Does the Market Anticipate? Can it? Should it?
summary
The gist
We explore a nuance to ’no arbitrage’: it can be suboptimal to act upon an arbitrage immediately; in such cases optimised trading can suppress the anticipation of predictable risky outcomes…
In short
The paper investigates whether immediate arbitrage is always optimal under model- or event-risk. By including pre-horizon risk disclosures, it shows that acting immediately on a mispricing can be suboptimal. This leads to a Status Quo Bias where traders wait, suppressing the anticipation of predictable risky outcomes, suggesting viability and efficiency are mutually enhancing at the cost of anticipation.
Key concepts
- No Arbitrage
- The principle that assets should not be bought or sold in a way that guarantees an immediate risk-free profit. The paper explores how this rule can be temporarily bypassed if acting immediately leads to a worse outcome due to future risks.
- Pre-horizon Risk-Outcome Disclosures
- A setting where the actual outcomes of model or event risks are revealed at a specific, finite time before they occur. This allows traders to make decisions based on future information, which changes the immediate optimal trading strategy.
- Status Quo Bias
- A behavioral tendency where agents prefer to maintain their current state rather than take action, especially when facing uncertain or risky outcomes. In this context, it arises because optimizing immediately might be worse than waiting for clearer risk resolution.
- Model- or Event-Risk
- The framework used to define uncertainty in asset pricing. It distinguishes between risks that are inherent in the underlying mathematical model (model-risk) and risks arising from specific, unpredictable external events (event-risk).
Terminology used across episodes
This episode discusses
The paper
Does the Market Anticipate? Can it? Should it? · Read on arXiv
KANGDA KEN WREN
University of York
Transcript
Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.
Ruth: I'm Ruth.
Warren: And I'm Warren, here with Iris.
Ruth: We're looking at the paper: "Does the Market Anticipate? Can it? Should it?".
Warren: Here's the gist of the paper. We explore a nuance to ’no arbitrage’: it can be suboptimal to act upon an arbitrage immediately; in such cases optimised trading can suppress the anticipation of predictable risky outcomes, creating an apparent Status Quo Bias <ref:2603.02187#pg2>.
Ruth: This is shown through continuoustime asset-pricing under model- or event-risk <ref:2603.02187#pg5>.
Warren: The Core Conflict The paper examines the conflict between ’no arbitrage’, ’information efficiency’ and ’risk anticipation’, and the role of the rate of ’signal-to-noise’ versus that of ’current return’ <ref:2603.02187#pg4>.
Ruth: It introduces a setting where outcomes are disclosed at a finite time, allowing for pre-horizon risk-outcome disclosures <ref:2603.02187#pg2>.
Title and authors: Ruth: So we’re talking about this paper now called "Does the Market Anticipate? Can it? Should it?" by Kangda Ken Wren, and generally speaking, it looks at a subtle point about what we usually think is no arbitrage.
Warren: Exactly. It suggests that even if you have an arbitrage opportunity right now, acting on it immediately might not be the best move because you could be suppressing some kind of future risk anticipation.
Iris: That sounds like it touches on the tension between having a good price and trying to figure out what's going to happen next, which is often where things get complicated in these models.
Ruth: Right. The core idea they are exploring is this conflict between what we call no arbitrage, which is about immediate pricing rules, and information efficiency, which is about how well prices reflect everything known and unknown.
Warren: And they introduce anticipation into that picture as a third element, looking at how much the rate of signal to noise versus current return matters in this setting.
Iris: They set up this scenario where outcomes, like whether there's a recession or some CEO change, are disclosed at a specific point in time tD, which is earlier than the total horizon T.
Ruth: And that timing is key because it allows them to show that it can be suboptimal to position right away for a mispricing or an arbitrage opportunity.
Warren: So instead of jumping in immediately on a perceived mispricing, you might choose a strategy that helps suppress the anticipation of those predictable risky outcomes.
Iris: That leads into what they call Status Quo Bias, which seems to be reinforced when we introduce ambiguity into the system using something called Bewley’s Inertia Axiom.
