Big Tech’s Mission to Replace the Banks Will End in Tears
by James Meadway
2 December 2020
From contact tracing to home working to forecasting lockdown breaches, Covid-19 has dramatically accelerated the digitisation of everyday life. It has also therefore reinforced the power of technology companies. The market valuation of the six largest tech firms rose by 38%, or about $1.7tr, in just the first half of this year, even as the rest of the economy tanked. For a period in May, shares in Zoom (a company unknown to most of us 12 months ago) were worth more than the seven largest airlines in the world combined.
Yet in the ever-expanding empire of big tech, one land remains to be fully conquered: finance. From big tech’s point of view, monetary systems are ripe for colonisation. Finance is and always has been an information-based business – the most basic operation of making a loan depends on knowing how likely the would-be borrower is to default, a rudimentary data-processing task – and that’s exactly the business big tech is in.
Of course, the conquest is in full swing. Amazon, Apple, Facebook and Google all now offer payment systems. Facebook acquired a European banking licence in 2016, and proposed a new global cryptocurrency, Libra, three years later. Apple has partnered with Goldman Sachs to launch a credit card. Google has linked up with US financial institutions to provide a digital front-end to their banking services. Big tech’s move into finance has, however, not always been smooth. Libra received hefty pushback from global central banks and regulators, with major economies from the US to India insisting on tight regulatory control, even outright bans (despite these setbacks, Libra is set for launch in early 2021, although in reduced form).
The government likes to think of fintech, “financial technology”, as an unproblematic new growth industry for post-Brexit; chief secretary to the Treasury John Glen has claimed it will be “vital” in “ensuring the economy bounces back, post-coronavirus”, whilst Boris Johnson used this year’s Tory conference speech to talk up the sector’s “excitement and verve”. Yet as a new report from the Finance Innovation Lab highlights, fintech presents profound and underappreciated problems.
First, the spread of data analysis into finance isn’t mainly about the plucky start-ups the Tories love to talk about. Banks and other financial institutions hold reams of data on their customers – and not only that, they use Machine Learning (ML) to process it; according to the Bank of England, two-thirds of financial services companies are using some form of ML, the foundation for Artificial Intelligence (AI). Big tech, however, is even better at accumulating and processing that data. And if there’s a plausible growth plan for the smaller fintech companies out there, it tends to lie in waiting for one of the giants to swallow them, just as the tech giants have been doing for years with startups. The economics of data – the more you have, the more valuable it becomes – pushes everything towards monopolies.
But there are other, bigger problems. Take credit-scoring. Our current credit-scoring system, overseen by companies like Experian, uses only a tiny amount of the relevant data available to generate a score. This means that large numbers of people are denied loans through no fault of their own, but simply because the financial system holds inadequate data on them. More data would mean a more accurate score. Zest AI, a fintech startup founded by Google’s former Google chief information officer Douglas Merrill, aims to do exactly this: use all sorts of data that credit score services don’t efficiently use – “All data is credit data,” says Merrill – to deliver a better score.
The upside of this is that it could means the markets for loans work more efficiently. The downside is that if “all data is credit data”, almost anything you do could in theory be used to generate your credit score without you necessarily knowing how or why.
More worryingly, nor would anyone else necessarily know how or why your data was being used; the foundation of Artificial Intelligence (AI) means allowing computers to make decisions that humans do not necessarily understand.
This really matters when, for example, an AI is racist. There is research being done into making AI decisions at least potentially knowable, and GDPR requires decisions made by an AI to be intelligible on demand. The danger is a drift towards China’s social credit system: access to finance and other services becoming dependent on machine-made decisions that can appear entirely arbitrary.
This unintelligibility also threatens to create new forms of systemic risk. We saw what happened in 2008, when poorly-regulated, and poorly-understood, financial systems are liable to overreach. The banks back then had extraordinary volumes of data, and apparently cutting-edge statistical techniques to manage making and selling loans, to the point where they believed the risks involved had been managed into irrelevance. This, as we now know, turned out to be wrong: the banks had missed the possibility that their activities had created a new, previously unseen kind of systemic risk. More data about what has happened does not always tell you what could.
The problem is potentially even worse with fintech. At least in 2008, it was possible to know in principle where the system had broken down, in the first instance: in the use of default statistics, a poor model of future default probability, and the excessive production of collateralised debt obligations. With an AI, it is not even necessarily possible to understand how your technology has produced a particular result. And if the banks of 2008 were deemed too big to fail, requiring massive government interventions to support them, big tech most definitely is. While the central banks have somewhat recognised these risks, their tendency has been to think there can be “nothing new under the sun” in the fusion between finance and tech. Their naivete should alarm us.
Capitalism is, of course, what is driving big tech to colonise more and more of our lives with data – in the case of finance, almost certainly beyond the point that the risks of doing so can be contained. The challenge is to separate the profits motive from the collection and use of data. The Lab report recommends, amongst other things, new non-profit and publicly-accountable financial institutions. The think-tank Common Wealth has produced recommendations for democratising big tech more generally. Another, more focused option is the creation of central bank digital currency, allowing people to open bank accounts directly with the Bank of England – something the bank itself is examining. But as long as profit alone is allowed to drive developments in fintech, the risks of catastrophic overreach remains.
James Meadway is an economist and Novara Media columnist.