Instrument-Appropriate Transaction Cost Models

A single basis-point cost assumption applied uniformly across asset classes will either kill viable strategies or greenlight doomed ones -- cross-asset cost synthesis requires instrument-specific models.

Why This Matters

Transaction costs have three components: explicit costs (commissions, exchange fees, taxes), spread costs (bid-ask crossing), and market impact (price movement caused by the trade itself). Their relative magnitudes vary dramatically by instrument class. An equity cost model applied to options is structurally wrong. A futures cost model that ignores roll costs is deceptively optimistic. A forex model that assumes stable spreads will fail during funding stress.

Cross-asset strategy evaluation makes this reality inescapable. In S&P 500 options, for example, the median round-trip cost on option premium can reach 1,091 bps -- roughly 100 times a typical equity cost assumption. Even a strong cross-sectional signal (IC = +0.068) is destroyed on execution. This primer synthesizes when standard cost models apply as-is and when they need adaptation across instrument classes.

Intuition

Think of transaction costs as friction, but friction that changes shape depending on the surface. Skating on ice requires different assumptions than running on sand. A backtest that models all surfaces as concrete will understate friction on some trades and overstate it on others, producing net-return estimates that are wrong in both directions.

The critical question for each instrument is: what is the correct scaling variable? For equities, costs scale primarily with trade size as a fraction of average daily volume (ADV). For options, costs scale with the option premium, not the underlying notional. For FX majors under normal conditions, costs are sub-basis-point, but they scale with funding stress and volatility. Getting the scaling variable wrong is not a matter of precision -- it is a category error.

Formal Core

Equities

Kyle's foundational model establishes that informed trading creates permanent price impact proportional to order flow, captured by the parameter $\lambda$ (Kyle's lambda) that measures market depth [ref:HLKJBBY9]. Almgren and Chriss derive optimal execution schedules that trade off temporary impact against timing risk, producing an efficient frontier of execution strategies parameterized by risk aversion [ref:EUKUI5IR].

Frazzini, Israel, and Moskowitz provide the empirical anchor. Using $1.7 trillion of live institutional executions across 21 developed equity markets (1998-2016), they find that real-world price impact follows a concave (approximately square-root) function of trade size as a fraction of daily volume [ref:8FR7V3BQ]:

$$\text{Impact} \approx \sigma \cdot c \cdot \left(\frac{Q}{\text{ADV}}\right)^{\delta}$$

where $\sigma$ is stock-level volatility, $Q$ is trade size, ADV is average daily volume, $c$ is a scaling constant, and $\delta \approx 0.5$ (the square-root law). Crucially, they find that costs for a patient institutional trader are an order of magnitude smaller than TAQ-based academic estimates, which aggregate across many trader types and overweight aggressive liquidity-demanding trades [ref:8FR7V3BQ]. Key parameters: stock-level ADV, volatility, and market capitalization.

Futures

Costs center on bid-ask spread (typically 0.5-2 ticks for liquid contracts), roll cost at contract expiry, and margin/funding. Impact is generally lower than equities for comparable notional sizes because futures markets are centralized with high ADV-to-open-interest ratios. However, roll costs are invisible in a naive backtest: a strategy that holds through expiry without modeling the spread paid when rolling from the expiring contract to the next creates a systematic upward bias in net returns.

Foreign Exchange

Karnaukh, Ranaldo, and Soderlind validate that daily data can accurately proxy FX transaction costs, but they demonstrate that liquidity evaporates globally when funding constraints (TED spread) and volatility (VIX) rise [ref:WPTELP7D]. Major pairs like EUR/USD have sub-basis-point quoted spreads under normal conditions; emerging-market and exotic crosses can be 10-50 times wider. FX carries no exchange fees but has implicit costs from the dealer markup and the overnight swap rate. Static cost assumptions will drastically underestimate execution slippage during market stress.

