Fredholm Approach to Nonlinear Propagator Models
Eduardo Abi Jaber, Alessandro Bondi, Nathan De Carvalho, Eyal Neuman, Sturmius Tuschmann · 2025
Read Fredholm Approach to Nonlinear Propagator Models by Eduardo Abi Jaber, Alessandro Bondi, Nathan De Carvalho, Eyal Neuman, Sturmius Tuschmann (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-04323.txt.
Introduction
Eduardo Abi Jaber and four co-authors solved a trading task with lasting price impact through a Fredholm frame. Their paper, from twenty twenty-five, deals with propagator models of market impact. A trade moves price, and that move fades in a set curve, here with power law tails. The trader holds a signal on where price will go and must time buys and sells. The best plan meets a nonlinear stochastic Fredholm equation with terms that look both ways in time.
They prove a sole best plan exists under a monotone term that marks the impact's bend. Past that term, a best plan still exists on a countable chance space. A loop scheme homes in on the plan, with a code run to show it.
Catalog record
| CREATOR | Eduardo Abi Jaber, Alessandro Bondi, Nathan De Carvalho, Eyal Neuman, Sturmius Tuschmann |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 118 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-04323
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Fredholm Approach to Nonlinear Propagator Models Eduardo Abi Jaber, Alessandro Bondi, Nathan De Carvalho, Eyal Neuman, Sturmius Tuschmann Abstract We formulate and solve an optimal trading problem with alpha signals, where transactions induce a nonlinear transient price impact described by a general propagator model, including power-law decay. Using a variational approach, we demonstrate that the optimal trading strategy satisfies a nonlinear stochastic Fredholm equation with both forward and backward coefficients. We prove the existence and uniqueness of the solution under a monotonicity condition reflecting the nonlinearity of the price impact. Moreover, we derive an existence result for the optimal strategy beyond this condition when the underlying probability space is countable. In addition, we introduce a novel iterative scheme and establish its convergence to the optimal trading strategy. Finally, we provide a numerical implementation of the scheme that illustrates its convergence, stability, and the effects of concavity on optimal execution strategies under exponential and power-law decay. Mathematics Subject Classification (2010): 93E20, 60H30, 91G80 JEL Classification: C02, C61, G11 Keywords: Optimal trading, nonlinear market impact, propagator model, power-law decay, square root-law, Fredholm equations, nonlinear stochastic control 1 Introduction Price impact refers to the empirical observation that executing a large order adversely affects the price of a risky asset in a persistent manner, leading to less favorable execution prices. Consequently, an agent seeking to liquidate a large order, known as a metaorder , must split it into smaller parts, referred to as child orders , which are typically executed over a period of hours or days. A fundamental question in this context concerns the impact of a metaorder as a function of its size. A large body of empirical and theoretical work suggests that this relationship is concave and in fact, well approximated by a square-root law, see e.g. [ 13 , 15 , 25 , 30 , 32 , 34 ] and Chapter 12.3 of [ 9 ] for a comprehensive introduction. The square-root price impact law states that the peak impact D^{\textrm{peak}} for a metaorder of volume X , executed over a time horizon T , is given by D^{\textrm{peak}}=Y\sigma_{T}\left(\frac{X}{V_{T}}\right)^{\delta},\quad\textrm{for }X\ll V_{T}, (1.1) where Y is a constant of order 1, V_{T} is the total traded volume in the market, and \sigma_{T} is the asset’s contemporaneous volatility, both measured over time T . The scaling exponent \delta is estimated to be in the range [0.4,0.7] , and is often taken as 0.5 as an approximation. The square-root law has been shown to be a good approximation in various markets (such as equities, foreign exchange, options, and even cryptocurrencies) and is general enough to include different types of market liquidity, broad classes of execution strategies, and various trading frequencies ( [ 7 , 12 , 33 , 34 ] ). While the square-root law provides a simple connection between the metaorder size and its impact on the mid-price, price impact evolves dynamically, and the reaction of the mid-price to the metaorder is mostly