Confirmation Bias in Research Notes

An analyst writes a bullish research note on a retailer. She believes the company is undervalued. In the weeks that follow, she reads earnings call transcripts with her thesis in mind. When management discusses strong inventory positions, she highlights it as evidence of confidence. When management mentions competitive pressure in one region, she frames it as localized and temporary. The same facts could support a bearish case—inventory strength might signal weak demand, pressure in one region foreshadows broader competitive threats. But her selective interpretation, guided by confirmation bias, sees only affirming evidence.

Confirmation bias is the tendency to seek, interpret, and recall information in ways that confirm a preexisting belief. In equity research, it warps the entire analysis chain: which data sources are consulted, how evidence is weighted, which counterarguments are acknowledged, which are dismissed.

Quick definition: Confirmation bias is the cognitive tendency to prioritize information that supports a preferred hypothesis and minimize or dismiss information that contradicts it.

Key takeaways

• Analysts begin with a thesis (bullish or bearish) and then selectively gather and interpret data to validate it.
• Evidence evaluation becomes asymmetric: bullish evidence receives charitable interpretation, bearish evidence is dismissed as "not fundamental" or temporary.
• Confirmation bias is particularly strong after a position is published or a recommendation is issued; the analyst has reputation stakes.
• Business cycle misread often stem from confirmation bias—analysts committed to a "recovery" narrative see positive signals and ignore negatives, or vice versa.
• "Stonewalling" counterarguments becomes a professional habit: "I know the margin story concerns you, but here's why it doesn't matter." The thesis is preserved.
• Confirmation bias links research quality to analyst humility; analysts who actively seek disconfirming evidence (devil's advocates) write better notes.
• Investors can detect confirmation bias in research by checking whether counterarguments are fairly represented or straw-manned.

How confirmation bias structures the analysis

An analyst decides to initiate coverage of a small-cap tech company with a "buy" recommendation. She believes the market has underappreciated the company's competitive positioning. From that moment, her analysis is shaped by confirmation bias. She reads competitive analyses and notices cases where the company wins; she downplays cases where it loses. She interviews management and hears optimistic forward guidance; she accepts it readily. She interviews a customer and hears moderate satisfaction; she frames it as loyalty, not complacency.

Each data point is processed through a confirmation filter. Information that affirms the buy thesis is integrated into the model with confidence. Information that contradicts it is subjected to skepticism. "That customer issue is a one-timer." "Competitor gained market share this quarter, but long-term we have the better product." "Margin pressure is industry-wide, not specific to our story."

The researcher is not consciously lying. Confirmation bias operates beneath conscious awareness. She genuinely believes she is being analytical, weighing evidence fairly. But the weighting is biased. Over time, the biased weighting accumulates. The research note that emerges represents not a balanced analysis of the company but a thesis-shaped interpretation of the evidence.

Consider a financial analyst covering a bank she rates as a buy. The bank's net-interest-margin (NIM) is compressing due to deposit competition and rising wholesale funding costs. This is material—NIM is 60% of the bank's profitability. A bear would highlight NIM compression as evidence of deteriorating franchise quality. The bullish analyst, committed to her buy rating, frames it differently: "NIM compression is industry-wide and temporary; deposit betas will normalize as the rate-hiking cycle ends." She may be right. But she is also engaging in confirmation bias, selectively accepting an optimistic interpretation of a material headwind because the interpretation confirms her thesis.

Thesis lock-in and recommendation stakes

Confirmation bias intensifies when an analyst publishes a recommendation. A public buy or sell rating creates reputational stakes. The analyst's name is on the research. Downgrading three months later looks like poor judgment or poor market timing. As a result, the analyst becomes unconsciously committed to defending her initial thesis. Disconfirming evidence is interpreted charitably; confirming evidence is seized.

This process is recursive. After publishing a buy rating, the analyst reads quarterly earnings and looks for evidence that supports the buy thesis. She finds some: revenue growth was in line, gross margins expanded slightly. She weights these heavily. She also reads that operating expenses grew faster than expected, but frames it as "growth investment" necessary for the bull case. She attends the quarterly earnings call and, listening for confirmation, highlights every optimistic comment and downplays every cautionary one.

Six months pass. The bullish thesis has not played out. The stock has underperformed. At this point, confirmation bias faces a test. Does the analyst update her view, or does she dig in? Research shows that many analysts dig in, finding new reasons to support the original thesis. The market is wrong. The catalyst is just delayed. A new product launch will drive reacceleration. Confirmation bias, now reinforced by sunk-cost fallacy (the analyst has invested ego in the thesis), becomes even stronger.

The research note from month six differs from the month-zero thesis in surface details but remains fundamentally bullish. Confirmation bias has reoriented the analysis to fit the conclusion.

Selective evidence gathering

One of confirmation bias's most consequential manifestations is selective evidence gathering. An analyst decides a company is a turnaround play. She then contacts suppliers, customers, and industry experts—but she frames questions in ways that invite confirmation. "Do you see this company as a potential leader given its new management?" rather than "What are the long-term competitive barriers in this industry, and does this company have them?" The former question is openin to confirmation, the latter to reality-testing.

She reads industry reports and retains data points that support the turnaround narrative. She ignores or downplays structural industry headwinds that might undermine it. She interviews management and accepts forward guidance readily because management's optimism aligns with her thesis.

