The Permanent Signature
Forty years. Five market crises. A complete technological revolution. The fingerprint never disappeared.
We have established that the fingerprint exists across markets. Every asset class. Every continent. Every exchange. The universality across space is complete.
But universality across space is not enough. A sceptic could argue that the signatures are a product of this particular era. That the rise of algorithmic trading amplified feedback. That electronic exchanges created memory that did not exist on the trading floor. That the fingerprint is an artefact of modernity, not a fundamental property of markets.
To close that argument, we need to demonstrate universality across time.
Has the fingerprint always been there? Or is it new?
We rolled a five-year measurement window across forty years of data, from 1984 to 2026, and computed the fingerprint at every quarter using Detrended Fluctuation Analysis. We then examined five natural experiments: moments when the structure of markets changed dramatically. If the fingerprint survived every transformation that financial markets have undergone in the last four decades, the temporal case is closed.
Forty Years
Figure 7.1 is the temporal equivalent of the scatter plot from Episode 4. Where that chart showed all markets at one point in time, this chart shows all time at once.
Figure 7.1: Rolling Hurst exponents of absolute returns, measured via Detrended Fluctuation Analysis in five-year trailing windows at quarterly intervals, across all available markets from 1984 to 2026. The bold dark line shows the cross-market median. The shaded band shows the 25th to 75th percentile range. The red dashed line at H = 0.5 marks the random walk baseline. Five vertical dashed lines mark the natural experiments: Black Monday (1987), the transition to electronic trading (2000), the Global Financial Crisis (2008), the Flash Crash and rise of HFT (2010), and the COVID crash (2020). The lower panel shows the number of markets available at each date, growing from 36 in the late 1980s to 68 by the mid-2000s. The cross-market median never approaches 0.5. In forty years of data, the Hurst exponent has fluctuated between approximately 0.64 and 0.82, but it never descends to the level that would indicate independence. The interquartile range sits permanently above 0.5.
The chart speaks for itself.
The cross-market median Hurst exponent has never, in forty years, fallen to 0.5. It has fluctuated. It has risen and fallen with market conditions. It has responded to crises, regime changes, and structural transformations. But it has never once reached the value that would indicate a random walk.
The decade-by-decade averages confirm the stability. In the late 1980s to 1994, dominated by pit trading, open outcry, and manual order execution, the median Hurst exponent was 0.757. In the 1995 to 2004 decade, spanning the transition to electronic trading, it was 0.707. In the 2005 to 2014 decade, encompassing both the Global Financial Crisis and the rise of high-frequency trading, it was 0.722. In the most recent period, 2015 to 2023, it was 0.720.¹
The range across four decades is narrow: 0.707 to 0.757. The fingerprint does not grow or shrink. It does not emerge with technology or fade with regulation. It persists, decade after decade, through every structural transformation the market has undergone.
The fingerprint is permanent.
Five Natural Experiments
Rolling windows show stability. But the most powerful test of a mechanism is what happens when the environment changes abruptly. If feedback is the cause, then moments when feedback conditions change should produce observable shifts in the fingerprint, not the disappearance of memory, but shifts in its intensity that correspond to the change in feedback.
We identified five natural experiments: events that dramatically altered market structure, regulation, or participant composition.
For each event, we compared the fingerprint in the five years before to the five years after, across all available markets. This window is long enough to produce stable estimates and short enough to isolate the effect of each event.²
Figure 7.2: Before-and-after comparison of the fingerprint around five natural experiments. The left panel shows Hurst exponents before (blue) and after (red) each event. The right panel shows tail exponents alpha on the same basis (lower alpha means fatter tails). In the Hurst panel, all bars remain well above 0.5 both before and after every event. The GFC produced the largest Hurst shift (+0.083), consistent with the intensification of feedback during the deleveraging cascade. In the tail exponent panel, Black Monday produced the most dramatic shift: alpha fell by 1.46, as portfolio insurance cascades amplified tails dramatically. The Flash Crash saw tails thin afterward (alpha rose by 0.76), consistent with circuit breakers successfully interrupting the most extreme feedback loops. But in every case, both before and after, the fingerprint is present. No event eliminated it. No structural change erased it. Feedback survived everything.
Black Monday, 1987
On October 19, 1987, the Dow Jones Industrial Average fell 22.6 percent in a single day. Portfolio insurance strategies created a feedback cascade: falling prices triggered automatic selling, which drove prices lower, which triggered more selling. Circuit breakers did not exist. The cascade ran until it exhausted itself.³
Before Black Monday, the cross-market Hurst exponent was 0.727. After: 0.695. A modest decline, consistent with post-crash regulatory interventions designed to dampen extreme feedback. The tail exponent fell by 1.46, the largest shift in our dataset, as the crash produced dramatically fatter tails. But the fingerprint remained firmly in persistent territory. The crash amplified feedback temporarily. The regulations that followed modestly dampened it. Neither created nor destroyed the underlying memory.
