The Market Mind Hypothesis, page 40
Apart from being able to better time-lock with other data (in our case excess volatility and volume-indicators), this pattern’s objectivity is also more likely to lead to correlation across noise traders. This is important as noise trades “will only matter if they are correlated across noise traders. If all investors trade randomly, their trades cancel out” (Shleifer and Summers 1990, p. 23). Related to such copied trading behaviour, cognitive science has shown that (experiencing) imitation of behaviour in general impacts the quality of human exchanges generally. In particular, it increases trust which facilitates trade (e.g. Van Baaren et al. 2003).
Our focus was on excess volatility (via our VD indicator) as the measure of market noise and its relationship with volume and other trading statistics. We used these indicators as rudimentary reflections of trading behaviour (no doubt, more advanced analyses are conceivable with additional datasets).
According to the literature, a positive relationship generally exists between volatility and trading activity. In the case of volume, for example, this is theoretically explained via the mixture of distribution hypothesis (MDH) and the sequential information arrival hypothesis (SIAH). This literature also includes research on microstructure. An early source in that regard is Tauchen and Pitts (1983). Referring to noisy personal information, they requested from future research insight into “the stochastic specification” of the dynamics involved “as more traders enter the market”. We hope to provide this via the stochastic interpretation of RP by showing that volume-indicators, reflecting traders’ reactions, change according to the accumulation in market noise around noisy informational events. This suggests, we propose, that increased market noise ‘invites’ on occasions more (smaller) traders because—combined with their own build-up in neuronal noise—it makes them sensitive, i.e. trigger happy, in their discretionary trading around such events. Again, inter-brain synchronisation due to shared attention and intention seems to play a role.
9.2.3
Readiness Potential
Readiness Potential (RP) is the English translation of the original German Bereitschaftspotential, suggesting a “readiness” to act. Technically, RP reflects cortical brain activity (measured via EEG), in particular a build-up of neuronal activity, ahead of voluntary action or movement. Tests of RP are generally done in a laboratory setting. Subjects, whose brains are monitored, are asked to (repeatedly) perform an instructed movement spontaneously without any forethought or pre-planning. Researchers are particularly interested in the brain activity leading up to those movements which they measure in narrow windows of a few seconds. Researchers are also interested in the timing of the movement itself and the perceived timing of the decision to move.
The RP has a rich history going back to the 1960s when it was first discovered. In 1983 neuroscientist Benjamin Libet (with colleagues) published seminal research showing (presumably) that the feeling of having decided to act, the so-called conscious decision, emerges long after the neuronal decision has already been made in the brain. Libet’s account emphasised the role of the unconscious, while the conscious decision is tacked onto the decision-making process after-the-fact without playing any causal role. His argument rested on the time course of the RP, whereby the gradual build-up of neuronal activity in the pre-motor area precedes not only the onset of movement, but also the conscious decision to initiate the movement.
Since then the RP has always been presumed to reflect a process of planning and preparation for movement. More recently, however, Schurger, Sitt and Deheane (2012)1 challenged this presumption by applying a stochastic accumulator model, including a so-called Libetus-interruptus task: subjects perform the standard Libet task, but are told that they may sometimes, i.e. randomly, be interrupted by an auditory “click,” in which case they are to perform the movement (in this case a button press) as quickly as possible. The authors found a slow build-up of neuronal activity preceding fast reactions to the click, that began long before the click itself. In short, they argue that RP might reflect ongoing stochastic fluctuations in brain activity whose crests tend to coincide with the onset of movement. This latter interpretation has since received additional support. Of particular interest is a study by Murakami et al. (2014) which found evidence for an accumulator process in area M2 of rats (homolog of human premotor cortex) performing a task where the rat could spontaneously abandon waiting for a large reward and instead opt for an immediate and certain, albeit smaller reward.
The gist of the argument from this stochastic RP interpretation is the following: when the imperative to act is weak, for example due to lack of evidence, the decision’s timing that leads to movement is partly determined by ongoing sub-threshold fluctuations in brain activity. To be clear, this timing is the precise moment at which the decision threshold is crossed. Time locking to movement onset ensures that these slow fluctuations are recovered in the event-locked average in the form of a gradual build-up or accumulation. By this account the RP does not reflect a goal-directed process and the real “decision” to initiate action is a threshold crossing event that happens very close in time to the onset of the movement itself. We translated this in the collective setting of NT.
9.2.4
Parallels Between NT and RP
The stochastic RP interpretation suggests that the process of neuronal accumulation preceding the onset of a trade may involve autocorrelated random fluctuations that influence the decision and thus the tipping point of execution. In particular, the precise moment, i.e. timing, at which the decision threshold is crossed is largely determined by spontaneous sub-threshold fluctuations in neuronal activity when the imperative for a trade, i.e. information, is weak or absent.
