Why history can't help understand the current crash
This latest crisis is unprecedented in many ways – and that makes it hard to apply lessons from past crashes.
The March meltdown was a tough time for almost all investors. Even quantitative investing strategies (see below), many of which are intended to be largely market neutral – ie, they try to exploit anomalies between the price of individual securities without taking a view on whether the overall market will rise or fall – were badly hit.
By the beginning of May, the HFR Global Hedge Funds index was down by around 4.5% from the start of the year, according to Bloomberg, but a subindex of equity market neutral funds had lost twice as much.
Why these funds performed poorly is fairly simple. In a crisis, past relationships between different securities break down – which leads to runaway losses for anybody who bets on them returning to normal quickly. Computer models that base their trading on history are very vulnerable at times like these. A human will grasp that any past patterns involving, say, cruise ships or airlines might no longer be useful when global travel is grinding to a halt. A computer lacks that insight.
Investing in the dark
Still, with the initial panic over, we’re as much in the dark as any misfiring quant model when making sense of what comes next. There’s little precedent for a global economic disruption on this scale.
Only three events in the modern era offer any real parallels: the crash of 1929-1933, the bear market of 1973-1974 and the financial crisis of 2007-2009. All three saw severe global market panics during times of huge upheaval (the Great Depression, the collapse of the Bretton Woods exchange-rate system and the oil crisis, and the global credit crunch).
In each case, stocks dropped by around 50% or even more (US stocks fell by almost 90% from 1929 and UK ones by almost 75% from 1973). All took several years to regain past highs in nominal and real terms (a decade or two in some cases). Dividends fell by 20%-30% in real terms (more than 50% for the US in 1929) and again took years or even decades to recover, although nominal dividends in the inflationary 1970s rebounded very quickly.
What’s curious today is that dividends are already being slashed: markets expect S&P 500 dividends to fall by 25% by 2022, eurozone ones to fall 40% and FTSE 100 payouts to fall more than 50%. Yet the S&P 500 is now down less than 10% this year and UK and eurozone stocks by less than 25% – and none were down more than 35% at the low. This is hard to square. It’s a struggle to see how this crisis can be over without the priciest markets becoming cheaper, unless dividends snap back. Still, the quant quake reminds us that history may yet be a poor guide in such an unprecedented event.
I wish I knew what quantitative investing was, but I’m too embarrassed to ask
Quantitative investing – also known as systematic investing – is the broad term for a wide range of different investment strategies that employ sophisticated computer-based mathematical models to identify and carry out trades.
Common quant strategies include factor investing, which looks at characteristics of an asset such as the profitability of a company. To take a simple example, an investing algorithm might buy stocks that appear cheap on measures such as price/book value and sell stocks that look dear. This is a classic value investing approach that a human manager might follow. But the algorithm will take into account far more metrics than a human manager might, while ignoring other considerations that might sway a human – such as whether they think the firm’s management is good.
Risk parity strategies allocate between different assets depending on how volatile they are and how much volatility the investor wants. Trend-following strategies (also known as managed futures or commodity trading advisers (CTAs)) look for trends in the price of assets and make decisions based on those. Statistical arbitrage is based on relationships that normally exist between the price of different securities. Event-driven strategies identify patterns around events such as earnings announcements or corporate actions. Systematic global macro decides whether to invest in countries, assets or sectors based on quant models that use economic data.
Proponents argue that quantitative investing helps remove biases and emotion from investment decisions, ensuring that they are based purely on data. While quant models are initially programmed by people, the role of humans in making individual investment decisions is greatly reduced. With some of the newer artificial intelligence-powered approaches, even the designers may not fully understand why a system chooses specific trades.