Ruth: It sounds like they are arguing that viability and efficiency can actually work together, but only if you trade anticipation for immediate action.
Warren: The theoretical framework they build involves model- or event-risk and how risk resolution happens before the full outcome is known.
Iris: They define market types where it’s either B-unsure given public data, or sure under a larger set of dataflow, which is what they call viability.
Ruth: And they introduce specific features like B-specific data and predictable disclosures at some time tD that is much smaller than the total time horizon T.
Warren: Then they develop a Risk-Neutral Equivalent pricing approach, adding this risk-specific data to the standard information structure framework to allow for that pre-horizon risk resolution.
Iris: Their key result is Proposition one which says that under B-risk with data governed by reference law piB0 WB over the time interval from zero to one, any fundamental theorem of asset pricing viable asset price must have this specific form: S t equals the expectation of a certain quantity minus a term called RP hat sub t <ref:2603.02187#pg1>.
Ruth: That formula shows that the price dynamic isn't just about current things; it has to account for this future resolution structure.
Title and authors: Warren: It directly supports the idea that you can’t just act on an arbitrage opportunity because there’s a constraint imposed by how those outcomes are disclosed over time.
Iris: The behavioral implication they draw is that for a predictable risk of change, it is suboptimal to long change unless the dataflow providing information about that change is quite intense.
Ruth: And this leads to something interesting: the market-clearing price at any time t, which they call S h et t t equals the expectation of some quantity at time two four minus one half, isn't a steady state and actually has dynamic behavior during periods of low RII, or risk intensity.
Warren: This dynamic behavior is what creates that apparent Status Quo Bias where the zero trader fair price at time zero, S zero t, acts like a steady state.
Iris: So they conclude that viability and efficiency are actually mutually enhancing things, but this happens at the expense of anticipating future changes.
Ruth: It’s a big idea for how we model markets that moves away from just assuming immediate reaction to any price change.
Warren: It means if you’re trading, you have to consider not just the arbitrage itself but also the anticipation of when and how much risk will resolve before you can actually execute your trade.
Iris: And that brings us to how AI systems could potentially use this: they could learn to distinguish between model- or event-risk and that B-specific data by using compound structures that allow for pre-horizon risk resolution.
Ruth: That sounds like an improvement in how AI can price things because it learns to handle the uncertainty structure more accurately when there's a known future disclosure point.
Warren: It means the AI could optimize trading strategies that are not optimal right away for arbitrage but suppress that anticipation, essentially delaying action until risk resolution is imminent.
Iris: That would help models account for those behavioral biases we see in markets where people seem to stick with the status quo even when things are changing.
Ruth: It's about making the model more robust by acknowledging that sometimes being efficient means not acting immediately on a known outcome.
Warren: The paper also suggests incorporating a price-of-risk, k At any time t, which they define as K one half minus K one half plus sigma pi plus t, which is directly related to the gain-to-loss ratio and risk aversion parameters.
Iris: That term gives us a concrete way to think about the risk component that needs to be priced into those pre-horizon dynamics we discussed.
Ruth: So, we’ve seen how this paper challenges the idea that immediate action on any arbitrage is always best, introducing a layer of anticipation management instead.
Warren: It’s about moving from a purely instantaneous view of trading to one that respects the timing of future information disclosure and risk resolution.
Iris: That’s what we have now regarding this paper "Does the Market Anticipate? Can it? Should it?".
The paper's summary: Ruth: So, to recap, this paper is arguing that even if you have an arbitrage opportunity right now, acting on it immediately might actually be suboptimal because you could be suppressing the anticipation of future risky outcomes.
Warren: Right. The core tension they're setting up is between the idea of no arbitrage and how much information efficiency means when you're dealing with unpredictable risk.
Iris: They introduce a setup where outcomes are disclosed at some point in time, let’s call it tD, which is before the full horizon T. This timing allows them to show that positioning right away for a mispricing isn't always the best move.
Ruth: So, instead of jumping on an arbitrage, you might choose a strategy that helps suppress the anticipation of those predictable risky events that are coming later.
