Price of Signal-Anticipation (PoSA)

• Price of Signal-Anticipation (PoSA) is a metric that compares enriched anticipatory policies against restricted decision rules across various domains.
• It captures the extra cost in voltage control, cost reduction in optimal execution, and performance gaps in portfolio optimization through comparative statics.
• Analytical bounds and spectral properties ensure that PoSA remains bounded and that efficiency losses diminish in large-scale systems.

Searching arXiv for the cited PoSA-related papers and nearby terminology to ground the article. arxiv_search(query="Price of Signal-Anticipation OR Signal-Anticipation in Local Voltage Control in Distribution Systems OR Static vs Adaptive Strategies for Optimal Execution with Signals OR Anticipatory Portfolio Optimization", max_results=10) arxiv_search(query="(Liu et al., 2018, Bellani et al., 2018, Alonso, 2 Jun 2026)", max_results=10) Price of Signal-Anticipation (PoSA) denotes an objective gap induced by acting on an anticipatory model rather than a restricted benchmark. In the distribution-systems literature, PoSA measures the extra global cost incurred when local Volt/Var controllers anticipate their own effect on voltage instead of taking the voltage signal as fixed (Liu et al., 2018). In optimal execution, the same term measures the cost reduction achieved by adaptive trading relative to a pre-committed static schedule in the presence of a predictive signal (Bellani et al., 2018). In anticipatory portfolio optimization, PoSA is defined as the realized control gap between an anticipatory policy that solves a richer decision problem and a naive myopic policy calibrated under a restricted estimator (Alonso, 2 Jun 2026). The shared theme is comparison between restricted and enriched decision rules; the sign and interpretation depend on whether the underlying problem is posed as cost minimization or reward maximization.

1. Conceptual definitions and sign conventions

Across the cited literatures, PoSA is not tied to a single formula. It is a domain-specific gap between two equilibria, policies, or controls that differ in how they treat signal dependence, feedback, or endogenous impact. The comparison is structurally similar across applications, but the benchmark and objective differ.

Setting	Benchmark comparison	PoSA definition
Local voltage control in distribution systems	Signal-taking network equilibrium $q^*$ vs signal-anticipating Nash equilibrium $q^a$	$\PoSA = F(q^a)-F(q^*)$
Optimal execution with signals	Optimal static strategy vs optimal adaptive strategy	$\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$
Anticipatory portfolio optimization	Naive policy $\theta_{\rm na}$ vs anticipatory policy $\theta_{\rm an}$	$$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$

In the distribution-network setting, PoSA is an inefficiency metric: it is the gap in the social cost $F$ between the network optimum under signal-taking control and the Nash equilibrium under signal-anticipating control (Liu et al., 2018). In the execution and portfolio settings, PoSA is instead the value of adaptivity or model enrichment: it is nonnegative under correct specification because the anticipatory policy solves a less restricted problem (Bellani et al., 2018, Alonso, 2 Jun 2026).

This suggests that PoSA is best regarded as a comparative statics concept for anticipatory decision-making rather than a universally signed welfare quantity. Its meaning depends on whether anticipation is modeled as strategic decentralization, adaptive response to predictive signals, or optimization under an enriched law.

2. PoSA in local voltage control on radial distribution networks

The formulation in “Signal-Anticipation in Local Voltage Control in Distribution Systems” considers an $n$-bus radial distribution network with positive-definite reactance matrix $X\in\mathbb R^{n\times n}$, fixed voltage component $q^a$0, and reactive injection vector $q^a$1. The linearized power-flow law is

$q^a$2

Each bus $q^a$3 has convex provisioning cost

$q^a$4

The signal-taking or “social” cost is

$q^a$5

with unique minimizer

$q^a$6

where $q^a$7 and $q^a$8 (Liu et al., 2018).