transient. Propagator models are a central tool for mathematically describing this decay phenomenon. They express price moves in terms of the influence of past trades and therefore capture the decay of price impact after each trade ( [ 9 , 18 ] ). The price distortion D_{t} is quantified by D^{X}_{t}=\int_{0}^{t}G(t,s)dX_{s},\quad t\geq 0, (1.2) where X_{t} describes the amount of shares of the asset executed by time t and G(t,s) is a positive semi-definite Volterra kernel, often referred to as a propagator . Obizhaeva and Wang [29] and Gârleanu and Pedersen [17] , along with follow-up papers, assume that the price distortion D_{t} decays exponentially over time, that is, G(t,s)=\kappa e^{-\rho(t-s)}\mathbbm{1}_{\{t>s\}} for some positive constants \kappa and \rho . On the other hand, Bouchaud et al. [9] (see Chapter 13.2.1 and references therein) report on empirical observations that the propagator G decays as a power-law of the lag, i.e., G(t,s)\approx(t-s)^{-\beta}\mathbbm{1}_{\{t>s\}},\quad t,s\geq 0, (1.3) with 0<\beta<1 , and these results are also supported by theoretical arguments. While the power-law behavior of price impact decay is strongly supported by market data, results on corresponding optimal liquidation strategies have been scarce. Power-law or other general propagators that do not exhibit exponential decay introduce non-Markovianity in the corresponding optimal trading problem, thereby rendering standard stochastic control tools, such as dynamic programming or FBSDEs, inapplicable. Recent progress has been made in the context of optimal execution with the linear propagator model, see [ 19 ] for the case without signals and [ 1 , 2 , 4 ] for the case with stochastic signals. However, incorporating both the nonlinearity arising from the square-root law in ( 1.1 ) and the power-law decay in ( 1.3 ) remains a long-standing open problem. As a first step in merging these models, Alfonsi et al. [5] introduced a nonlinear propagator model, motivated by the local shape density of the limit order book and its reaction to market orders. In their model, the price impact I_{t} of a metaorder (X_{t})_{t\geq 0} is given by I^{X}_{t}=h(D^{X}_{t}),\quad t\geq 0, (1.4) where h is an increasing, nonnegative, and concave function, and D^{X} is defined in ( 1.2 ). Note that the case h(x)=x corresponds to the linear propagator model, while the case h(x)=\text{sign}(x)|x|^{\delta} , with \delta\in(0,1) , is inspired by ( 1.1 ) (see [ 8 , 11 , 21 ] ). Alfonsi et al. [5] then solved the associated execution problem, which involves minimizing the expected execution costs, C(X)=\mathbb{E}\left[\int_{0}^{T}h(D^{X}_{t})dX_{t}\right], (1.5) in the special case where G(\cdot) is an exponentially decaying function, i.e., the Markovian case, obtaining closed-form, deterministic optimal strategies. Hey et al. [21] generalized the framework of [ 5 ] by incorporating a predictive trading signal (also referred to as alpha ) into the price process. From a modeling perspective, this means that the unaffected price process is no longer a martingale. The agent therefore faces a tradeoff between exploiting the signal to make an immediate profit and managing price impact costs. The objective functional then changes to maximizing the expected revenue functional: R(X)=\mathbb{E}\left[\int_{0}^{T}\left(\alpha_{t}-I_{t}^{X}\right)dX_{t}\right], (1.6) where the signal (\alpha_{t})_{t\geq 0} captures the drift term information of the unaffected asset price. In addition, inventory risk and a penalty for terminal inventory can be incorporated into ( 1.6 ). The optimal execution strategy with respect to ( 1.6 ) is expected to be adapted to the signal’s filtration. Hey et al. [21] worked within the Markovian setup and derived a first-order condition that characterizes an implicit connection between price distortion and the signal. A more tractable relation between impact and alpha was then established in the case where the function h(x) in ( 1.4 ) follows a power-law, and the optimal price impact was derived in closed form. The model in [ 21 ] can be regarded as a generalization of the exponentially decaying linear propagator model studied in [ 24 , 26 ] , for which closed-form solutions were derived. For additional work on cost misspecification in the setting of Hey et al. [21] , we refer to [ 20 ] , and a similar problem for the general linear propagator model was studied in [ 28 ] . The main goal of this work is to study the optimal execution problem in which the price impact is a nonlinear