Transient Impact in Spot FX Markets

• Transient Impact in Spot FX is defined as the time-decaying effect of trades on the mid-price due to market resilience and liquidity dynamics.
• Mathematical models using propagator functions and exponential decay kernels capture the non-monotonic, time-dependent response of FX prices.
• Optimal execution strategies integrate transient impact to balance inventory risk, transaction costs, and dealer resilience for improved market making.

Transient impact in spot foreign exchange (FX) markets refers to the phenomenon whereby the price impact of trades is neither entirely permanent nor purely instantaneous, but rather decays over a finite time scale. The mid-price response to order flow is characterized by a non-monotonic, time-dependent function reflecting resilience and mean-reversion in the market. Empirical evidence demonstrates that in spot FX, as in equity markets, impact is best modeled as a transient process, with both the amplitude and decay shape tightly linked to market liquidity and execution strategies. The mathematical modeling and optimal control of trading and quoting under transient impact have become central problems in FX microstructure, market making, and algorithmic execution research.

1. Transient Impact: Empirical Phenomena and Definitions

In spot FX, trades exert an immediate influence on the mid-price, but this effect relaxes over time due to market resilience and subsequent order flow. The empirical response function $R_i(\tau)$ for currency pair $i$ measures the average mid-price return after a trade, conditional on the trade’s sign, as a function of lag $\tau$ in either trade-time or calendar-time units. The form is

$$R_i^{(\mathrm{scale})}(\tau) = \langle r_i^{(\mathrm{scale})}(t-1,\tau) \varepsilon_i^{(\mathrm{scale})}(t) \rangle ,$$

where $r_i$ is the relative mid-price change, and $\varepsilon_i$ is the trade sign. Large-scale data for FX majors over multiple years reveal a universal non-Markovian signature: $R(\tau)$ rises to a peak at a characteristic lag $\tau_{\max}$ (of order $10^3$ trades or a few minutes), then decays as a slow power law $R(\tau) \sim \tau^{-\beta}$ with $\beta\sim 0.2$–$0.4$ (Londono et al., 2021).

Once trades are grouped by their average bid-ask spread, liquidity-dependence appears: pairs with wider average spreads display stronger peak impact and shallower decay. This relationship, summarized in Table 1, mirrors equity market findings and confirms that spot FX exhibits universal transient-impact fingerprints.

Spread Group	Average Spread (pips)	$R_\mathrm{max}$ (trade time, 2015)
G1	1–3	$1.0 \times 10^{-4}$
G2	3–10	$3.0 \times 10^{-4}$
G3	$>10$	$6.0 \times 10^{-4}$

The non-monotonic, slowly decaying shape of the response function validates the use of transient impact models in FX pricing and execution (Londono et al., 2021).

2. Mathematical Modeling of Transient Impact in Spot FX

The canonical transient impact model (or "propagator" model) postulates that every trade at time $s$ contributes to the current price via an impact function that decays over time. The mid-price dynamic is:

$$S_t = S_0 + \int_0^t f(v(s)) G(t-s)\, ds + \sigma W_t,$$

where $v(s)$ is the signed trading rate, $f(v)$ is the instantaneous impact function, $G(u)$ is the decay kernel, and $W_t$ is Brownian noise (Curato et al., 2014). For spot FX dealers, a tractable case uses an exponential decay kernel:

$G(u) = k e^{-\beta u}, \quad u \geq 0,$

where $k$ is the base impact and $\beta$ the resilience parameter. The impact decay timescale $\tau_\mathrm{imp} = 1/\beta$ is empirically found to be of order minutes, matching the natural inventory-rebalancing timescale of active FX dealers (Barzykin, 19 Jan 2026).

Limiting cases of this model recover standard frameworks:

• $\beta \to 0$: $G(u) = k$ yields purely permanent impact (as in Almgren–Chriss).
• $\beta \to \infty$: $G(u) \propto \delta_0(u)$ gives purely instantaneous (temporary) impact.

Nonlinear generalizations adopt empirically fitted forms $f(v) = \mathrm{sign}(v)|v|^\delta$, $\delta < 1$, and $G(u) \sim u^{-\gamma}$, $0 < \gamma < 1$ (Curato et al., 2014), to capture concavity and long-range decay.

3. Optimal Execution and Market Making under Transient Impact

FX market makers and execution algorithms face a complex objective: they must weigh inventory risk, spread revenues, transaction costs, and the impact propagation from both client trades and their own hedging in the interdealer market. The control variables consist of bid/ask skews $\delta_t^{n,b/a}$ set for various client order sizes and the continuous hedging speed $v_t$.

