Volume-To-Volatility Ratio (VVR)

• VVR is an umbrella term describing diverse quantitative approaches that link trading volume with price variability, emphasizing weak raw correlations paired with pronounced extreme-volatility signals.
• Researchers utilize conditional distributions, extreme value statistics, and entropy-weighted estimators to integrate volume and volatility for improved risk forecasts and execution decisions.
• Empirical findings show that higher trading volume often correlates with extreme volatility, while structural decompositions reveal that volume’s impact depends on liquidity and market dynamics.

Searching arXiv for the cited papers and related Volume-To-Volatility Ratio literature. arXiv search query: "Volume-To-Volatility Ratio volume volatility trading volume realized volatility order slicing LMV" Volume-To-Volatility Ratio (VVR) denotes, in the broadest research sense, a class of constructions that compare trading activity with price variability. In the arXiv literature considered here, however, VVR does not appear as a standardized named scalar such as “volume divided by volatility.” Instead, the volume–volatility nexus is operationalized through conditional distributions, envelope statistics for extreme volatility, execution heuristics that jointly forecast range and volume, entropy-based volatility estimators weighted by volume shares, spread formulas in which volume and volatility enter jointly, and structural decompositions of volume variation into volatility-driven and liquidity-driven components (Zheng et al., 2014, Chattopadhyay et al., 2024, Vinte et al., 2022, Sarkissian, 2016, Bucci et al., 6 Jun 2026). In that sense, VVR is best treated as an umbrella label for related but non-equivalent quantitative objects.

1. Conceptual status in the literature

A consistent feature of the cited corpus is the absence of an explicit, universally adopted VVR formula. “Predicting market instability: New dynamics between volume and volatility” studies the conditional law $P(g\mid v)$, the scaling collapse of those conditionals, the local maximum volatility (LMV), and a joint conditional probability for forecasting next-day extreme volatility, but it does not define a literal volume-to-volatility ratio (Zheng et al., 2014). “Volatility-Volume Order Slicing via Statistical Analysis” similarly does not introduce a named VVR; instead, it combines a volatility proxy based on price range with forecast trading volume to determine execution decisions (Chattopadhyay et al., 2024). “A Volatility Estimator of Stock Market Indices Based on the Intrinsic Entropy Model” comes closest to a ratio-based object through the volume share

$p_i=\frac{q_i}{Q},$

yet this is explicitly a volume-to-total-volume share used to weight volatility estimation, not a direct ratio of volume to volatility (Vinte et al., 2022).

The same pattern holds in market microstructure and structural time-series work. “Spread, volatility, and volume relationship in financial markets and market making profit optimization” does not define VVR as a named measure, but it derives spread equations in which volatility and volume jointly determine liquidity costs (Sarkissian, 2016). “A Structural Matrix Autoregressive Model for the Joint Dynamics of Volume, Volatility, and Returns” does not define VVR either, but it estimates contemporaneous and dynamic transmission between realized volatility and trading volume and decomposes volume variation into informative and liquidity components (Bucci et al., 6 Jun 2026). By contrast, “IVE: Enhanced Probabilistic Forecasting of Intraday Volume Ratio with Transformers” is explicitly about intraday volume ratio for VWAP execution rather than any volatility-linked ratio; it is therefore adjacent to, but not itself, a VVR study (Lee et al., 2024).

Paper	Closest VVR-like object	Function
(Zheng et al., 2014)	$P(g\mid v)$, LMV, joint conditional probability	Extreme-volatility prediction
(Chattopadhyay et al., 2024)	Forecast range + forecast volume	Order slicing
(Vinte et al., 2022)	$p_i=q_i/Q$	Volume-weighted volatility estimation
(Sarkissian, 2016)	$\Delta=\lambda s\sigma\sqrt{n/V}$	Spread–liquidity relation
(Bucci et al., 6 Jun 2026)	$c_{21}$, FEVD informative share	Structural volume–volatility decomposition
(Lee et al., 2024)	Intraday volume ratio	VWAP support

This suggests that VVR is not a single settled metric but a family resemblance across models that quantify how trading activity conditions, weights, amplifies, or transmits volatility.

2. Distributional and predictive formulations

The most direct empirical treatment of a VVR-like relation appears in the study of volume-conditional volatility. For 30 DJIA stocks over 1990–2007, normalized volatility and normalized logarithmic volume are defined as

$$g_i(t)=\left|\frac{R_i(t)-\langle R_i(t)\rangle}{\sigma_R}\right|, \qquad v_i(t)=\frac{\tilde Q_i(t)-\langle \tilde Q_i(t)\rangle}{\sigma_{\tilde Q}},$$

and the conditional density is found to be well described by a power law with exponential cutoff,

$P(g\mid v)\sim g^{-\xi}e^{-\varsigma g},$

with linearly volume-dependent parameters

$\xi=\alpha v+a,\qquad \varsigma=\beta v+b.$

Maximum likelihood estimates reported in the supplied summary are approximately $\alpha=0.4$, $p_i=\frac{q_i}{Q},$0, $p_i=\frac{q_i}{Q},$1, and $p_i=\frac{q_i}{Q},$2 (Zheng et al., 2014).

