How market volatility responds to a preceding price trend: a lag-resolved measurement in cryptocurrencies and equities
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.
Ruth: I'm Ruth.
Warren: And I'm Warren, here with Iris.
Ruth: We're looking at the paper: "How market volatility responds to a preceding price trend".
Warren: Here's the gist of the paper. We quantify the response of market volatility to a preceding price trend as a function of lag at minute resolution, investigating its asymmetry—the excess volatility following a fall relative to a rise of the same magnitude.
Ruth: The sample comprises trades in three major cryptocurrencies on one exchange and three large US stocks and the S&P 500 ETF between 2020 and 2026.
Warren: For cryptocurrencies, the asymmetry is positive, aligning with the leverage effect; it is maximal at a lag of 20 minutes and diminishes at negative lags where volatility follows the trend.
Ruth: For Bitcoin and Ether, this asymmetry decays to half by a lag of approximately one and a half hours.
Title and authors: Ruth: To keep going with the paper "How market volatility responds to a preceding price trend: a lag-resolved measurement in cryptocurrencies and equities," we need to talk about what the authors are actually measuring. It’s not just looking at raw volatility, but specifically how it relates to the direction of a previous price move, measured at minute resolution.
Warren: Right, so they are quantifying this response as a function of lag in minutes. They aren't just looking at correlation; they are setting up this specific setup where they look at the excess volatility following a fall versus a rise of the same magnitude.
Iris: The sample size is quite broad for this kind of work, covering three major cryptocurrencies on one exchange and three large US stocks plus the S andP five hundred ETF across two thousand twenty to two thousand twenty-six <ref:2611.04533#pg1>. That gives them a good scope to test their findings.
Ruth: And they found that for the cryptocurrencies, this asymmetry is positive, which aligns with what you expect from the leverage effect in that space. It follows the trend; it's greatest at the first measurable lag of twenty minutes but gets weaker at negative lags where the trend itself dictates the volatility.
Warren: And they also found that for Bitcoin and Ether specifically, this asymmetry decays to one half by a lag of roughly one and a half hours. That’s a specific time scale for how long that effect lasts after an impulse moves the price up or down.
Iris: They then do some decomposition under time reversal, which splits this asymmetry into two parts: T-even and T-odd. The interesting part is that only the T-odd component exhibits an arrow of time; it’s the one that actually carries direction in time.
Ruth: That means they can isolate whether volatility responds before or after a price move, which is important for telling the difference between news impact and actual volatility feedback. For Bitcoin and Ether, this T-odd part was detected at the first lag of twenty minutes.
Warren: And at ten and twenty minutes, they see that the T-odd part is four to five times bigger than the T-even part, which points to a purely one-sided response over time in those specific windows.
The paper's summary: Ruth: Moving on to the core summary of "How market volatility responds to a preceding price trend: a lag-resolved measurement in cryptocurrencies and equities," they are essentially describing how this lag-resolved measurement works and what the results tell us about time directionality.
Warren: They start by stating that they quantify the response of market volatility to a preceding price trend as a function of lag at minute resolution. The core concept is looking at the asymmetry—the excess volatility following a fall relative to a rise of the same magnitude.
Iris: For cryptocurrencies, they find this asymmetry is positive, matching what you expect from the leverage effect in that space. It follows the trend; it's greatest at the first measurable lag of twenty minutes but gets weak at negative lags where the trend itself dictates the volatility.
Ruth: And for Bitcoin and Ether specifically, they found that this asymmetry decays to one half by a lag of roughly one and a half hours. That’s a specific time scale for how long that effect lasts after an impulse moves the price up or down.
Warren: Then they do decomposition under time reversal, which splits this asymmetry into two parts: T-even and T-odd. The interesting part is that only the T-odd component exhibits an arrow of time; it’s the one that actually carries direction in time.
Iris: This means by decomposing it this way, they can see if there's a directional effect or just some statistical artifact at play in different markets.
The paper's improvements: Ruth: Now let’s talk about the suggested improvements the authors made to this measurement method, because they aren't just presenting a finding, they’re suggesting ways to make it better. They think they could use scale-free amplitude measurements, like Delta s, which are invariant to rescaling epoch volatility and allow them to compare different assets and years fairly.
Warren: That sounds like a way to get a more robust comparison when you're looking at things across different markets where the baseline volatility levels are so wildly different. It lets you normalize the data better before you even start comparing.
Iris: They also suggest they can calculate specific time scales for how fast this response decays, using models like b plus a e minus l over tau, which gives them a quantitative measure of how long market volatility sticks around after an impulse moves the price.
Ruth: So, summarizing these improvements, it's about making the measurement more robust by using scale-free amplitude measurements and better models for decay time scales.
Warren: And that whole simulation part with shuffled data is key because it confirms that this isn't some random noise they just happened to find.
Iris: One limitation they mention is that the method, as presented, doesn't fully resolve the issue of coupling without temporal direction, which means there’s still a piece of ambiguity left in what the statistic itself generates.
Conclusion: Ruth: So to wrap up this paper "How market volatility responds to a preceding price trend: a lag-resolved measurement in cryptocurrencies and equities," the main point is that volatility reacts differently depending on the asset class and how you look at time directionality.
Warren: For cryptocurrencies, there's a clear one-sided temporal pattern at short lags, and for equities, it seems more about the sign of a move over inner trends.
Ruth: It’s a step toward better understanding how different asset classes process market information over short timescales, giving us concrete numbers instead of just qualitative observations.
Warren: The paper shows that the response dynamics are not uniform across markets, which is something we need to keep watching as markets evolve.