The optimal policy is derived by solving a Hamilton–Jacobi–Bellman (HJB) equation for the risk-adjusted value function $V(t,q,x)$, where $q$ is inventory and $x$ the transient impact state:

$$\begin{align} 0 = \partial_t V &- \beta x(q + \partial_x V) - \frac{1}{2} \gamma \sigma^2 q^2 \ &+ H_E(p_E) + \sum_n \Delta^n [H_\mathrm{OTC}^n(D_{q+}^n V) + H_\mathrm{OTC}^n(D_{q-}^n V)]. \end{align}$$

Here, $p_E$ aggregates impact-adjusted inventory and value gradients, while $H_E$ and $H_\mathrm{OTC}^n$ encode execution and OTC quoting logic, respectively (Barzykin, 19 Jan 2026).

Assuming a quadratic value function ansatz, one obtains explicit feedback forms for the optimal quoting and hedging policy. The skews and hedging rate depend nontrivially on both inventory $q$ and transient impact state $x$, with sensitivity to the decay kernel via $B_0 = \beta/(\beta + \omega)$, $\omega$ being the risk-relaxation rate.

4. Comparative Statics, Calibration, and Empirical Evidence

Key parameters in transient-model-based FX execution include:

• Resilience $\beta$: The magnitude of $B_0 = \beta/(\beta + \omega)$ governs the response to transient impact state. As $\beta$ increases, quoting and hedging policies become more sensitive to $x$, exploiting mean-reversion in the impact decay.
• Risk-aversion $\gamma$: Appears as $A_0 \propto \sqrt{\gamma}$; higher $\gamma$ leads to wider spreads, more aggressive inventory-flattening, and greater aversion to risk.
• Transaction costs $\psi$, $\eta$: Determine execution thresholds; higher costs increase the “no-trade” (pure-internalization) region.

Empirical calibration in spot FX, using real transactional data, suggests $k=0.005$ bp/M, $\beta=1000$ day$^{-1}$ (implying $\tau_\mathrm{imp} \approx 1.4$ minutes), and typical hedging half-lives $\omega^{-1} \approx 2$ minutes (Barzykin, 19 Jan 2026). Monte Carlo simulations reveal that accounting for transient impact materially improves P&L for large inventory shocks, confirming the importance of resilience modeling in dealer algorithms.

5. Transient Impact and Nonlinear Execution in Spot FX

Optimal execution in the presence of transient impact requires addressing both the decay of past impact and the nonlinearities of immediate price response. For realistic concave-impact and power-law decay kernels, the optimal execution schedule becomes front-loaded: concentrate more volume likely early in the interval to minimize cost, as later trades have diminished impact due to decaying memory (Curato et al., 2014).

However, without proper regularization, strongly concave impact models may admit pathological, even arbitrage-generating “oscillating” execution strategies. Remedies include introducing explicit spread costs or modeling a convexification of the impact function at large trading rates to maintain economic consistency and practical implementability.

Adaptive strategies that include predictive signals or drift ($A_t$) in the mid-price dynamically balance between exploiting exogenous prediction, minimizing temporary cost, and timing the recovery from self-induced transient distortions. In models with both temporary and transient impact, the optimal trading rate $\hat v_t$ is computed as a linear feedback in inventory, transient state, and anticipated signal, and can be implemented via coupled forward-backward SDEs with explicit coefficients (Neuman et al., 2020).

6. Theoretical and Practical Implications for FX Dealers

Transient impact modeling bridges the gap between purely instantaneous and permanent price impact approaches. The empirically observed impact-recovery timescale, comparable to the dealer’s risk management horizon, makes the explicit inclusion of resilience essential for optimal market making and hedging in spot FX (Barzykin, 19 Jan 2026).

For typical small, frequent executions, the difference between transient and permanent models is minor; however, for discretionary or large trades, incorporating transient impact provides significant improvements in expected performance and risk control. Calibration to empirical response curves, as exhibited in large-sample studies (Londono et al., 2021), is crucial for accurate parameterization and robust operation across liquidity regimes.

In summary, the FX transient impact literature establishes that optimal order scheduling, quoting, and risk management must integrate decaying impact kernels, liquidity-adjusted amplitude, and the interplay with risk preferences and predictive signals for market participants to achieve superior execution and P&L outcomes in spot FX. Empirical, theoretical, and numerical results consistently validate the relevance of this paradigm for both academic and practitioner applications (Barzykin, 19 Jan 2026, Londono et al., 2021, Curato et al., 2014, Neuman et al., 2020).