The same paper introduces the local maximum volatility statistic

$p_i=\frac{q_i}{Q},$3

which extracts the upper envelope of volatility in each volume bin. Ordinary same-day volume–volatility correlation is described as weak, but the correlation between logarithmic volume and LMV is reported as very strong; for Boeing, the supplied example gives roughly $p_i=\frac{q_i}{Q},$4 for volume–LMV, versus about $p_i=\frac{q_i}{Q},$5 in the figure discussion for same-day volume–volatility, while the general conclusion emphasizes weak overall correlation in raw levels (Zheng et al., 2014). The central asymmetry is explicit: low volatility can occur across nearly all volume levels, but high volatility occurs much more often when volume is high.

Predictive conditioning reinforces that asymmetry. The joint conditional probability

$p_i=\frac{q_i}{Q},$6

is used to study whether today’s volume and volatility quintiles improve prediction of next-day extremes. The top-volume/top-volatility quintile pair is reported to raise the probability of a top-1% next-day volatility event to about three times the unconditional probability, whereas the bottom-bottom quintile reduces it to about half (Zheng et al., 2014). A plausible implication is that any useful VVR-like summary should be tail-sensitive rather than based only on average volatility.

3. Execution-oriented formulations

In optimal execution, the closest analog to VVR is not a static ratio but a decision rule that balances predicted liquidity against predicted risk. “Volatility-Volume Order Slicing via Statistical Analysis” uses two estimated quantities: a volatility proxy given by the intraday high-low price range and a liquidity or activity proxy given by trading volume. At 5-minute frequency, range is defined as

$p_i=\frac{q_i}{Q},$7

with additional decomposition into $p_i=\frac{q_i}{Q},$8 and $p_i=\frac{q_i}{Q},$9 (Chattopadhyay et al., 2024).

The forecasting architecture is a two-stage EWMA plus Metropolis-Hastings MCMC system. Both range and volume are modeled with log-normal distributions estimated by MLE. Range sampling conditioned on volume uses the acceptance rule

$P(g\mid v)$0

while EWMA smoothing is given by

$P(g\mid v)$1

To align daily EWMA and 5-minute range scales, the paper introduces the adjusted scaling factor

$P(g\mid v)$2

with $P(g\mid v)$3 days (Chattopadhyay et al., 2024).

The practical role of these objects is execution control. The paper does not provide a literal formula such as

$P(g\mid v)$4

but its logic is explicitly VVR-like: predicted range acts as expected volatility, predicted volume acts as expected market capacity, and order slicing is adapted to both. Reported forecast accuracy is Average MAPE for Volume: 26.23% and Average MAPE for Range: 35.88% (Chattopadhyay et al., 2024). The contribution is therefore heuristic and forecast-driven rather than a threshold rule based on a single scalar VVR.

A separate execution-related strand appears in intraday volume-ratio forecasting for VWAP. The IVE model predicts intraday volume ratio at one-minute scale with a Transformer encoder-decoder and a Student’s $P(g\mid v)$5-distribution head, outputting the mean and standard deviation of volume ratios. It is not a volatility-ratio model, but it shows how ratio-like volume targets can be embedded in live execution systems. The Korean-market live trading experiment reports average execution performance 4.82 bp better than Market VWAP, standard deviation 34.59 bp, and beat ratio 59% (Lee et al., 2024). This suggests that, in practice, VVR-adjacent execution models may separate volume-ratio forecasting from volatility handling rather than collapsing both into a single metric.

4. Volume-weighted volatility estimation

An alternative route to VVR-like thinking treats volume not as a denominator or comparator, but as a weighting scheme inside the volatility estimator itself. In the intrinsic entropy model, daily volume shares over an $P(g\mid v)$6-day window are defined as

$P(g\mid v)$7

where $P(g\mid v)$8 is daily traded volume and $P(g\mid v)$9 is total volume over the estimation window. The model explicitly interprets $p_i=q_i/Q$0 as “entropic probability” or “market credence” assigned to the corresponding price level (Vinte et al., 2022).

The resulting estimator follows a Yang–Zhang-style decomposition: $p_i=q_i/Q$1 The component expressions supplied in the summary include

$p_i=q_i/Q$2

along with OHLC-based price-change terms modulated by $p_i=q_i/Q$3 or $p_i=q_i/Q$4 (Vinte et al., 2022). The essential point is that a price move supported by a larger fraction of total market activity contributes more heavily to the volatility estimate.

