The Reflexive Market

The most developed framework for understanding how belief shapes markets is George Soros’s theory of reflexivity, articulated in The Alchemy of Finance (1987) and refined over four decades of subsequent work, including his peer-reviewed treatment in the Journal of Economic Methodology (Soros, 2013).

The starting point is the observation that participants in financial markets are not passive observers. They form expectations about future market behaviour, and they trade on those expectations. Their trading actually shapes the prices they were trying to anticipate. This creates a feedback loop with four components.

First, participants form beliefs about market fundamentals. Second, those beliefs drive trading behaviour. Third, that trading behaviour moves prices. Fourth, the moved prices themselves become evidence that participants use to form new beliefs.

Consider a concrete example. In the late 1990s, technology stocks rose rapidly. The rising prices reinforced the narrative that “internet companies are revolutionary.” That narrative drew more investors in, who bought, which drove prices higher. The narrative seemed validated by price action. More investors arrived. The cycle continued until the underlying fundamentals could no longer support the prices, at which point the cycle reversed: falling prices undermined the narrative, which led to selling, which drove prices lower.

This is not just a story of irrationality. It is a structural property of markets. The market’s price reflects collective belief, but collective belief is partly formed by the market’s price. The two are entangled. They cannot be separated cleanly.

Soros calls this the human uncertainty principle: the cognitive function (forming beliefs about reality) and the manipulative function (acting on those beliefs and changing reality) cannot both be performed independently. Participants who think they are simply analysing markets are also, by their analysis and trading, partly creating the markets they analyse.

The framework’s implication is that any analysis treating fundamentals as the sole driver of prices will miss systematic patterns. Sometimes prices reflect fundamentals reasonably well. Sometimes they diverge from fundamentals in ways that the reflexive feedback can explain. Knowing which condition currently holds is itself a critical analytical question.

Reflexivity describes the structural feedback. Narrative economics describes what specifically propagates through that feedback.

Robert Shiller, working in parallel with Soros over decades, developed a framework for understanding how stories about markets spread and influence behaviour. His earlier work, Irrational Exuberance (2000/2015), documented how successive market bubbles were accompanied by specific narratives that gained widespread acceptance, then faded. His more recent Narrative Economics (2019) treats this systematically: narratives, like viruses, spread through populations according to identifiable patterns, and they exert real economic effects.

What is a market narrative? It is a compact story that explains why some development matters. “AI will transform every industry.” “Crypto is the future of money.” “Inflation is transitory.” “Houses always appreciate.” These are not predictions in the technical sense. They are framings that organise how participants interpret incoming information.

Narratives matter because participants do not respond to data directly. They respond to data filtered through the narratives they currently hold. A modest disappointment in earnings can crash a stock when the prevailing narrative is “this company can do no wrong.” The same disappointment may be ignored when the prevailing narrative is “this company has long-term challenges, but management is steady.” The data is the same. The narrative determines its meaning.

Shiller’s empirical work shows that narratives spread according to dynamics borrowed from epidemiology: they have transmission rates, recovery rates, and equilibrium spreads. Some narratives reach near-universal acceptance before fading. Others persist as background context for years. Understanding which narratives are currently dominant and how they are evolving is a research question, not a side observation.

For markets, this has direct implications.

First, price action that seems disconnected from fundamentals often has a narrative explanation. The market did not ignore the data; it interpreted the data through a frame that gave it different meaning.

Second, turning points in markets are often preceded by turning points in narratives. The market peaks when the prevailing bullish narrative reaches saturation; it bottoms when the prevailing bearish narrative does the same. Tracking narrative cycles can provide leading indicators that pure price analysis cannot.

Third, sentiment indicators capture some of this but not all. A bullish sentiment reading is downstream of a bullish narrative. The narrative is the root; sentiment is the symptom.

The practical analytical concept that emerges from reflexivity and narrative dynamics is expectation divergence: the gap between what participants currently expect and what is likely to actually happen.

This gap is where money is made or lost. A stock priced for perfection is one whose price reflects optimistic expectations across many dimensions: earnings growth, margin expansion, market share, management quality. If reality matches expectations, the stock is fairly priced. If reality exceeds expectations, the stock should rise. If reality falls short, the stock should fall, sometimes dramatically, even when reality is still good by absolute standards.

The same principle applies in reverse. A stock priced for catastrophe reflects pessimistic expectations. If reality matches, the stock is fairly priced. If reality is merely bad, the stock can rally significantly because the gap between bad and catastrophic is positive in expectation terms.

This is the analytical core of value investing as practised by serious practitioners: it is not about buying cheap stocks. It is about identifying situations where market expectations and likely reality diverge in your favour. The cheap stock that is cheap because expectations correctly anticipate its problems is not a value opportunity. The expensive stock whose expectations underestimate its growth is.