Warren: That’s the status quo bias they call it. It shows up when people stick to their current position because they anticipate something risky will happen soon, even if there's a small price mispricing right now.
Iris: The theoretical framework they build involves model- or event-risk and how risk resolution happens before the full outcome is known. They define viability based on market types, contrasting a market that’s unsure with one that’s sure under more dataflow.
Ruth: And they use this to show that standard continuous-time asset pricing needs to be updated by adding data features, like B specific data and those predictable disclosures at tD much smaller than the total time T.
Warren: The authors develop a Risk Neutral Equivalent pricing approach after adding that risk-specific data, which leads to Proposition one. That proposition basically states that any viable asset price has a very specific form involving an expectation of some quantity minus this reference price term RP hat sub t.
Iris: That formula is the mechanism they use to show how the price dynamic must behave, showing it’s not just about what’s happening right now but also how those future resolutions are factored in.
Ruth: The behavioral implication they draw is pretty direct: for any predictable risk of change, it's suboptimal to long that change unless you have really intense dataflow providing information about it.
Warren: And this means the market-clearing price at any time t isn't just a steady state; it actually changes depending on periods of low risk intensity. That dynamic behavior is what creates that apparent status quo bias where the zero trader fair price at time zero acts like a steady state.
Iris: The conclusion they land on is that viability and efficiency are actually mutually enhancing, but only if you trade anticipation for immediate action. It’s about being efficient by not anticipating future changes too much.
Ruth: So, what this means for us is that if we model markets using these structures, we have to account for the timing of information disclosure and risk resolution rather than just assuming immediate reaction to any price movement.
Warren: If you're trading, you gotta consider the cost of suppressing that anticipation. It’s not just about catching a mispricing; it’s about managing when you get to act on it once the risk resolves itself.
Iris: That leads us into how AI systems could use this concept. They could learn to distinguish between general model- or event-risk and that specific B data by using compound structures that allow for pre-horizon risk resolution, improving how they price uncertain outcomes.
Ruth: That means the AI could optimize strategies that are not immediately optimal for arbitrage but instead suppress the anticipation of those future risky events, effectively delaying action until risk resolution is imminent.
Warren: It helps models account for those behavioral biases we see in markets where people seem to stick with the status quo even when things are changing, because they’re not just reacting to today’s price.
Iris: And they also suggest incorporating a specific price of risk at any time t, which they define as K one half minus K one half plus sigma pi plus t. That term gives us a concrete way to think about the risk component that needs to be priced into those pre-horizon dynamics we discussed.
Ruth: So, we’ve seen how this paper challenges the idea that immediate action on any arbitrage is always best, forcing us to consider anticipation management instead.
Warren: It’s about moving from a purely instantaneous view of trading to one that respects the timing of future information disclosure and risk resolution.
Iris: That’s what we have now regarding this paper, but it leaves us wondering how robust these findings are when the dataflow isn't perfectly structured as they assume.
The paper's improvements: Tom: So, we’re looking at how the authors themselves suggest ways to make this model better than what they presented on page five of that arXiv submission.
Ruth: It seems like they point toward using AI to actually learn how to handle that distinction between different types of risk data. They suggest using compound structures that let the AI see both model- or event-risk and those specific B data features together in a way that allows for pre-horizon risk resolution.
Warren: That makes sense, because right now the model assumes you can cleanly separate those things, but they are not independent in reality. If the AI can learn that dependency, it improves its ability to price uncertain outcomes across different risk types.
Iris: From my econometric perspective, that’s interesting because it addresses the identification issue. It suggests a way to structure the data so you can test if the mechanism is actually working by checking how well it handles these compound structures before drawing conclusions about pricing.
Ruth: Exactly. It moves us away from assuming a perfect separation of risk types and towards letting the data flow dictate how we resolve risk over time, which is what they call pre-horizon risk resolution.
Warren: For a trader, that means the AI could become much better at forecasting not just the current price movement but also *when* that movement will happen relative to when you get your key pieces of future information.
