6.6 Toxic Order Flow & VPIN

Basic Prerequisites: Adverse Selection & the Glosten–Milgrom Model (the information-cost foundation this builds on) and Probability & Measure Theory (Poisson processes, mixtures). (these are the folder-level prerequisites for pages 02–06; page 01 states its own, smaller, entry requirements)

1. Intuition & Practical Objective

Some order flow is toxic: it arrives in one-sided waves that are informed - the counterparty knows the efficient price is about to move against you. When a market maker fills such flow, the fill itself is a loss signal. The extreme case is the May 6, 2010 flash crash: liquidity vanished because market makers detected an overwhelming, one-sided informed (mostly institutional selling) wave and pulled their quotes - precisely the toxicity-triggered liquidity evaporation that VPIN is designed to anticipate.

This folder is the toxic-order-flow-and-VPIN topic-folder for Pillar 6 (Market Making & Microstructure). It is a hub: it (a) gives you the fast formula-and-model lookup below, and (b) routes you to six sub-pages that walk from first-principles intuition through PIN (probability of informed trading), the EKOP Poisson model, and VPIN (volume-synchronized probability of informed trading) with bulk-volume classification. Primary verified sources: Hasbrouck Empirical Market Microstructure (Ch 6, the PIN model) and the primary papers - Easley, Kiefer & O’Hara (1997, PIN); Easley, López de Prado & O’Hara (2012, VPIN) and (2011, flash crash); Lee & Ready (1991, trade signing); Andersen & Bondarenko (2014, the mandatory critique).

The one-sentence essence. “Measure the probability that the arriving order flow is informed - either as PIN, the fraction of flow from informed traders in a Poisson model, or as VPIN, the rolling normalized order-flow imbalance in volume time; when either spikes, market makers’ quotes are being picked off, and the correct response is to widen or step aside.”

2. Mathematical Ground Truth & Derivations

Notation (Hasbrouck Ch 6; EKOP 1997): $α$ = probability an information event occurs on a given day; $μ$ = Poisson arrival rate of informed traders (on an event day); $ϵ$ = Poisson arrival rate of uninformed traders on each side.

PIN - the probability of informed trading (EKOP 1997; Hasbrouck eq. 6.4)

Over a sample, informed orders arrive at total rate $αμ$ ; uninformed at rate $2 ϵ$ (one Poisson stream of intensity $ϵ$ per side). PIN is the informed fraction of the flow:

PIN = \frac{αμ}{αμ + 2 ϵ} \in [0, 1]

Only the product $αμ$ is identified - $α$ and $μ$ individually are imprecise, PIN is stable/estimable (Hasbrouck Ch 6).
$PIN = 0$ with no informed traders; $PIN \to 1$ as informed flow dominates.
PIN is the trade-count measure (buys/sells per day, Poisson mixture, MLE); VPIN is its volume-synchronized generalization.

VPIN - volume-synchronized probability of informed trading (ELO 2012)

Divide the tape into equal-volume buckets of $V$ shares each; in bucket $τ$ , classify buy-volume $V_{τ}^{B}$ and sell-volume $V_{τ}^{S}$ (so $V = V_{τ}^{B} + V_{τ}^{S}$ ). Over a rolling window of $n$ buckets:

VPIN = \frac{\sum _{τ = 1}^{n} V _{τ}^{S} - V _{τ}^{B}}{n V} \approx \frac{αμ}{αμ + 2 ϵ}

Because $E [∣ V^{S} - V^{B} ∣] \approx αμ$ and $E [V^{B} + V^{S}] = αμ + 2 ϵ$ , VPIN estimates exactly the PIN ratio - in volume time (ELO eq. 9).
Balanced noise flow ( $V^{B} \approx V^{S}$ ) → VPIN → 0; one-sided panic (all sells) → VPIN → 1.
The ELO default: $V = \frac{1}{50}$ of average daily volume and $n = 50$ (a rolling “daily” VPIN).

Bulk-volume classification (ELO 2012 Appendix A) and Lee–Ready (1991)

Quantity	Formula
Tick-rule signing (Lee–Ready)	$q_{t} = + 1$ if $P_{t} > M_{t}$ , $- 1$ if $P_{t} < M_{t}$ , else $sign (P_{t} - P_{t - 1})$
Bulk buy-volume in bucket $τ$	$V_{τ}^{B} = \sum_{i} Z (\frac{Δ P _{i}}{σ _{Δ P}})$ , $Z = Φ$ standard-normal CDF
Bulk sell-volume	$V_{τ}^{S} = V - V_{τ}^{B}$

Fast lookup (job #1): PIN and VPIN are the same object - the informed fraction of flow - estimated on two clocks (trade-count vs volume). The formulas above are transcribed from Hasbrouck Ch 6 and ELO 2012 eq. 9 + Appendix A (verified from the primary PDF); the numbers below were re-executed and reproduced exactly.