Signal-anticipating behavior changes the local optimization problem because each bus accounts for its own self-sensitivity. The corresponding global objective is

$q^a$9

with unique minimizer

$\PoSA = F(q^a)-F(q^*)$0

The paper shows that the interaction among buses becomes a game, and that the Nash equilibrium coincides exactly with the global minimizer of $\PoSA = F(q^a)-F(q^*)$1 (Liu et al., 2018).

Under Assumptions A1–A2, described as strictly decreasing, Lipschitz droop functions, the best-response update

$\PoSA = F(q^a)-F(q^*)$2

converges to the unique Nash equilibrium $\PoSA = F(q^a)-F(q^*)$3. The paper further establishes asymptotic global stability and states that signal-anticipating voltage control has a less restrictive convergence condition than signal-taking control (Liu et al., 2018).

Within this framework, PoSA is defined by

$\PoSA = F(q^a)-F(q^*)$4

The object being measured is therefore the extra social cost created by strategic signal-anticipation at equilibrium rather than the local private objective $\PoSA = F(q^a)-F(q^*)$5.

3. Closed-form representation, spectral bounds, and scaling laws

A central result of the distribution-systems analysis is an exact quadratic representation of PoSA. Defining

$\PoSA = F(q^a)-F(q^*)$6

the paper proves

$\PoSA = F(q^a)-F(q^*)$7

The corresponding worst-case normalized quantity is

$\PoSA = F(q^a)-F(q^*)$8

This converts the inefficiency induced by signal-anticipation into a spectral property of the network and control-cost matrices (Liu et al., 2018).

The upper bound in Theorem 4 is

$\PoSA = F(q^a)-F(q^*)$9

The lower bound in Theorem 5 is

$\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$0

Using Weyl’s inequality, the analysis further shows

$\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$1

and hence

$\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$2

independent of the network size $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$3 (Liu et al., 2018).

Two consequences are emphasized. First, PoSA is universally bounded by a constant determined by the smallest inverter cost $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$4, so it does not grow arbitrarily with the size of the network. Second, the average loss per node vanishes: $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$5 The paper interprets this as a desirable property: no mechanism is needed to mitigate the signal-anticipating behavior because the efficiency loss stays bounded and the per-node loss disappears in large networks (Liu et al., 2018).

For a special case with a line graph of length $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$6, uniform line reactance $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$7, and uniform $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$8, the inverse of $\mathrm{PoSA}=C_{\rm static}-C_{\rm adaptive}$9 has explicit eigenvalues

$\theta_{\rm na}$0

and the resulting closed-form upper bound for $\theta_{\rm na}$1 remains $\theta_{\rm na}$2 as $\theta_{\rm na}$3 (Liu et al., 2018).

The theoretical analysis is complemented by numerical experiments on a 42-bus SCE feeder with 5 PV inverters at buses 2, 12, 26, 29, and 31. The Volt/Var droop, following IEEE 1547.8, is

$\theta_{\rm na}$4

with deadband $\theta_{\rm na}$5 p.u., cost $\theta_{\rm na}$6, and inverter limits enforced. $\theta_{\rm na}$7 was computed by sampling $\theta_{\rm na}$8 (Liu et al., 2018).

$\theta_{\rm na}$9	$\theta_{\rm an}$0
5	0.004
10	0.0025
15	0.0017
20	0.0012
25	0.0009
30	0.0007

In these tests, the loss is below $\theta_{\rm an}$1 and decreases with larger $\theta_{\rm an}$2. The convergence tests also confirm that the signal-anticipating update converges under significantly larger $\theta_{\rm an}$3, equivalently smaller $\theta_{\rm an}$4, than the standard signal-taking scheme (Liu et al., 2018).

4. PoSA in optimal execution with predictive signals

In “Static vs Adaptive Strategies for Optimal Execution with Signals,” PoSA is introduced in a different optimization setting: a trader liquidates inventory over $\theta_{\rm an}$5 while observing a short-term predictive signal (Bellani et al., 2018). The signal $\theta_{\rm an}$6 is an Ornstein–Uhlenbeck process,

$\theta_{\rm an}$7

and the unaffected mid-price satisfies

$\theta_{\rm an}$8

so that $\theta_{\rm an}$9. Inventory evolves under trading speed $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$0 as

$$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$1

and instantaneous linear impact with parameter $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$2 yields execution price

$$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$3

The expected net reward is

$$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$4

with running risk penalty $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$5 and terminal penalty $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$6 (Bellani et al., 2018).

