Sharp thoughts on Friday from Andrew Haldane, executive director for financial stability at the Bank of England, on the changing topology of the market — including the rise of high-frequency trading.
As well as analysing the rise of fragmentation, nano-speed arbitrage, co-location and even the causes of the flash crash, he brings up, for example, the changing nature of market-making and specifically the way the business now assesses risk.
Consider, for example, bid-ask spreads as an indicator of liquidity. How exactly are those bid-ask spreads determined nowadays?
In most cases, says Haldane, the spreads are a reflection of the risk market makers take to manage both an inventory-management problem (how much stock to hold and at what price to buy and sell, before that inventory loses value) and an information-management problem (the possibility of trading with someone better informed about true prices than themselves).
Given that, Haldane notes:
The bid-ask spread, then, is the market-makers’ insurance premium. It provides protection against risks from a depreciating or mis-priced inventory. As such, it also proxies the “liquidity” of the market – that is, its ability to absorb buy and sell orders and execute them without an impact on price. A wider bid-ask spread implies greater risk in the sense of the market’s ability to absorb volume without affecting prices. This basic framework can be used to assess the impact of the changing trading topology on systemic risk, moving from analysing market microstructure to market macrostructure.
The flash crash, however, taught us an important lesson here.
While on the surface the new market structure has contributed to tighter bid-ask spreads — partly due to the speed with which bids and offers can now be processed — it’s also given rise to a new form of risk. The fact that bid-ask spreads don’t always tell you the full story.
As Haldane notes:
But bid-ask spreads can sometimes conceal as much as they reveal. For example, by normalising on volatility, Chart 8 air-brushes out what might be most interesting: normalising volatility might normalise abnormality. It risks falling foul of what sociologists call “normalisation of deviance” – that is, ignoring small changes which might later culminate in an extreme event.
Essentially, he says, clear inefficiencies or abnormalities may be being masked by the way things are traded nowadays.
He goes on:
HFT liquidity, evident in sharply lower peacetime bid-ask spreads, may be illusory. In wartime, it disappears. This disappearing act, and the resulting liquidity void, is widely believed to have amplified the price discontinuities evident during the Flash Crash. HFT liquidity proved fickle under stress, as flood turned to drought.
While that in itself is not new news, Haldane says the critical point is that non-normal patterns in prices have begun to appear at much higher frequencies. He points to a recent study which suggests that “since around 2005, stock price returns have begun to exhibit fat-tailed persistence at 15 minute intervals. Given the timing, these non-normalities are attributed to the role of HFT in financial markets”.
This type of non-normality can be quantified via something known as the Hurst coefficient, derived from observations of Nile Delta flooding.
With that coefficient in mind, Haldane notes:
The Hurst coefficient summarises this behaviour in a single number. For example, a measured Hurst equal to 0.5 is consistent with the random walk model familiar from efficient markets theory. A Hurst coefficient above 0.5 implies fatter tails and longer memories. In his study, Smith finds that the Hurst coefficient among a selection of stocks has risen steadily above 0.5 since 2005. In other words, the advent of HFT has seen price dynamics mirror the fat-tailed persistence of the Nile flood plains.
HFT market-makers are thus bound to be influenced by the increased abnormality in prices because their inventory risk increases:
This change in price dynamics will in turn influence market-making behaviour. Consider the problem facing an HFT market-maker. They face inventory risk from market fluctuations and information risk from adverse selection. Pricing these risks means forming a guess about the future path of prices. The greater the potential range of future prices, the larger the insurance premium they will demand.
This has implications for the dynamics of bid-ask spreads, and hence liquidity, among HFT firms. During a market crash, the volatility of prices (σ) is likely to spike. From equation (1), fractality heightens the risksensitivity of HFT bid-ask spreads to such a volatility event. In other words, liquidity under stress is likely to prove less resilient. This is because one extreme event, one flood or drought on the Nile, is more likely to be followed by a second, a third and a fourth. Reorganising that greater risk, market makers’ insurance premium will rise accordingly.
On top of this increased tendency to compensate for inventory-risk, also comes the tendency to over compensate for information risk, perhaps even more acutely, says Haldane. As a result, pricing becomes near-impossible and with it the making of markets.
That, then, is how you end up in a situation where Accenture shares fall to 1 cent while the shares of Sotheby’s rise acutely to $99,999.99 – nothing more than the limits of an HFT market-maker’s quoting obligation.