3. Computational Implementation - PIN + VPIN in one pass (stdlib only)

4. Failure Modes & First-Principles Breakdowns

Hub signposts - the folder’s failure-mode analysis lives in 05 · Failure Modes & Practice. In one line each:

Classification is not truth. Tick-rule and bulk classification infer buy/sell from prices, so any price move (volatility, a public trend) can be misread as informed one-sided flow (04 · VPIN).
VPIN is not portable across settings. Its absolute level depends on the arbitrary bucket size $V$ and window $n$ - a fixed “VPIN > 0.3” threshold is meaningless across instruments (05).
High VPIN ≠ imminent crash. Andersen & Bondarenko (2014) show VPIN is dominated by volume/volatility and did not robustly predict the flash crash; it is a flow-imbalance gauge, not a crash oracle (05).

5. References

Easley, D., Kiefer, N. & O’Hara, M. (1997), The information content of the trading process, J. Empirical Finance 4, 159–186. PIN, the Poisson-mixture informed-fraction. Primary PDF: 37_Easley_1997... in corpus.
Easley, D., López de Prado, M. & O’Hara, M. (2012), Flow toxicity and liquidity in a high-frequency world, Review of Financial Studies 25(5), 1457–1493. The formal VPIN paper: volume-synchronized imbalance, bulk-volume classification, eq. 9 and Appendix A. Primary PDF: 33Easley2012_flow_toxicity_and_liquidity_in.pdf.
Easley, D., López de Prado, M. & O’Hara, M. (2011), The microstructure of the “flash crash”: flow toxicity, liquidity crashes, and the probability of informed trading, J. Portfolio Management 37(2), 118–128. The applied claim that VPIN spiked before May 6, 2010. Primary PDF: 34_Easley_2011....
Lee, C. M. C. & Ready, M. J. (1991), Inferring trade direction from intraday data, J. Finance 46(2), 733–746. The signing algorithm every flow metric rests on. Primary PDF: 36_Lee_1991....
Andersen, T. G. & Bondarenko, O. (2014), VPIN and the flash crash, J. Financial Markets 17, 1–46. The essential critique
Easley, D., Hvidkjaer, S. & O’Hara, M. (2002), Is information risk a determinant of asset returns?, J. Finance 57(5), 2185–2221. The asset-pricing use of PIN. Primary PDF: 38_Easley_2002....
Hasbrouck, J. (2007), Empirical Market Microstructure, OUP

6. Connected Graph Bridges

Foundational base: Adverse Selection & Glosten–Milgrom · Probability & Measure Theory · Econometrics & Time Series
Sibling topic (in-pillar): Spread Decomposition & the Roll Model (measuring the adverse-selection/impact component empirically) · Inventory Management & Quote Skewing (why makers widen) · Avellaneda–Stoikov Optimal Quoting (reservation pricing over toxic flow)
Sub-pages (in-folder): 01 From Zero · 02 Probability of Informed Trading · 03 The EKOP Model · 04 VPIN · 05 Failure Modes · 06 Advanced Extensions

Beginner: start at 01 · Practitioner: start at 05

Kwant Atlas

Explorer

6.6 Toxic Order Flow & VPIN

1. Intuition & Practical Objective

2. Mathematical Ground Truth & Derivations

PIN - the probability of informed trading (EKOP 1997; Hasbrouck eq. 6.4)

VPIN - volume-synchronized probability of informed trading (ELO 2012)

Bulk-volume classification (ELO 2012 Appendix A) and Lee–Ready (1991)

3. Computational Implementation - PIN + VPIN in one pass (stdlib only)

4. Failure Modes & First-Principles Breakdowns

5. References

6. Connected Graph Bridges

6.6.1 Toxic Order Flow from Zero

6.6.2 Probability of Informed Trading (PIN)

6.6.3 The EKOP Model

6.6.4 VPIN

6.6.5 Failure Modes & Real-World Practice

6.6.6 Advanced Extensions