CompanyGraph
Recent experience dominates evaluation even when longer-term base rates suggest different expectations, creating systematic extrapolation of current conditions that conflicts with the tendency of metrics to revert toward historical averages.
Why recency bias and mean reversion collide most forcefully at exactly the moments when expectations are most distorted.
Introduction
Ask someone who has experienced three consecutive years of strong market returns what they expect from the next year, and the answer will be influenced by those recent returns more than by the full historical distribution. Ask the same question after three years of poor returns, and the expectation shifts accordingly. This is recency bias: the structural tendency to weight recent experience disproportionately when forming expectations about the future.
This bias exists in tension with a widely observed statistical property: mean reversion. Many financial and business metrics tend, over time, to move back toward their long-term averages. High profit margins attract competition and tend to moderate. Low profit margins trigger restructuring and tend to improve. Extreme growth rates normalize. Extreme valuations adjust. The tendency is not mechanical or guaranteed, but it is persistent enough across many metrics and time periods to constitute a structural feature of the data.
The interaction between these two phenomena creates a characteristic pattern. Recency bias leads participants to extrapolate current conditions precisely when mean reversion makes current conditions least likely to persist. After periods of extreme performance in either direction, expectations tend to be most misaligned with the structural tendencies of the system.
Core Concept
Recency bias reflects a real property of the environment: recent information is often more relevant than distant information. Conditions change, and older data may reflect structural conditions that no longer exist. The problem is not that recent data receives attention but that it receives disproportionate attention relative to its actual informational advantage over longer-term data. The weight given to recent experience exceeds what a careful assessment of relevance would assign.
The mechanism operates through availability and vividness. Recent experiences are more easily recalled, more emotionally salient, and more vivid than distant ones. Three years of strong returns are not just data points; they are lived experience accompanied by specific memories, emotions, and narratives. These experiential qualities make recent data psychologically weightier regardless of its statistical informativeness.
Mean reversion, by contrast, is a statistical tendency that operates across populations and over time. It is driven by identifiable structural forces: competition erodes abnormal profits, demand responds to abnormal pricing, management changes follow abnormal performance, capital flows toward abnormal returns and away from abnormal losses. These forces do not operate on predictable timelines, and exceptions exist, but the tendency is robust across many domains of financial and business data.
The collision between recency bias and mean reversion creates a structural dynamic. At the moment when conditions are most extreme, and thus most likely to revert, expectations formed by recency bias are most strongly extrapolating those extreme conditions. High margins lead to expectations of continued high margins precisely when competitive forces are most likely to compress them. Depressed valuations lead to expectations of continued depression precisely when the conditions for recovery may be accumulating.
Structural Patterns
- Extrapolation at Extremes — At the peak of strong performance, expectations for continued strength are highest. At the trough of weak performance, expectations for continued weakness are highest. This pattern consistently places consensus expectations at maximum divergence from mean-reverting tendencies.
- Margin Expectations — Companies experiencing historically high profit margins attract estimates that project continuation. Structural forces that tend to compress margins, including competition, customer pushback, and input cost cycles, receive less weight in analysis because the recent experience of high margins is more salient.
- Growth Rate Anchoring to Recent Periods — Recent growth rates serve as the baseline for projections even when base effects, market saturation, or competitive entry make those rates structurally difficult to sustain. The more impressive the recent growth, the higher the projected growth, and the greater the gap between expectation and structural probability.
- Cyclical Misinterpretation — Cyclical peaks are mistaken for structural levels because the recent experience of peak conditions dominates evaluation. Cyclical troughs are similarly treated as permanent conditions. The structural property of cyclicality itself is underweighted relative to the most recent phase of the cycle.
- Risk Perception Cycles — After extended periods of stability, risk perception decreases. After periods of volatility, risk perception increases. Both assessments are driven by recent experience rather than structural analysis of the conditions that produced the recent experience.
- Capital Allocation Timing — Capital tends to flow toward sectors, strategies, and asset classes with strong recent performance and away from those with weak recent performance. This pattern directs capital toward areas where mean reversion is more likely and away from areas where mean reversion would be beneficial.
Examples
Corporate earnings cycles demonstrate the pattern clearly. During economic expansions, corporate profits tend to grow and margins tend to expand. Analyst estimates during these periods project continued improvement. When the cycle turns, the adjustments come gradually as each quarter's results reset expectations. The base rate of earnings cyclicality is well established, yet each cycle produces a period where recent strong earnings lead to estimates that extrapolate those earnings as a new normal.
Commodity markets illustrate recency bias in pricing. During supply shortages, prices rise sharply, and forecasts project continued high prices. During periods of oversupply, prices decline, and forecasts project continued low prices. The structural tendency of commodity prices to mean-revert is driven by supply response: high prices incentivize production that eventually creates oversupply, and low prices curtail production that eventually creates shortage. This cycle is structural and well-understood, yet recent price levels consistently dominate expectations.
Housing markets provide a long-cycle example. During periods of sustained price appreciation, expectations of continued appreciation strengthen, often supported by narratives about structural changes in demand, supply constraints, or demographic shifts. These narratives may contain genuine structural elements, but the extrapolative expectations they support tend to overshoot what those structural elements justify. The same dynamic operates in reverse during sustained declines, where structural narratives about excess supply or demand weakness support expectations of continued decline.
Risks and Misunderstandings
The most important misunderstanding is that mean reversion is a timing tool. Observing that a metric is extreme relative to its historical range tells us that reversion is structurally more likely, not when it will occur. Extreme conditions can persist for extended periods, and the forces driving reversion operate on unpredictable timelines. Mean reversion is a statistical tendency, not a prediction of imminent reversal.
Another common error is assuming that all metrics mean-revert. Some changes are genuinely structural rather than cyclical. A company that achieves a structural competitive advantage may sustain margins that appear extreme relative to historical averages without reverting. Distinguishing structural change from cyclical extremes requires analysis that goes beyond the statistical pattern itself. The tendency to mean-revert is a default assumption that specific structural analysis can override.
It is also tempting to believe that quantitative awareness corrects recency bias. Knowing the historical average of a metric provides an anchor against pure extrapolation. But the psychological weight of recent experience is not easily offset by historical statistics. Building processes that systematically incorporate base rates alongside recent trends is more effective than relying on individual judgment to balance the two.
What Investors Can Learn
- Compare current conditions to historical ranges — Placing current metrics in historical context reveals how extreme they are. Extreme values are not guaranteed to revert but are structurally more likely to moderate than to extend further.
- Question the permanence of recent trends — When analysis relies heavily on the continuation of recent conditions, consider what structural forces exist that would tend to moderate those conditions over time.
- Recognize cyclicality — Many financial and business metrics are cyclical. Understanding where in the cycle current conditions sit provides structural context that pure trend extrapolation misses.
- Distinguish structural change from cyclical extremes — Some deviations from historical norms reflect genuine structural change. Others reflect cyclical extremes that will revert. The distinction matters enormously and requires analysis beyond pattern recognition.
- Consider base rates alongside recent data — Long-term averages and distributions provide information that recent experience alone does not. Neither dominates; both contribute to a more complete picture than either provides alone.
Connection to CompanyGraph's Philosophy
The tension between recency bias and mean reversion is a structural feature of how human evaluation interacts with statistical tendencies in financial data. Describing this interaction, including its limits and exceptions, provides context for interpreting market behavior without predicting specific outcomes. This kind of structural observation, attentive to both human and statistical patterns, reflects CompanyGraph's approach to understanding what is rather than forecasting what will be.