Iris: One final thought: the method doesn't fully resolve the issue of coupling without temporal direction, meaning there’s still some ambiguity left in what the statistic itself generates.
Ruth: Exactly. That’s where future work needs to focus, trying to build a measurement that can cleanly separate those two effects.
Warren: Good point. For now, this paper gives us a solid framework for analyzing these response dynamics across different asset classes with minute resolution data between two thousand twenty and two thousand twenty six.
Gülşah Bıçkı, Taylan Akdoıan
Ozyeğin University
q-fin.ST
Submitted: 2026-11-03
Updated: 2026-11-03
Comments: 37 pages, 8 figures, to be submitted to "Physica A: Statistical Mechanics and its Applications"
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 83/100
The gist: We quantify the response of market volatility to a preceding price trend as a function of lag at minute resolution…
Key concepts
- Lag-resolved measurement
- This involves measuring how market volatility changes after a price move at specific time intervals (lags), down to minute resolution. This allows researchers to see exactly how quickly and in what direction volatility responds following a trend, distinguishing between immediate news impact and slower feedback loops.
- Asymmetry
- This refers to the difference in volatility response when a price moves up versus when it moves down by the same amount. The paper investigates 'excess volatility' after a fall compared to a rise, which helps identify directional market effects like leverage.
- Time Reversal Decomposition
- This technique separates the observed asymmetry into two parts: T-even and T-odd components. The T-odd component is crucial because it captures the temporal directionality—whether volatility responds before or after the price trend, which helps isolate true market feedback from simple correlation.
- T-odd/T-even Components
- These are mathematical parts of a signal that reveal the temporal nature of a response. The T-odd component is what exhibits an 'arrow of time,' meaning volatility reacts in a specific sequence relative to the price change, confirming whether the market is reacting ahead of or lagging behind the trend.
Summary
We quantify the response of market volatility to a preceding price trend as a function of lag at minute resolution, investigating its asymmetry—the excess volatility following a fall relative to a rise of the same magnitude. p. 0 The sample comprises trades in three major cryptocurrencies on one exchange and three large US stocks and the S&P 500 ETF between 2020 and 2026. p. 0 For cryptocurrencies, the asymmetry is positive, aligning with the leverage effect; it is maximal at a lag of 20 minutes and diminishes at negative lags where volatility follows the trend. p. 0 For Bitcoin and Ether, this asymmetry decays to half by a lag of approximately one and a half hours. p. 0 Decomposition under time reversal separates this asymmetry into T-even and T-odd components, only the latter exhibiting an arrow of time; for Bitcoin and Ether, the T-odd part is detected at the first lag, reaching 0. p. 04 to 0.9 times the T-even part at 10 and 20 minutes. p. 0 At a lag of 20 minutes, this ratio indicates a purely one-sided response in time. p. 2 Conversely, no detectable T-odd part is found in the straight-line fit over two and a half trend scales for equities. p. 0 The index fund's asymmetry depends on the fit: across the full range, it is large at both positive and negative lags but of opposite signs, forming a mirror; over inner trends, however, the positive side is four times larger than the negative side. This suggests the fund reacts more to the sign of a move than its magnitude. p. 0 The return-variance correlation displays an arrow of time. p. 0 Simulated and shuffled data without coupling yield zero with errors matching correct size and recover the time scale of a one-sided response embedded in simulated markets, suggesting this result is a market response function where price response to a trade is defined [43]. p. 0 The impulse is the trend of the log price over 20 minutes, and the response is the log variance over a subsequent 20-minute window. p. 2 The quantity studied is this asymmetry, derived from fits of log variance against trend on both sides of zero. p. 2 In Bitcoin and Ether, the T-odd part is comparable to the T-even part, with a T-odd component reaching 4 sigma to 5 sigma from zero at that window, consistent with a wholly one-sided temporal response. p. 0
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI can quantify market volatility response to preceding price trends by using a lag-resolved measurement at minute resolution, distinguishing between news impact curves and coupling without temporal direction. This allows the system to identify whether asymmetry follows the trend or if it is due to a coupling without temporal direction, which is crucial for interpreting effects like the leverage effect.
-
Improved AI can perform time reversal decomposition of market asymmetry into a T-even part and a T-odd part, allowing it to isolate the component that
carries an arrow of time.
This capability enables the system to determine if volatility responds before or after a price move, differentiating between news impact and volatility feedback. -
Improved AI can utilize scale-free amplitude measurements, like Delta s, which are invariant to rescaling of epoch volatility and enable comparison across assets and years. This allows the system to compare the relative magnitude of leverage effects across different cryptocurrencies and equities regardless of their respective volatility levels.
-
Improved AI can calculate specific time scales for response decay, such as the half-decay lag, based on fitted models like b plus a e minus l over tau. This provides a quantitative measure of how long market volatility persists after a trend impulse, allowing for the identification of different response dynamics (e.g., one and a half hours for Bitcoin and Ether).
-
Improved AI can perform rigorous null hypothesis testing using data without coupling, such as sign shuffles and simulated markets. This ensures that any detected asymmetry is not merely an artifact of the statistic itself or a coupling without temporal direction, providing confidence in the interpretation of results.
Sources
Related papers
- PanelGP: A Probabilistic Factor Model for Asset-Return Panels
- Physics-Informed Singular-Value Learning for Cross-Covariances Forecasting in Financial Markets
- A general technique for explicit quantile representations
- A Controlled Comparison of Deep Learning Architectures for Multi-Horizon Financial Forecasting: Evidence from 918 Experiments
- Who Are a Firm's Economic Peers? Mapping a Latent Information Field with Language Models
- Neural Realized Measures: Forecast-Supervised Intraday Aggregation for Volatility Forecasting