Electronic Trading, 2000
The transition from pit trading to electronic order books fundamentally changed market microstructure. Latency collapsed. Order flow became visible. Algorithmic participants entered. The entire execution infrastructure was replaced.⁴
Before: H = 0.716. After: H = 0.726. A slight increase. The fingerprint did not merely survive the transition to electronic trading. If anything, the increased speed and transparency of electronic markets amplified feedback slightly by making price information available faster to more participants.
The Global Financial Crisis, 2008
The collapse of Lehman Brothers triggered the most severe deleveraging cascade in modern history. Margin calls forced liquidation across every asset class simultaneously. Correlation spiked. Liquidity vanished. The feedback loop between price, margin, and forced selling operated at full intensity.⁵
The Hurst exponent leapt from 0.683 to 0.767, the largest memory shift in our dataset. The crisis did not weaken the fingerprint. It intensified it. Feedback operated at maximum intensity during the deleveraging cascade, and the statistical signature responded exactly as the mechanism predicts: more feedback, deeper memory. Tail exponents fell by 0.33, consistent with fatter tails during the crisis.
The Flash Crash and Rise of HFT, 2010
On May 6, 2010, the S&P 500 fell nearly 10 percent in minutes before recovering. The event exposed the fragility of high-frequency trading ecosystems and led to new circuit breakers designed to interrupt feedback cascades.⁶
After the Flash Crash, tails thinned substantially (alpha rose by 0.76), the largest tail-thinning shift in our dataset. This is consistent with circuit breakers successfully truncating the most extreme feedback loops. But memory persisted: H = 0.769 before, 0.738 after. The circuit breakers trimmed the tails. They did not touch the memory.
COVID, 2020
The COVID crash of March 2020 was the fastest bear market in history. The S&P 500 fell 34 percent in 23 trading days. The feedback cascade was global, simultaneous, and amplified by algorithmic risk management systems that reduced exposure automatically as volatility spiked.
Before COVID: H = 0.648. After: H = 0.713. An increase of 0.065. Tails fattened modestly (alpha fell by 0.39). The pattern is consistent with every other crisis: feedback intensified, memory deepened, tails fattened. The fingerprint did not weaken. It responded.
What Survived
Consider what the fingerprint has survived.
It survived the largest single-day crash in market history. It survived the complete replacement of the physical trading infrastructure. It survived the worst financial crisis since the Great Depression. It survived the introduction of circuit breakers designed specifically to interrupt feedback cascades. It survived the fastest bear market ever recorded.
It survived the transition from human pit traders to algorithmic execution. It survived the rise of passive investing. It survived the explosion of derivatives markets. It survived the globalisation of capital flows. It survived regulatory reforms, exchange mergers, and the entry of high-frequency trading firms that reshuffled the participant base entirely.
Nothing killed it. Nothing weakened it. Nothing moved it below the random walk boundary.
The fingerprint is not a product of technology. It is not a product of regulation. It is not a product of any particular era, market structure, or participant composition. It is a product of the one thing that has not changed across four decades: the fact that market participants observe price and condition their behaviour on what they observe.
Feedback is permanent because observation is permanent.
As long as participants can see price, they will react to it. As long as they react to it, feedback will operate. As long as feedback operates, the fingerprint will persist. This is not a historical accident. It is a structural inevitability.
The Complete Case
Seven episodes in, every dimension of the argument is closed.
Universal across markets: sixty-eight markets, eight asset classes, every continent.
Universal across signatures: memory, persistence, and fat tails co-occur in every market.
Causally demonstrated: removing feedback eliminates the fingerprint. Restoring feedback restores it.
Mechanistically explained: a phase transition at roughly twenty-five percent feedback intensity separates the random walk from reality.
Temporally stable: the fingerprint has persisted through four decades and five structural transformations.
The case is complete.
What remains is the closing argument. The next two episodes step back from the data and ask what it all means. Episode 8 synthesises the full body of evidence into a single, cohesive picture. Episode 9 delivers the verdict.
