give a weighting to each outcome. It can consider multiple events at once, and decide if a behaviour change is innocuous, or forms a wider pattern of potential financial instability. It could then contact that person to ask if they want to connect with someone who can help them. It would work like this:

1. Risk monitoring: Through the utilisation of AI-powered algorithms, banks can continuously monitor various behavioural indicators to assess the financial stability of individuals in real-time. These indicators include not only increased credit card usage and gambling activities but also changes in spending habits, such as shopping at discount stores or splitting payments using "buy now, pay later" services. By analysing these patterns, AI systems can assign a risk rating to borrowers, providing an ongoing assessment of their financial health. This real-time risk monitoring allows banks to proactively identify individuals who may be at a higher risk of defaulting on their mortgage payments.
2. Early intervention: With the implementation of AI-based risk monitoring, banks can establish threshold levels for risk ratings. When a borrower's risk rating falls below a predetermined threshold, it triggers an alert for early intervention. The AI can contact the borrower via email, SMS, WhatsApp, or other messaging service, and connect them with a representative of the bank, who will offer support and assistance before the financial situation becomes too severe. Early intervention can involve engaging in open and proactive communication with the borrower to understand their challenges and explore potential solutions. By intervening at an early stage, banks can help struggling mortgage payers develop personalised plans to address their financial difficulties, potentially preventing the escalation of problems and reducing the likelihood of mortgage defaults.
3. Targeted assistance: Recognising that each borrower's financial situation is unique, banks can provide targeted assistance tailored to their specific needs. This assistance may include offering alternative repayment plans that align with the borrower's current financial capabilities, refinancing options to reduce monthly mortgage payments, or providing financial counselling to help improve budgeting and money management skills. By customising assistance, borrowers are more likely to regain financial control and find sustainable solutions that alleviate their mortgage payment challenges. These targeted interventions aim to empower borrowers, reduce financial stress, and minimise the risk of mortgage defaults. Furthermore, banks can collaborate with external partners, such as housing counselling agencies or community organisations, to provide comprehensive support services and resources to further assist borrowers.

The goal is not only to help individuals, but to ensure the financial stability of the economy as a whole. Individual mortgages are like blocks in a giant game of Jenga. One or two can be knocked out, but as that number grows, the situation moves closer to total collapse. Using AI to monitor behavioural data across the mortgage market to identify emerging trends, potential risks, and systemic vulnerabilities allows targeted measures and policies to address specific challenges before they escalate into broader financial crises. If AI is given access to comprehensive behavioural data, we can use it to make informed decisions, implement early warning systems, and enforce prudent lending practices, reducing the probability of a large-scale financial crash.

Of course, there will be some people who feel that this kind of monitoring conflicts with their right to privacy. It’s not a baseless argument. As AI becomes stronger, we will increasingly have to decide where we strike the balance. However, if AI enables us to spare future generations from another economic collapse, it feels a little difficult to argue against it. We've already traded away our privacy for a lot less than that.