Arguably this is often the case in an investment environment characterised by uncertainty and incomplete knowledge (or even lack of efficiency, for that matter). But it seems especially applicable to technically oriented noise traders. Why? Because timing is key in their motivation. In addition, the evidence to back-up their trade is, well, noise. In short, this is why we stated earlier that, like subjects receiving the Libetus-interruptus in a laboratory RP-setting, technical traders are primed to respond to the market’s version of random interruptions in the form of moving-average crossings. Of course, not all moving-average crossings result in a ‘noisy’ trade. But those that do trigger it do so because some decision variable has already been building up to threshold. To wit, based on our working hypothesis we forecast that the specific profile, i.e. steepness, in the build-up of excess volatility (again, quantified as VD) can explain some characteristics in trading behaviour, as captured in the variance in volume-indicators.
Moreover, in markets uncertainty (or rather risk) is generally expressed in volatility, with higher volatility reflecting more uncertainty and thus, arguably, less evidence. So, increasing volatility weakens the (informational) imperative to trade, thereby increasing the susceptibility of trading decisions to subthreshold neuronal fluctuations. And here we arrive at the potential reflexive connection we mentioned before between markets and minds. Not only does this connection suggest a possible neuronal explanation for a positive feedback-loop, connected (cross-brains) through traders, between noisy prices and noisy neurons. It also provides us with a testable hypothesis.
In short, we propose that noise trading activity following technically significant events, i.e. (Fibonacci-based) moving-average crossings, is related to the accumulation of excess volatility ahead of these events and has an internal, i.e. neuronal, contributor. We summarize our interpretation of RP translated into a NT setting in Table 9.1.
Table 9.1:Readiness Potential (RP) versus Noise Trading (NT).
Item in RP setting Item (proxy) in NT setting
Neuronal noise (build-up)
Event, i.e. interruption
Decision
Action/movement
Subject type/behaviour
Excess volatility (build-up)
Price crosses its MA
Trading decision
Execution of trade
Trader type/behaviour (via volume and other trading statistics)
For example, RP’s “voluntary self-initiated movements” are translated in the NT setting as pushing buttons on a keyboard, mouse clicks, or other actions that execute trading decisions. The broader “imperative to move” in RP we interpret in the NT settings simply as the urge to trade, possibly motivated by active management, including the drive to reach a particular target or outperform some benchmark.
9.2.5
Findings from the Pilot Project
Here I will summarise our main findings from our investigations. The first focused on determining the “general profile” of the build-up in excess volatility around moving-average crossings to see if, pattern wise, it approximately reflects that of RP’s neuronal noise. The second focused on the relationship between build-up in excess volatility and changes in trading statistics around moving-average crossings, which we then interpret in terms of trader type/behaviour. Here we allowed ourselves some leeway, for two reasons. First, there is no practical comparison for RP’s ‘subject type/behaviour’ in the NT setting. Second, to infer trading behaviour we had to make do with the dataset we were given.
In Figure 9.1, the chart at the top (1A) shows the general profile of build-up in neuronal noise, culminating in a spike, around an RP-event. Ignoring different timescales, etc., this is roughly mirrored by the two charts below it, reflecting the general profile of build-up in price noise around moving average crossings (both up [in blue] and down [in red]) for both the SPY (1B) and the USO (1C).
Source: Schotanus and Schurger (2020, p. 207)
Figure 9.1: Neuronal and price noise.
In conclusion, both data sets—neuronal (1A), respectively price noise (1B and 1C)—obviously result from different measurements and, consequently, show variations in profile. The clearest differences are that the build-up in price noise is less gradual and that it lingers. This is probably partly due to the relative lower frequency of the price data and thus in measurements. Additionally, we also relate this to our overall hypothesis and see it as a consequence of the non-lab collective conditions. Whereas the EEG Build-up (1A) is an aggregation of EEG measurements on isolated individuals, in markets (1B and 1C) individual neuronal noise and collective price noise reflexively interact, possibly including contagion (see the discussion in the next section). Despite these differences the overall similar shape of their related graphs broadly confirms what we hoped to find.
Our second investigation consisted of running a large number of regressions of VD Build-up against trading statistics around moving-average crossings. Our general finding was that VD Build-up could explain some aspects of the (assumed) typical behaviour of discretionary technical traders. In other words, the data showed footprints of these traders as being active around ‘default’ moving average crossings.
9.2.6
Discussion of Results
Fischer Black famously stated that “noisy trading puts noise into the prices” (Black, 1986, p. 532). Inspired by this we applied the RP framework of decision and action to NT. As we hypothesised, the variance in volume-indicators following our choice of technical events often suggests a trading behaviour that is characteristic for discretionary technical traders whose timing is sensitive to a combination of weak evidence and accumulating neuronal noise. We submit that the latter’s footprints are, collectively for this group, reflected in excess volatility. In other words, we have potentially identified an ‘internal’, that is an embodied, contributor to Black’s “cumulative noise” (Black, 1986, p. 532). Implicitly therefore, market noise and mind noise could reflexively feed on one another, with noise traders particularly susceptible to participate. This could help explain, for example, bubbles and crashes in stock prices, which are patterns emerging from, in our case, (accumulated) randomness.