Iris: And it connects back to the behavioral side too, because if the system can better model those predictable disclosures, it might actually lessen that Status Quo Bias by making the timing of action more explicit in the pricing structure.
Ruth: So, this isn't just about adding more numbers; it’s about changing how we structure our models to account for the fact that risk isn't always smooth and continuous across time.
Warren: It means that when we build these trading strategies, they shouldn’t just focus on finding the mispricing; they should focus on optimizing the timing of when you commit capital based on when you expect that risk resolution to actually occur.
Iris: The authors themselves flag a limitation though, which is that this whole setup relies heavily on the dataflow being structured in a very specific way—that predictable disclosure happening at some point tD much earlier than the total time T.
Ruth: Right, so if the real world doesn't have that kind of clean data structure where outcomes are disclosed early, then this specific pre-horizon resolution framework might not be applicable. It’s a constraint on the data we can realistically expect to get from markets.
Warren: So for someone listening who just wants to know what it means: it means sophisticated trading models could start incorporating an "anticipation manager" that decides whether to act now or wait for a clearer signal about future risk resolution.
Iris: It’s a step toward building market models that are more honest about the timing of information, even if we can't perfectly capture every detail of those future disclosures.
Conclusion: Ruth: So we’ve looked at how this paper, "Does the Market Anticipate? Can it? Should it?", suggests that you shouldn't just act on an arbitrage opportunity if you can suppress the anticipation of future risky outcomes.
Warren: Right. It boils down to finding a way to manage the timing of your trade based on when you expect that risk itself to be resolved, instead of acting purely on today's mispricing.
Iris: The main point they drive home is that viability and efficiency are actually things that help each other, but only if you trade away anticipation for immediate action. It’s a trade-off they have to make in their model structure.
Ruth: If you were listening just for the big picture, this paper implies that we need to rethink how we price assets when there's predictable risk of change over time. It’s not just about the current return; it’s about the whole future structure of information release.
Warren: From a trading side, it means that if you try to run a strategy based on immediate arbitrage, you might be fighting against the natural market dynamic which is trying to keep things stable until that future risk resolves itself. The drawdown comes from acting too soon before the resolution window opens.
Iris: I think what changes for someone who just listens is realizing that in any complex market model, you can’t treat all risks as happening instantly; they have a time component built into how they are priced. That structure is what gives the price dynamic its flavor.
Ruth: It does, but it also puts constraints on the data we use. The authors admit that this specific framework relies on outcomes being disclosed at some point tD much earlier than the full horizon T, which isn't always true in real markets. That’s a limitation to keep in mind.
Warren: Yeah, so they’re giving us a sophisticated tool, but it only works if the market data has that specific structure of predictable risk-outcome disclosures ahead of time. It stops working if the information is just random noise without that timing element.
Iris: That’s exactly where future work might go—figuring out how to apply this idea when we're dealing with much less structured, more ambiguous data flows in the real world.
Ruth: Exactly. So, "Does the Market Anticipate? Can it? Should it?" shows us a way to get smarter about anticipating risk resolution in financial models without just throwing away all efficiency for the sake of waiting.
Warren: It’s a nuanced way to look at trading, shifting the focus from just finding an edge now to managing your position based on when you think that edge will actually become real.
Iris: It’s a good exercise in checking if our assumptions about how information flows across time hold up under different types of risk measures.
Ruth: We’ll take this idea about pre-horizon resolution and keep it in mind as we look at the next paper on dynamic star-shaped risk measures. That has some interesting connections there.
More episodes
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- The day in conversation for 2026-11-18
- Daily Summary for 2026-11-18
- 2601.05274-On the use of case estimate and transactional payment data in neural networks for individual loss reserving
- 2009.00868-Loss-Given-Default Modeling by Post-Last Passage Time Process
- 2606.30070-Financial Resilience Evaluation: From Conditional Expectations to Dynamic Convex Risk Measures
- 2212.12687-Measuring price impact and information content of trades in a time-varying setting
- Daily Summary for 2026-11-17
- 2401.08323-Dynamic portfolio selection under generalized disappointment aversion