For the foundation’s framework, expectation divergence is a central concept because it links the otherwise abstract notion of expectations to a measurable analytical question: what does the price imply about expected outcomes, and how does that compare to the best available estimate of likely outcomes?

Reading expectations from prices requires reverse-engineering. Discounted cash flow models can be inverted: given the current price, what growth rate must investors be assuming? Given the current multiple, what assumptions about future earnings are embedded? These are valuation regime questions. The foundation’s research does not claim to do this work better than dedicated equity analysts. It uses these techniques where appropriate to characterise the implicit expectations embedded in market prices, then asks whether those expectations are consistent with reality as the foundation’s other layers, structure, cycles, and broader narrative analysis, read it.

What does reading expectations look like in practice?

Several sources of evidence feed into this work, each with their own strengths and limitations.

Sentiment indicators attempt to measure investor positioning and emotional state directly. AAII surveys, put/call ratios, investor confidence indexes, fund manager positioning surveys. These data are useful but limited. They capture stated beliefs (which may differ from acted-on beliefs) and they often produce signals that are coincident or lagging rather than leading.

Positioning data attempts to measure what investors are actually doing rather than what they say. Futures positioning (Commitments of Traders reports), options positioning, ETF flows, margin debt levels. This data is more meaningful than survey sentiment but is also harder to interpret. Extreme positioning often precedes turns, but extreme is itself a moving target.

Narrative analysis tracks the stories currently dominant in markets. This is harder to measure but in some ways the most valuable. It can be done qualitatively by tracking financial media, social media, conference themes, and analyst language. Quantitative natural- language methods exist but are still developing. The foundation’s approach combines qualitative tracking with attention to specific markers: when a narrative reaches near-universal acceptance, when contrarian voices begin emerging, when the prevailing frame stops explaining incoming data well.

Implied probability extraction uses derivative markets to back out market-implied probabilities of future outcomes. Options prices contain information about implied volatility and skew that reveals what the market is collectively pricing as likely. This is a quantitative window into expectations that is not available through surveys.

The discipline of integrating these sources is significant. None alone is sufficient. Sentiment surveys can be at extremes for weeks before a turn. Positioning data can be misread. Narratives are slippery. Implied probabilities reflect market participants who may themselves be wrong.

What the foundation does is read these multiple sources through each other, a diffractive reading in the terms of Concept I, looking for confluence: when sentiment is extreme and positioning is stretched and the dominant narrative is being challenged by recent events and implied probabilities are pricing the consensus view as near-certain, the conditions for a turn are present even if the timing remains uncertain.

The expectations layer, like every other in the framework, has limits the foundation takes seriously.

First, sentiment and positioning data are notoriously prone to false signals. Extreme readings can persist for extended periods before resolving. A sentiment reading is not a timing signal. The foundation’s discipline includes treating sentiment as context, not as trigger.

Second, narratives are subjective. Identifying the dominant narrative requires judgment, and different analysts reading the same media environment may identify different narratives. The foundation acknowledges this and tries to characterise narratives concretely, with specific framings and specific assumptions, rather than vaguely as just bullish or bearish.

Third, expectation divergence requires an estimate of likely reality against which expectations are compared. This estimate is itself uncertain. Foundation work cannot claim privileged access to fundamental truth. It can only contrast market-implied expectations with the best available analysis of likely outcomes, and acknowledge the second-order uncertainty in that comparison.

Fourth, the reflexive feedback can persist longer than fundamental analysis would suggest is justified. “The market can stay irrational longer than you can stay solvent” is Keynes’s famous warning. The framework respects this. Identifying that expectations diverge from likely reality does not mean immediate convergence; it means the conditions for eventual convergence are present, possibly without specific timing.

Fifth, narrative analysis can degenerate into projection. Analysts may identify as the dominant narrative what is actually their own framing of events. The discipline requires careful attention to evidence: what specific language is being used in widely-read sources? What specific assumptions are being made? Without this empirical anchor, narrative analysis becomes assertion.

The expectations layer is not optional in serious market analysis. The structural and cyclical layers describe the forms within which markets move. The probability layer describes how to weight scenarios. But the actual drivers of price action over short and medium horizons are very often expectations and their shifts, not fundamentals directly.

This is the empirical observation behind reflexivity: that markets are not just measuring something external, but also producing what they appear to measure through the feedback between belief and price. Ignoring this is to systematically miss what moves markets.

The foundation’s commitment to the expectations layer follows from this observation. It does not claim to predict where prices will go from sentiment alone. It claims, more modestly, that any analysis ignoring the expectations dynamic will be incomplete. The framework integrates expectations with structure, cycles, and probability into a more complete reading.

The work continues. Narrative analysis methods are still developing. New sources of expectations data emerge each year as financial markets generate more granular information about positioning and belief. What endures is the underlying recognition: markets are partly stories that participants tell themselves, and those stories shape what comes next. Reading the stories carefully, alongside the other analytical layers, is one of the things the foundation does.