In 2008 legendary investor Warren Buffett threw out a challenge to his peers and rivals: he invited anyone of them to invest $500,000 in a portfolio of hand-picked hedge funds whose performance, over the course of ten years, and taking expenses into account, would match or do better than the same sum that Buffett would put into in an unmanaged index fund.
Ted Seides of investment firm Protegé Partners took up the gauntlet. While Buffett’s half-million would simply track the aggregate fortunes of the five hundred large companies that make up the American stock exchange’s S&P 500, Seides selected five hedge funds that he, a Wall Street wizard of no mean ability himself, predicted would outperform them.
Nine years later Seides conceded an early defeat. Although he had done well, increasing the value of his pot by $220,000, Buffett’s gains weighed in at almost four times as much.
The point of the wager had been two-fold: that the high fees charged by hedge fund managers could potentially represent a complete loss of value for money; and that, in any event, returns on an investment spread widely within the fortunes of a healthy stock market would likely do no worse in the long run than one precision-engineered to succeed. The virtues of so-called passive investment had been spelt out.
Since then, automated financial investment models have become big business. Alongside current AI applications, human attempts to play the money market, for all their bravura, are now widely perceived as both needlessly expensive and senselessly risky.
Passive investment is the earliest and, by some lights, still the most efficient way of putting an investment fund on auto-pilot. The idea, invented in the 1970s by American financial strategist Jack Bogle, is simple but was, at the time, counter-intuitive: to invest money in funds that would simply track major stock market indexes rather than try to beat them. Low-maintenance, and therefore low-cost, the investments asked for little more human input than patience.
But there was always scope for automation to do more.
Since the 1980s those whose business is the active management of financial assets have sought out ways in which their stock-in-trade decision-making might be made more systematic, less discretionary and better informed by data-rich insight; in other words, more automated.
Investments given over to such control systems have come to be known as Quantitative funds. They use mathematical models and vast data fields instead of individually informed human judgement-calls to assess the relative attractiveness of potential assets and make predictions about the likely quality of returns.
Although the quantitative methodology has inherent limitations (to be offset, some suggest, by the less measurable insights that can be derived from complementary, qualitative research), there is no doubting the performance power of those hedge funds which – Bridgewater Associates and AQR Capital Management to name only today’s market leaders – espouse its approach.
The growing role AI has to play in the world’s money markets is clearly visible in the impact it is having on one sector in particular: the increasingly mainstream investment vehicles known as exchange-traded funds (ETFs). Traditionally a form of passive investment, these bundles of financial assets track the value of a market index, though they are also traded on stock exchanges through the day. Their popularity is a testament to, among other things, the well-established efficiency of the passive investment principle.
None of which means, of course, that one ETF performs as well as any other, nor that computer power and mathematical modelling don’t make a big difference. Leading ETFs are now starting to use proprietary algorithms to analyse such macroeconomic data as market volatility and interest rates with a view to predicting market changes and, on that basis, periodically rebalancing the composition of the fund.
The fact that basically passive, strongly automated and therefore strikingly low-cost investment structures are now the marketplace norm is a break with tradition from which there may be no going back – especially if the shift is understood within a larger demographic context.
The digital revolution has democratised, through an abundance of online services, the kind of investment opportunities that an earlier generation, typically older and of higher net worth, would have explored with the aid of expensive financial advisors. The online services – Betterment and Wealthfront are two of the key players – are popularly known as Robo-advisors.
These digital investment mechanisms deploy the same kind of algorithms that were pioneered in the early days of quantitative fund management and have been trickling down through the system ever since. Individual investors are allocated assets (typically via ETFs) automatically, with the assets subject to reinvestment and the allocation subject to rebalancing also automatically.
The intelligence of the platforms’ behaviour extends to practices such as tax-loss harvesting. The selling of assets at a loss in order to reduce short-term tax liability is, in fact, a good example of the kind of financial manipulation that is difficult to achieve manually but straightforward work for the right software.
Impressive as advances in automated investment techniques have been so far – advances which have benefited an unprecedented number of individuals as well, arguably, as the health of the stock market itself – their future is by no means certain. It is not known, ultimately, how good a fit artificial intelligence is for the unique dynamics of the marketplace. Can it do much more with the kind of data it currently has at its disposal? If it needs more, will there be enough – or could there be too much? Will the new AI insights be game-changers or dead-ends? And should their work be complemented by human economic theory, or will they somehow challenge and replace it?
Whatever the future, it will be a far cry from the days, not so very long ago, when technology meant ticker tape and slide rules, and when the best investments were made by people whose abilities seemed to defy analysis.