Empirically, the paper reports that intrinsic entropy estimates are on a much lower numerical scale than standard estimators, while exhibiting a much higher coefficient of variation. For the S&P 500 over 5-day windows, the supplied mean estimates are approximately $p_i=q_i/Q$5 for close-to-close, $p_i=q_i/Q$6 for Parkinson, $p_i=q_i/Q$7 for Garman-Klass, $p_i=q_i/Q$8 for Rogers-Satchell, $p_i=q_i/Q$9 for Yang-Zhang, and $\Delta=\lambda s\sigma\sqrt{n/V}$0 for intrinsic entropy; the intrinsic entropy coefficient of variation is reported as $\Delta=\lambda s\sigma\sqrt{n/V}$1, versus roughly $\Delta=\lambda s\sigma\sqrt{n/V}$2 to $\Delta=\lambda s\sigma\sqrt{n/V}$3 for the others (Vinte et al., 2022). The paper further concludes that the estimator may be especially useful for short-term trading horizons of roughly 5 to 11 days. In VVR terms, the key object is not a ratio of volume to volatility, but a volume-fraction weighting that changes the construction of volatility itself.

5. Microstructure and structural-system interpretations

In market microstructure, volume and volatility jointly determine spread through a regime-dependent relation. The basic law in “Spread, volatility, and volume relationship in financial markets and market making profit optimization” is

$\Delta=\lambda s\sigma\sqrt{n/V}$4

which yields

$\Delta=\lambda s\sigma\sqrt{n/V}$5

This is the low-volume, liquidity-dominated relation: spread increases with volatility and decreases with volume (Sarkissian, 2016). The same paper then generalizes spread to a composition of liquidity price and impact price,

$\Delta=\lambda s\sigma\sqrt{n/V}$6

with the impact term linked to money flow per unit time. The crucial result is non-monotonicity: for small normalized volume $\Delta=\lambda s\sigma\sqrt{n/V}$7, $\Delta=\lambda s\sigma\sqrt{n/V}$8, while for large $\Delta=\lambda s\sigma\sqrt{n/V}$9, $c_{21}$0. Thus, more volume first tightens spreads and later widens them once impact dominates (Sarkissian, 2016). A plausible implication is that any VVR-like metric that assumes “more volume relative to volatility is always better” is structurally incomplete.

A structural time-series version of the same problem appears in the SMAR model for daily trading volume, realized bipower variation, and returns. The realized volatility proxy is

$c_{21}$1

and the normalized contemporaneous structural matrix is reported as

$c_{21}$2

The supplied interpretation identifies 0.5075 as the contemporaneous effect of a one-standard-deviation realized-volatility shock on trading volume (Bucci et al., 6 Jun 2026). The corresponding forecast error variance decomposition for trading volume labels the volatility-driven share as the informative component and the own-volume share as the liquidity component. At $c_{21}$3, the supplied shares are 20.4% informative, 79.4% liquidity, 0.2% returns, and 0.0% cross-asset spillovers; at $c_{21}$4, they are 7.2%, 35.8%, 0.1%, and 56.9%, respectively (Bucci et al., 6 Jun 2026).

These results shift the VVR discussion from static comparison to decomposition. The nearest analog to VVR in this framework is not $c_{21}$5, but the relative importance of volatility-driven versus liquidity-driven trading activity.

6. Causal evidence, limitations, and common misconceptions

Causal work on Bitcoin futures reinforces that volume–volatility interaction need not imply stability. Using C-ARIMA, the regulated CME futures launch is reported to have increased volatility by +116% on average in the first week in the baseline model and by +139% in the mediation-adjusted volatility model controlling for volume; the corresponding first-week CME effect on volume is about +39%, while CBOE shows a small negative volume effect of about −18% and no significant volatility effect (Menchetti et al., 2021). Lagged volume terms in the volatility model are reported as positive and significant. This is direct evidence of a positive volume–volatility relationship, but not evidence that a high level of volume necessarily improves market quality.

Several misconceptions are therefore excluded by the cited literature. First, VVR is not a standard named metric in this arXiv corpus; treating it as a settled universal formula would be inaccurate (Zheng et al., 2014, Chattopadhyay et al., 2024, Vinte et al., 2022, Bucci et al., 6 Jun 2026). Second, higher volume is not uniformly stabilizing: in microstructure models, additional volume can eventually increase spread through impact (Sarkissian, 2016), and in causal Bitcoin evidence, volatility can rise more strongly than volume after a structural intervention (Menchetti et al., 2021). Third, not every volume ratio is a VVR. The intraday volume ratio studied for VWAP by IVE is a pure execution-volume target rather than a volatility-linked ratio (Lee et al., 2024).

Finally, one bibliographic caution is explicit. For (Duan et al., 2022), the supplied content is a placeholder stating that no PDF is available; accordingly, it provides no usable evidence for any model connecting volume, volatility, and stock pricing, and no support for a VVR formulation (Duan et al., 2022). Within the available arXiv record, VVR is therefore best understood not as a canonical formula, but as a compact label for a heterogeneous set of empirical and structural mechanisms through which trading volume and volatility are related.