The earlier comment on ‘correlated’ noise traders is relevant in that regard—especially considering the possibility of cross-brain synchronisation—and we would like to make one further clarification of the type of similarities between the mind and the market, this time metaphorically and anecdotally. Specifically, we link spontaneous volatility to the EMH, in the sense of clarifying how a pattern of sustained build-up during major turning points challenges market efficiency.
Initially it seems that the self-interest underlying choices by individuals naturally results in those choices being independent. This then supports the argument of the random walk leading to equilibrium. How could spontaneous volatility disturb this? In other words how could randomness turn into a pattern which shifts equilibrium? Dehaene (2014) describes a cascade of neuronal activity as an “avalanche” (p. 131) culminating in a “global ignition” (p. 135), with neurons bursting into widespread coordinated activation, leading to the emergence of consciousness. Echoing some of Kelso’s work, he compares this to the way an audience begins with a few random claps and then erupts into synchronous applause. In similar terms, but now implying such cross-brain synchronisation, here is a quote from Sornette (in Bastiaensen et al., 2009) characterising the build-up of market activity, when a crash is spontaneously emerging from randomness:
The audience expresses its appreciation with applause. In the beginning, everybody is handclapping according to their own rhythm. The sound is like random noise. There is no imminence of collective behavior. This can be compared to financial markets operating in a steady-state where prices follow a random walk. All of a sudden something curious happens. All randomness disappears; the audience organizes itself in a synchronized regular beat, each pair of hands is clapping in unison. There is no master of ceremony at play. This collective behavior emanates endogenously. It is a pattern arising from the underlying interactions. This can be compared to a crash. There is a steady build-up of tension in the system … and without any exogenous trigger a massive failure of the system occurs. There is no need for big news events for a crash to happen. (Bastiaensen et al., 2009, p. 1; emphasis added)
In that regard, our proposition provides cognitive support to Black’s assertion that “noise in the sense of a large number of small events is often a cause factor much more powerful than a small number of large events can be” (Black, 1986, p. 529).
Of course, by “jumping on the bandwagon” (DeLong et al., 1990, p. 379) rational speculators could also contribute to the accumulation of excess noise ahead of the technical events we described. Even if their ex-ante motivation seems rational, our argument applies to them as well: the exact timing of their ‘front-running’ trade is likely to be disproportionally affected by their respective neuronal noise, particularly when excess volatility is increasing, weakening any evidence of actual technical events (let alone any fundamental information) that acts as the ‘rational’ imperative to trade. In other words, as long as the reason to trade is motivated by an information source that is noisy (in our case a crossing of a moving average) and coloured by uncertainty (reflected in volatility), all discretionary traders, including rational front-runners, are susceptible to RP’s accumulation breaking the threshold that triggers the decision to trade. This argument, however, does not hold for so-called mechanical (or system) traders whose decisions are objective, in the sense that they are coded and outsourced to computers, with execution fully automated (but see our comment on machines later).
In many respects, this project was speculative. Still, both NT and RP are important topics in their respective fields with, we think, potential linkages that deserve further research. Appendix 2, for example, contains a brief description of a potential follow-up project that will involve human traders in a real-life setting so that we avoid artificial laboratory conditions. For that purpose we will use non-intrusive tools to gather both high-frequency market and behavioural data that will help to expand upon the thesis of this pilot project. Our goal is to build a (machine-learning) model that shows how human decision making together with modern information delivery systems contributes to reflexive market dynamics. We anticipate that it could contribute to improving risk management.
Hopefully it can also inspire others to follow-up in similar research directions. Obvious suggestions include using different securities, frequencies, and/or events (e.g. based on fundamental data). Also, our research could potentially be enriched with data from (retail) day-trading accounts, whereby technical market events can be time-locked to actual trades.
Finally, the big elephant in the room: what about ‘the machines’? As we already pointed out, we know that accumulation of randomness, i.e. the build-up of noise, toward a tipping point is a phenomenon that occurs elsewhere in nature. However, we prefer to leave this for others to contemplate.
9.3
The Market Speaks its Mind
I need to know what is happening in the markets … I hooked up a music synthesizer to the computer, linked it to the interface between the computer and quote screen, and generated a program that would give a musical summary of the markets. I used piano tones for stocks, strings for interest rates, the cello for short-term rates, and the violin for the 30-year bond. The Japanese yen was registered with the high flute, corresponding to the favorite instrument in Japan, the shakuhachi. The English horn, the French horn, and the Alpenhorn stood in for the other currencies.
