Hasbrouck (1991) Trade VAR & Price Impact Lab

Model Joel Hasbrouck's (1991) structural Vector Autoregression (VAR) framework to isolate the permanent information content of trades from transitory microstructure frictions.

MARKET REGIMES:

Permanent Impact I(∞)

$0.082

8.2¢ per trade shock

Trade Information Share (R_w²)

Immediate Impact I(0)

$0.080

b0 = $0.080 immediate

Transitory Impact (Friction)

-$0.002

Momentum reinforcement

VAR Parameters

VAR Lag Order (K): 5

Number of autoregressive lags in the quote and trade equations.

Order Flow Persistence ($\rho_x$): 0.35

Autocorrelation of trade signs (positive indicates institutional order splitting).

Immediate Quote Impact ($b_0$): $0.08

Immediate contemporaneous midquote revision upon trade execution.

Private Trade Innovation ($\sigma_{v2}$): 0.25

Volatility of unexpected informed order flow shocks.

Public News Shock ($\sigma_{v1}$): 0.40

Volatility of trade-uncorrelated public news announcements.

Sample Trades ($T$): 500

Total consecutive transactions in the econometric sample.

Microstructure Analytics Visualizer

Hasbrouck VMA Representation: $\begin{pmatrix} r_t \\ x_t \end{pmatrix} = \begin{pmatrix} \alpha(L) & \beta(L) \\ \gamma(L) & \delta(L) \end{pmatrix} \begin{pmatrix} v_{1,t} \\ v_{2,t} \end{pmatrix}$. The cumulative impulse response $I(\tau) = \sum_{s=0}^\tau \beta_s$ traces the permanent quote revision caused by an unexpected trade $v_{2,t} = 1$.

Sensitivity: VAR Lags ($K$) vs Persistence ($\rho_x$) on Permanent Impact $I(\infty)$ ($)

Evaluates how lag length and order persistence affect the estimated permanent informational impact per unexpected trade.

Sensitivity: Private Vol ($\sigma_{v2}$) vs Public Vol ($\sigma_{v1}$) on Information Share $R_w^2$ (%)

Quantifies the proportion of fundamental quote revision variance driven by private trades versus public news announcements.

Mathematical Foundations: Hasbrouck's (1991) VAR Price Impact Model

In modern electronic markets, trade executions carry private information regarding asset valuation while generating temporary dealer friction. In his landmark 1991 Journal of Finance paper, Joel Hasbrouck formalized a structural econometric method to disentangle permanent private information from transient liquidity noise.

1. Bivariate VAR Specification

Let $r_t = q_t - q_{t-1}$ be the midquote revision following trade $t$, and $x_t \in \{+1, -1\}$ be the signed order flow. The system is specified with $K$ lags:

$$r_t = b_0 x_t + \sum_{k=1}^K a_k r_{t-k} + \sum_{k=1}^K b_k x_{t-k} + v_{1,t}$$ $$x_t = \sum_{k=1}^K c_k r_{t-k} + \sum_{k=1}^K d_k x_{t-k} + v_{2,t}$$

The contemporaneous trade term $b_0 x_t$ enforces recursive causal ordering: orders arrive first and trigger immediate quote adjustments by liquidity providers.

2. Impulse Response & Permanent Impact

Inverting the VAR into its Vector Moving Average (VMA) form isolates the quote response to an unpredicted trade shock ($v_{2,t} = 1$):

$$I(\tau) = \sum_{s=0}^\tau \beta_s, \quad I(\infty) = \lim_{\tau \to \infty} I(\tau) = \sum_{s=0}^\infty \beta_s$$ $$I_{\text{trans}}(\tau) = I(\tau) - I(\infty)$$

$I(\infty)$ measures the permanent informational revision per trade innovation, while $I_{\text{trans}}(\tau)$ captures transient inventory rebalancing and quote bounce.

3. Beveridge-Nelson Variance Decomposition ($R_w^2$)

Decomposing quotes into a pure random-walk fundamental efficient price $w_t$ and stationary microstructure noise $s_t$ ($q_t = w_t + s_t$), the variance of permanent increments $\Delta w_t$ is:

$$\sigma_w^2 = \sigma_{w,x}^2 + \sigma_{w,v}^2 = \left(\sum_{s=0}^\infty \beta_s\right)^2 \sigma_{v2}^2 + \left(\sum_{s=0}^\infty \alpha_s\right)^2 \sigma_{v1}^2$$ $$R_w^2 = \frac{\sigma_{w,x}^2}{\sigma_w^2} = \frac{\left(\sum \beta_s\right)^2 \sigma_{v2}^2}{\left(\sum \beta_s\right)^2 \sigma_{v2}^2 + \left(\sum \alpha_s\right)^2 \sigma_{v1}^2}$$

$R_w^2$ (Trade Information Share) establishes the fraction of fundamental market price discovery generated by private trading activity rather than public news releases.

Hasbrouck (1991) Market Microstructure Mastery Quiz

Related Market Microstructure & Quantitative Trading Simulators

Kyle's Lambda & Microstructure Lab

Model Pete Kyle's (1985) insider trading equilibrium, linear price impact lambda, market depth, and private info variance.

Glosten-Milgrom Microstructure Lab

Simulate Bayesian quote updating, endogenous adverse selection spreads, and sequential informational price discovery.

EKOP PIN Microstructure Lab

Estimate the Easley-Kiefer-O'Hara-Paperman (1996) Probability of Informed Trading (PIN) via numerical maximum likelihood estimation.

VPIN & Order Flow Toxicity Lab

Model Easley, López de Prado, and O'Hara's (2012) Volume-Synchronized Probability of Toxicity and flash crash cascade dynamics.

Hasbrouck Information Share Lab

Decompose multi-venue price discovery and cointegration leadership using Joel Hasbrouck's (1995) VECM and Cholesky bounds.

Roll Effective Spread Lab

Estimate bid-ask bounce friction and effective spreads from serial covariance in transaction price returns (Roll 1984).

Frequently Asked Questions

Joel Hasbrouck's (1991) model is a seminal empirical market microstructure framework published in the Journal of Finance. It uses a bivariate Vector Autoregression (VAR) of midquote price revisions (r_t) and signed trade order flow (x_t) to quantify the informational content of trades, isolating permanent price impact from transitory microstructure noise.

By inverting the estimated VAR system into a Vector Moving Average (VMA) impulse response function, the model traces the multi-period quote response to an unexpected trade shock. As the lag horizon goes to infinity, the cumulative response converges to the permanent price impact I(infinity), representing private information. Any temporary quote overshoot that decays back to the permanent level constitutes transitory friction such as dealer inventory holding costs or order processing bounce.

Hasbrouck specifies a triangular causal ordering where incoming trade orders hit existing displayed quotes first, prompting market makers to immediately update their posted bid and ask quotes. This causal protocol allows trade sign x_t to affect quote revisions r_t contemporaneously via parameter b0, whereas quote revisions do not affect trade orders in the same instant.

The Beveridge-Nelson decomposition models quote prices as a random walk fundamental value plus stationary microstructure noise. The trade information share (Rw^2) measures the proportion of total fundamental quote variance that is explained by unpredicted private trade flow innovations relative to exogenous public news announcements, typically ranging between 15% and 55% in liquid equity markets.

While Pete Kyle's (1985) lambda measures static linear price impact and Richard Roll's (1984) model estimates effective spread purely from first-order serial return covariance, Hasbrouck's (1991) VAR unifies both by modeling the full dynamic multi-lag propagation of order flow persistence, transitory inventory mean-reversion, and permanent informational price discovery simultaneously.