Bid-Ask Bounce

Transaction prices mechanically alternate between the bid and ask quotes without any change in the underlying fundamental value. When a trade occurs at the bid, the next trade is more likely to occur at the ask, and vice versa.

• Creates a sawtooth pattern in transaction prices
• Oscillation frequency depends on order flow imbalance
• Magnitude is bounded by the bid-ask spread
• Most pronounced in markets with wide spreads and low volume

Bid-ask bounce induces negative first-order autocorrelation in observed returns, even when true returns are serially independent. This statistical artifact arises because price changes reverse direction with each successive trade.

• First-order autocorrelation coefficient is typically negative
• Magnitude is proportional to the square of the spread
• Can cause mean reversion signals in high-frequency data
• Distorts volatility estimates and risk models

The oscillation between bid and ask artificially inflates realized volatility estimates. Observed variance exceeds true variance by a component directly related to the bid-ask spread.

• Realized variance = True variance + Spread-induced noise variance
• Effect is magnified at higher sampling frequencies
• Creates a volatility signature plot that declines as sampling interval increases
• Can mislead risk management and option pricing models

The Roll (1984) model provides the foundational framework for estimating the effective bid-ask spread from the serial covariance of price changes. It decomposes observed prices into a random walk plus a bounce component.

• Effective spread = 2 × √(-Cov(Δpₜ, Δpₜ₋₁))
• Assumes equal probability of buy and sell orders
• Breaks down when serial covariance is positive
• Forms the basis for modern transaction cost estimation

Bid-ask bounce creates a downward bias in tests of market efficiency and return predictability. The negative autocorrelation can mask genuine momentum effects or exaggerate mean reversion.

• Variance ratio tests are biased toward rejecting the random walk
• Can create false evidence of short-term overreaction
• Requires bias correction using quote midpoint returns
• Critical consideration for high-frequency strategy backtesting

The standard correction for bid-ask bounce is to use quote midpoint returns instead of transaction price returns. The midpoint represents the consensus value between bid and ask, eliminating the mechanical oscillation.

• Midpoint = (Best Bid + Best Ask) / 2
• Removes the alternating trade direction effect
• Requires synchronized quote and trade timestamps
• May still contain inventory and information effects

The primary measure of execution cost directly linked to bid-ask bounce. It is calculated as twice the absolute difference between the trade price and the midpoint at the time of the trade. This captures the round-trip cost of a transaction, quantifying the implicit cost of crossing the spread. A wider effective spread indicates higher liquidity cost and greater potential for bounce-induced volatility in observed returns.

A classic market microstructure model that explicitly estimates the effective bid-ask spread from the serial covariance of transaction price changes. The model assumes that the observed negative serial correlation in returns is entirely caused by bid-ask bounce. The estimated spread is calculated as:

• S = 2 × sqrt(-Cov(ΔP_t, ΔP_{t-1})) This provides a simple, trade-based proxy for the spread when quote data is unavailable.

The broader category of high-frequency price disturbances that bid-ask bounce falls under. Microstructure noise encompasses all deviations of observed transaction prices from the efficient price (the true fundamental value). This noise:

• Creates spurious volatility in realized variance calculations
• Biases high-frequency beta and correlation estimates
• Necessitates noise-robust estimators like Two-Scales Realized Volatility (TSRV)

A trading benchmark that is less sensitive to bid-ask bounce than arrival price benchmarks. Because VWAP averages prices over a full trading horizon, the transient oscillations caused by bouncing between the bid and ask are smoothed out. However, aggressive execution to match the VWAP schedule can still incur spread costs on each individual fill, making it essential to account for bounce when evaluating VWAP performance.

The difference between the expected price of a trade and the actual execution price. Bid-ask bounce is a direct contributor to slippage, especially for market orders that execute immediately at the opposing quote. A buy market order fills at the ask, and if the next trade occurs at the bid, the observed price change reflects bounce rather than a true market movement, creating the illusion of negative slippage.