The paper compares two problems. In the static problem, $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$7 is a deterministic pre-committed strategy satisfying the fuel constraint $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$8. In the adaptive problem, $$\mathrm{PoSA}=J(\theta_{\rm an})-J(\theta_{\rm na})$$9 is chosen dynamically in the full filtration. The static solution yields a closed-form optimal inventory path $F$0 obtained from the first-order condition and a second-order ODE with boundary conditions $F$1, $F$2. The adaptive solution is obtained from the HJB equation, leading to a unique optimizer

$F$3

and, under the OU signal,

$F$4

The resulting static and adaptive costs are

$F$5

(Bellani et al., 2018).

PoSA is then defined as

$F$6

In the OU case, the paper shows that the linear-in-$F$7 term cancels and derives the instantaneous-impact formula

$F$8

It also states that $F$9, so $n$0 (Bellani et al., 2018).

The interpretation in this literature is that adaptivity to the evolving signal reduces execution costs relative to a schedule fixed at time $n$1. The paper characterizes the parameter dependence directly from the closed-form expression: PoSA increases with signal volatility $n$2, decreases with mean-reversion $n$3, tends to $n$4 as $n$5, tends to $n$6 as $n$7, and for long horizons $n$8 with moderate $n$9 grows roughly linearly with $X\in\mathbb R^{n\times n}$0 (Bellani et al., 2018).

The paper also treats transient impact. In that case, no closed-form fully adaptive solution is known. Let $X\in\mathbb R^{n\times n}$1 denote the static optimal inventory in the transient-impact problem and $X\in\mathbb R^{n\times n}$2 the “piecewise-static with $X\in\mathbb R^{n\times n}$3 updates” strategy. Then

$X\in\mathbb R^{n\times n}$4

The numerical evidence reported is that $X\in\mathbb R^{n\times n}$5 as $X\in\mathbb R^{n\times n}$6 (Bellani et al., 2018).

5. Anticipatory portfolio optimization and the generalized PoSA framework

“Anticipatory Portfolio Optimization” extends the PoSA concept into a broad decision-theoretic framework in which anticipation can be informational, dynamic, or performative (Alonso, 2 Jun 2026). A portfolio is anticipatory when its optimizer acts on a richer model than the myopic, price-taking estimator used to calibrate it. With feasible set $X\in\mathbb R^{n\times n}$7, return law $X\in\mathbb R^{n\times n}$8, information sets $X\in\mathbb R^{n\times n}$9, and concave utility $q^a$00, the realized objective is

$q^a$01

The naive policy $q^a$02 solves the restricted problem with $q^a$03 frozen at the naive fixed-point, while the anticipatory policy $q^a$04 solves

$q^a$05

PoSA is defined as

$q^a$06

under correct specification (Alonso, 2 Jun 2026).

Several classical formulations appear as special cases. For log utility under initial enlargement of filtration, the public Brownian motion admits the decomposition

$q^a$07

where $q^a$08 is the information drift, and the extra expected log-wealth is exactly

$q^a$09

Under standard hypotheses this equals the relative-entropy or mutual-information cost of passing from the public to the enriched filtration; in the discrete-signal case it reduces to $q^a$10 (Alonso, 2 Jun 2026).

In mean-variance form, if a signal $q^a$11 refines the unconditional mean $q^a$12 into $q^a$13, with

$q^a$14

then the myopic Markowitz rule is $q^a$15 and the signal-adapted one is $q^a$16. Their value gap is

$q^a$17

This gives a closed-form quadratic value of signal resolution (Alonso, 2 Jun 2026).

