Market Impact Model: Estimating Trade Cost Before Execution

Key Takeaways

• Market impact has a permanent component that reprices the asset and a temporary component that reverts once trading stops.
• The square-root law gives impact proportional to volatility times the square root of the order's fraction of average daily volume.
• Buying 500,000 shares in a 10-million-share stock costs roughly 31 basis points of expected impact using a typical square-root estimate.
• Portfolio managers budget pre-trade impact before entering positions, since large orders in thin names can consume most of the expected alpha.

What It Is

Market impact is the adverse price change caused by your own trading. Buying pushes the price up before you finish buying. Selling pushes it down. A market impact model is a formula that estimates the expected size of that move as a function of trade size, horizon, volatility, and average market volume.

Impact models are used for three things: estimating cost before the trade, choosing an execution schedule during the trade, and evaluating execution quality after the trade. Every serious buy-side desk and every execution broker runs some version of one.

The Intuition

A trade has two kinds of price effects. One is temporary: your order consumes resting liquidity, the price dislocates while you trade, and it reverts once you stop. The other is permanent: the trade leaks information, and the market repriced the asset in response. You pay both. The permanent part stays in the price and is what other participants see. The temporary part shows up as extra slippage only for you.

Good impact models separate these components. They also recognize that splitting a large parent order into many small child orders does not make impact disappear. If you trade 5 percent of a day's volume, you pay roughly the same cost whether you spread it over one hour or four, as long as you do not push participation too high.

How It Works

The Almgren-Chriss framework (2001) writes total expected cost as the sum of permanent impact, temporary impact, and a risk term. In their linear specification:

Permanent impact: g(v) = gamma * v
Temporary impact: h(v) = epsilon * sign(v) + eta * v

Where v is the trading rate in shares per unit time, gamma and eta are calibration constants, and epsilon captures a fixed per-trade component such as half the bid-ask spread. Optimizing expected cost plus lambda * variance(cost) yields a closed-form schedule. The optimal holding trajectory decays from initial size to zero as a hyperbolic sine shape, faster for more risk-averse traders.

The square-root law, validated across equities, futures, options, and cryptocurrencies, says that the impact of a metaorder of total size Q is approximately:

impact ~ Y * sigma * sqrt(Q / V)

Where:

Q     = total shares in the parent order
V     = average daily volume
sigma = daily return volatility
Y     = a dimensionless constant, typically between 0.5 and 1

The striking empirical finding is that impact depends on the fraction of daily volume traded, not on how slowly or quickly you execute within a reasonable range. Almgren, Thum, Hauptmann, and Li (2005) found a similar exponent of roughly 0.6 on a large Citigroup dataset of US equity orders.

Worked Example

A portfolio manager wants to buy 500,000 shares of a stock that trades 10 million shares per day on average. Volatility is 2 percent per day. Using the square-root law with Y = 0.7:

impact = 0.7 * 0.02 * sqrt(500,000 / 10,000,000)
       = 0.7 * 0.02 * sqrt(0.05)
       = 0.7 * 0.02 * 0.2236
       = 0.00313

That is roughly 31 basis points, or about 0.31 percent of price. On a $100 stock, the manager should budget about $0.31 per share of expected impact, or about $155,000 total on the parent order. That is the number pre-trade TCA would show. Almgren-Chriss would then add a risk term: trading faster reduces exposure to adverse drift but raises impact; trading slower does the opposite.

Common Mistakes

1. Assuming impact scales linearly with size. A Kyle-style linear model is fine for small orders on short horizons. For parent orders that are a meaningful slice of daily volume, empirical work points to a concave, roughly square-root relationship. Linear extrapolation overestimates the cost of very small orders and underestimates the cost of large ones relative to the square-root prediction.

2. Calibrating on the wrong sample. Models fit to a month of calm markets break in a stressed tape. A serious desk re-estimates parameters through different regimes and flags when realized impact drifts far from prediction.

3. Ignoring the horizon-independence finding. Many traders assume that trading more slowly always lowers impact. Above a moderate participation rate, slowing down mostly adds timing risk without helping impact, because total Q is what matters most.

4. Not distinguishing permanent from temporary impact. If you measure only the round-trip slippage, you lose the piece that matters for other portfolio managers and for post-trade analysis. Decompose when evaluating execution.

5. Treating the constant Y as universal. Bouchaud and colleagues find the square-root law holds broadly, but the prefactor Y varies by asset class, venue, and liquidity regime. Use your own data for calibration, not a generic number.

Frequently Asked Questions

Q: What is a market impact model in simple terms? It is a formula that estimates how much a trade moves the price against itself, taking order size, average daily volume, and volatility as inputs and producing an expected cost in basis points before the trade starts.

Q: How does a market impact model affect investment decisions? Portfolio managers use it to pre-screen trade sizes, cutting back positions where estimated impact exceeds expected alpha, and to set realistic budgets for execution cost that feed into strategy capacity estimates.

Q: What is a real-world example of a market impact model? Buying 500,000 shares of a stock with 10 million shares average daily volume and 2 percent daily volatility costs roughly 31 basis points of expected impact under the square-root law with a typical Y of 0.7.

Q: How can investors use market impact models? Investors calibrate the square-root model on their own trade history, use it to size orders below a participation threshold where impact is acceptable, and monitor post-trade realized impact versus model prediction to detect when market conditions have shifted.

Q: How is a market impact model different from Kyle's lambda? Lambda is a linear per-unit price response estimated from short-interval regressions, suitable for small orders. Impact models use a concave square-root or power-law function appropriate for large parent orders where impact grows slower than linearly with size.

Sources

1. Almgren, R. and Chriss, N. (2001). "Optimal Execution of Portfolio Transactions." Journal of Risk, 3, 5-40. https://www.smallake.kr/wp-content/uploads/2016/03/optliq.pdf
2. Almgren, R., Thum, C., Hauptmann, E., and Li, H. "Direct Estimation of Equity Market Impact." https://www.cis.upenn.edu/~mkearns/finread/costestim.pdf
3. Bouchaud, J.-P. "The Square-Root Law of Market Impact." https://bouchaud.substack.com/p/the-square-root-law-of-market-impact
4. Gatheral, J. "Optimal Execution, Square-Root Law and Models." Baruch College MFE. https://mfe.baruch.cuny.edu/wp-content/uploads/2012/09/Chicago2016OptimalExecution.pdf

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.