Options

Costs scale with the option premium, not the underlying notional. In S&P 500 options, 1,091 bps round-trip on premium means that even a strong signal is destroyed by execution friction. Bid-ask spreads are wider for out-of-the-money options, illiquid strikes, and during high-volatility regimes. Delta-hedging adds additional cost from the underlying leg. A backtest that converts option costs to notional-equivalent basis points will understate friction by one to two orders of magnitude.

Crypto

Costs combine exchange fees (typically 2-10 bps on centralized exchanges), funding rates on perpetual swaps, and potentially severe slippage in thin order books. Decentralized exchange (DEX) execution adds gas fees and MEV-related adverse selection. In crypto perpetuals, strategies can turn negative when regime changes interact with thin cost margins.

How It Works in Practice

Madhavan provides a practitioner framework connecting microstructure theory to execution design: temporary impact (price pressure that reverts after the trade) and permanent impact (information content impounded into the price) require different modeling treatment [ref:T57YYLDF]. Temporary impact drives the urgency-patience trade-off; permanent impact sets the fundamental cost floor.

Capacity as a cost concept. Chan demonstrates that multi-day impact decay makes true strategy capacity 3.5-10 times smaller than naive estimates that assume impact resets each day [ref:E7AGR99D]. When impact decays slowly, splitting trades over multiple days provides less cost relief than expected because unreverted price pressure accumulates. Swinkels documents that momentum profits -- historically the strongest equity anomaly -- erode above modest capital levels due to price impact concentrated on the long side [ref:WNAQCRQZ].

The portfolio-to-execution linkage is inseparable from cost modeling. Cerniglia and Fabozzi emphasize that portfolio construction and execution must be jointly designed: the allocator determines the trade list, the trade list determines the cost, and the cost feeds back into net performance [ref:2YXWIIVF].

Worked Example

Consider two stylized strategies, both with positive gross alpha:

US firm characteristics (equities). Round-trip cost of approximately 12.5 bps using the Frazzini et al. square-root model [ref:8FR7V3BQ]. A gross Sharpe of 1.2 survives to a net Sharpe above 0.8. The dominant cost component is market impact, scaling with trade size relative to ADV. The strategy has meaningful capacity headroom.

S&P 500 options cross-section. Round-trip cost of 1,091 bps on premium. A gross Sharpe that would be attractive in equities is destroyed: the signal's IC of +0.068 cannot overcome the cost drag when translated into net returns. The dominant cost component is the bid-ask spread, scaling with premium. Capacity is severely constrained.

Same modeling pipeline. Same evaluation framework. Radically different cost regimes. A uniform 10-bps assumption would greenlight the options strategy and understate the equity strategy's advantage.

Practical Guidance

Identify the dominant cost component per instrument. For equities: market impact. For futures: spread plus roll. For FX: spread, regime-dependent. For options: bid-ask on premium. For crypto: spread plus funding plus slippage.

Use the correct scaling variable. Equities and futures scale costs with trade size / ADV. Options scale with premium. FX scales with spread, which itself scales with funding stress and volatility.

Model regime dependence. FX liquidity evaporates during funding crises [ref:WPTELP7D]. Option spreads widen during volatility spikes. Crypto spreads blow out during market dislocations. Static cost assumptions are worst-case wrong precisely when they matter most.

Validate against live data when possible. Frazzini et al. demonstrate that TAQ-based cost models overstate equity impact by an order of magnitude relative to patient institutional execution [ref:8FR7V3BQ]. Backtests calibrated to academic estimates may reject profitable strategies.

Test capacity sensitivity. If a strategy's net performance turns negative at modest AUM, the signal is real but the instrument's cost structure prevents profitable implementation at scale [ref:E7AGR99D].

Where It Fits in ML4T

Chapter 18 develops transaction cost theory, market impact models, and optimal execution frameworks. Chapter 20 applies these models across equities, futures, FX, options, and crypto. This primer shows when Chapter 18's models apply directly and when instrument-specific adaptation is required. The portfolio construction primer (Primer 03) explains how the allocator generates the trade list that these cost models then evaluate. Adjacent topics include the IC-to-Sharpe conversion gap and regime robustness across asset classes.