Next
Endnotes
References
- The decade-by-decade stability of Hurst exponents is consistent with findings in: Rama Cont, “Empirical Properties of Asset Returns: Stylized Facts and Statistical Issues,” Quantitative Finance, 2001. Our contribution extends this observation with systematic rolling-window analysis across 68 markets over 40 years using Detrended Fluctuation Analysis (DFA), demonstrating that the signatures persist continuously with remarkably stable intensity. DFA is preferred over R/S for rolling windows because it has lower finite-sample bias and is robust to non-stationarity. See: C.-K. Peng et al., “Mosaic Organization of DNA Nucleotides,” Physical Review E, 1994; and Kantelhardt et al., “Multifractal Detrended Fluctuation Analysis of Nonstationary Time Series,” Physica A, 2002.
- Natural experiment methodology. For each event, we defined a “before” window of 1,260 trading days (approximately 5 years) ending at the event date, and an “after” window of 1,260 trading days beginning at the event date. For Black Monday (1987), 504-day (2-year) windows were used because insufficient data preceded the event for 5-year estimation. All markets with sufficient data in both windows were included. For each market and each window, we computed: Hurst exponent via DFA on absolute returns, and Hill tail exponent with k = sqrt(N). Cross-market averages were computed for the before and after windows.
- The Black Monday crash of 1987 is widely attributed to portfolio insurance strategies that created a positive feedback loop between price declines and automated selling. See: Report of the Presidential Task Force on Market Mechanisms (Brady Report), 1988; and Mark Rubinstein, “Portfolio Insurance and the Market Crash,” Financial Analysts Journal, 1988. Our data shows that Black Monday produced the largest tail exponent shift in our dataset (alpha fell by 1.46), consistent with the extreme feedback amplification that characterised the event.
- The transition to electronic trading has been studied extensively. Terrence Hendershott, Charles Jones, and Albert Menkveld, “Does Algorithmic Trading Improve Liquidity?,” Journal of Finance, 2011. Our finding that the fingerprint slightly strengthened after electronification is consistent with the hypothesis that faster information dissemination accelerates feedback rather than eliminating it.
- The GFC deleveraging cascade is documented in: Markus Brunnermeier, “Deciphering the Liquidity and Credit Crunch 2007-2008,” Journal of Economic Perspectives, 2009; and Tobias Adrian and Hyun Song Shin, “Liquidity and Leverage,” Journal of Financial Intermediation, 2010. The Hurst exponent increase of 0.083 during the crisis, the largest in our dataset, is consistent with feedback intensification during the deleveraging cascade.
- The Flash Crash of May 6, 2010, is documented in the SEC/CFTC Joint Report, “Findings Regarding the Market Events of May 6, 2010.” The tail exponent increase of 0.76 after the Flash Crash (the largest tail-thinning shift in our dataset) suggests that Limit Up-Limit Down circuit breakers successfully truncated the most extreme feedback cascades while leaving the underlying memory-generating mechanism intact.
Methodology
- Rolling window specification. We used trailing 5-year (approximately 1,260 trading day) windows evaluated at quarterly intervals. At each quarter-end, we identified all markets with at least 1,000 trading days of history and computed DFA exponents on absolute returns. Cross-market statistics (median, 25th and 75th percentiles) were computed at each step. The number of contributing markets grows from 36 in the late 1980s to 68 from the mid-2000s onward.
- Detrended Fluctuation Analysis (DFA). For a series of length N: (1) compute the cumulative sum of mean-centred values (the “profile”). (2) Divide into non-overlapping windows of size n. (3) In each window, fit a linear trend and compute the root-mean-square residual. (4) Average across windows to get the fluctuation function F(n). (5) Repeat for increasing n from 10 to N/4, spaced geometrically by factor 1.5. (6) Estimate the scaling exponent as the slope of log F(n) versus log n via OLS. This exponent is equivalent to the Hurst exponent H: H > 0.5 indicates long-range persistence, H = 0.5 indicates no memory. DFA is preferred for rolling windows because it is less sensitive to sample size than R/S and is robust to polynomial trends in the data.
- Hill estimator. For a sample of absolute returns sorted in descending order, the Hill estimator with k = sqrt(N) order statistics: alpha = k / sum(ln(x_i / x_{k+1})) for i = 1 to k. Lower alpha indicates fatter tails.
Figures
- Figure 7.1: Upper panel: rolling 5-year DFA Hurst exponents across all available markets, 1987-2024. Bold line: cross-market median. Shaded band: 25th to 75th percentile. Red dashed line: H = 0.5 (random walk). Vertical dashed lines mark five natural experiments. Lower panel: number of markets contributing at each quarterly evaluation date.
- Figure 7.2: Before-and-after comparison of DFA Hurst exponents (left panel) and Hill tail exponents alpha (right panel) around five natural experiments. Blue bars: window before event. Red bars: window after event. Change magnitudes and directions annotated. All values computed as cross-market averages.
This research series is drawn from The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.