The most general finite-horizon formulation in the paper is an LQG model with stacked holdings $q^a$18, discount $q^a$19, transaction-cost matrix $q^a$20, and risk term $q^a$21. The objective is

$q^a$22

with score $q^a$23, precision $q^a$24, and price-taker precision $q^a$25. The exact three-way decomposition for the full anticipatory value over the restricted price-taking path $q^a$26 is

$q^a$27

where $q^a$28 is the resolved-score covariance (Alonso, 2 Jun 2026).

Expanding the $q^a$29-norm produces an information term, an impact term, a forecast term, and a signed forecast-impact cross term: $q^a$30

$q^a$31

$q^a$32

$q^a$33

With the $q^a$34-angle $q^a$35 defined by

$q^a$36

the paper derives the sharp bounds

$q^a$37

It also gives the orthogonal projection split

$q^a$38

when $q^a$39 in the $q^a$40-geometry (Alonso, 2 Jun 2026).

In the special case of permanent impact, the paper states that price-taking allocation

$q^a$41

changes into

$q^a$42

This makes the impact component of anticipation explicit as a precision adjustment that counts market impact twice instead of once (Alonso, 2 Jun 2026).

The stationary infinite-horizon extension endogenizes information covariance through Kalman error reduction and represents the impact-anticipation gap as

$q^a$43

a discounted Lyapunov trace. The paper also distinguishes correctly specified, vacuous, and misspecified anticipation. Correctly specified anticipation yields nonnegative PoSA; vacuous anticipation gives $q^a$44; misspecified anticipation can be harmful. The estimation penalty is

$q^a$45

so the expected value of estimated anticipation is

$q^a$46

A robust DRO formulation under a 1-Wasserstein ball adds the dual-norm penalty $q^a$47 to the mean term, and the bias-variance-optimal scalar shrinkage is $q^a$48 (Alonso, 2 Jun 2026).

6. Interpretation, scope, and recurring misconceptions

A recurrent source of confusion is that identical terminology is used for different objective orientations. In the voltage-control setting, PoSA is an inefficiency gap in a cost-minimization problem: signal-anticipation moves the system from the social minimizer $q^a$49 to the strategic equilibrium $q^a$50, and PoSA measures the resulting cost increase (Liu et al., 2018). In the execution and portfolio settings, PoSA is a value-of-adaptation quantity in reward-maximization problems: adaptive or anticipatory optimization improves on a static or naive benchmark, so the reported gap is beneficial under the maintained model (Bellani et al., 2018, Alonso, 2 Jun 2026).

A second misconception is that anticipation is automatically either harmful or beneficial. The cited papers support neither universal claim. In local voltage control, anticipatory behavior is strategic and does strictly increase the steady-state social cost, but the increase is spectrally bounded, remains below a constant independent of network size, and has vanishing average loss per node as $q^a$51 (Liu et al., 2018). In execution and portfolio optimization, anticipatory or adaptive control is valuable only insofar as the richer model captures decision-relevant structure. The portfolio framework is explicit that vacuous anticipation has zero value and misspecified anticipation can be negative after estimation error is accounted for (Alonso, 2 Jun 2026).

A third misconception is that PoSA always refers to pure information advantage. The modern portfolio treatment shows that anticipation has at least three faces: information, planning, and impact correction. The finite-horizon LQG formula separates an information trace from an inverse-precision norm, and the norm expansion further separates impact, forecast, and forecast-impact interaction (Alonso, 2 Jun 2026). This suggests that PoSA can index a broad family of enriched-control effects, including enlarged filtrations, horizon forecasts, and performative feedback through deployment laws.

Taken together, the literature presents PoSA as a rigorous comparative device for quantifying what changes when agents stop treating signals, forecasts, or market responses as exogenous. In power networks it measures the welfare loss of strategic local anticipation; in optimal execution it measures the cost savings from adapting to a predictive signal; in anticipatory portfolio theory it becomes a general realized control gap that unifies information value, forecast value, and impact correction within a common quadratic geometry (Liu et al., 2018, Bellani et al., 2018, Alonso, 2 Jun 2026).