An alternative approach—seeking disconfirming evidence actively—would look different. The analyst would ask: "What would have to go wrong for this turnaround to fail?" Then she would search specifically for evidence that failure is occurring. She would interview competitors about their own upside. She would read industry reports searching for evidence that the company's category is in structural decline. She would interview customers with skepticism: "Is the company really winning share, or are you just returning to normal purchasing after a period of under-ordering?" This approach is harder and slower. But it produces more accurate analysis.

The counterargument stonewalling pattern

A tells pattern of confirmation bias in research is how analysts handle counterarguments. A well-reasoned sell-side report should fairly present the bull and bear cases, then explain why the bull case wins. In reality, confirmation-biased research often strawmans the bear case: "Some worry about margin compression, but we believe the operating leverage story is intact." The bear case is not fully developed; it is acknowledged and waved away.

A more honest treatment: "Margin compression is a real risk. The company faces input cost inflation and customer price resistance. In a recessionary scenario, margins could compress 200 basis points, which would reduce EBITDA by $80M. However, we believe the company's pricing power and operating leverage should limit compression to 100bp, supporting our EBITDA estimates." This acknowledges the risk seriously, quantifies it, and explains why you believe it won't fully materialize. It invites disagreement; it does not dismiss it.

Confirmation bias drives analysts toward the strawman. The counterargument is acknowledged but not treated as a serious threat to the thesis because, consciously or unconsciously, the analyst is not interested in evidence that might refute her view.

Business cycle calls and confirmation bias

Few areas show confirmation bias more clearly than business cycle calls. In the late stages of an expansion, optimistic analysts anchor on the strength of recent data and dismiss warnings. "Profit growth is 20%, margins are expanding, valuations are reasonable for the growth rate." All true statements, if you ignore that these are late-cycle markers. The analyst focuses on recent strength and frames it as the new normal. Contrarian warnings about cycle risk are dismissed as pessimism.

Conversely, after a recession has begun, pessimistic analysts anchor on recent weakness. They dismiss early signs of recovery because the pain of the downturn is still vivid. Confirmation bias—seeking evidence that things are worse than consensus believes—causes them to miss the turning point.

Both patterns are driven by confirmation bias interacting with recency bias. Recent data is vivid. The analyst forms a thesis based on recent data. She then selectively interprets new data as confirming that the recent trend will persist.

A disciplined analyst fighting confirmation bias would apply a cycle checklist: "What typically marks the late stage of expansion? High capacity utilization, labor-market tightness, margin compression from wage pressure, credit stress. Do I see these? If I see most, the bull case should be qualified, even if recent growth is strong." This structure resists confirmation bias by making implicit assumptions explicit and testable.

Confirmation bias in model-building

Even the mechanical process of building a financial model can be warped by confirmation bias. An analyst builds a revenue forecast for a manufacturing company. Her thesis is that the company will gain share in a growing market. She models revenue growth at 8% annually. She then builds a margin model. Her thesis is that the company will leverage scale. She models incremental margins at 12%.

These assumptions are not implausible. But they are chosen partly because they confirm her bull thesis. A fresh analyst might model 6% growth and 8% incremental margins, leading to a lower valuation. The first analyst's assumptions are not wrong—they are confirmation-biased. They are chosen to support the bull case.

The bias deepens when the analyst tests sensitivity. She checks: what if growth is 7%? EBITDA falls 15%. What if margins are 10%? EBITDA falls 20%. These sensitivities confirm to her that the thesis is base-case and the downside is downside. She is not biased by asking: what if growth is 9% and margins are 14%? That would confirm a bull thesis. By exploring downside more than upside, she is reifying her base case as the balanced case.

A corrected sensitivity analysis would be symmetric: show equal detail of upside and downside cases, assign them equal credence, and describe what would have to be true for each to occur. This discipline resists confirmation bias.

Facing contrary evidence: the stonewall, the reframe, or the exit

When an analyst encounters evidence that clearly contradicts her thesis, she faces three options: stonewall (deny the evidence's importance), reframe (interpret the evidence differently), or exit (update the thesis). Confirmation bias strongly favors stonewall or reframe.

Example: An analyst has a buy rating on a manufacturer. The company announces a major customer loss. The customer represented 8% of revenue. This is material and negative. The analyst's options:

1. Stonewall: "It's a loss, but we don't think it indicates competitive weakness. The customer was price-sensitive and we chose not to match price."
2. Reframe: "This creates an opportunity to invest in higher-margin products for the remaining customer base. Net-net, EBITDA could be less impacted than revenue suggests."
3. Exit: "We're downgrading to hold. This signals loss of share in a key segment."

Confirmation bias points toward option 1 or 2. The analyst's thesis is that the company has durable competitive advantages. A customer loss contradicts that. But by reframing—the customer was low-margin, the loss is strategic—the analyst preserves the thesis while acknowledging the evidence.

The reframe is sometimes correct. But confirmation bias makes it too attractive. Over time, the analyst uses reframe to dismiss too many disconfirming signals, and the thesis becomes increasingly divorced from reality.

Common mistakes

Mistake 1: Selective evidence gathering. An analyst researches a company to validate a bull thesis. She talks to customers and management but not competitors. She reads industry reports that support sector growth but skips reports questioning structural tailwinds. The evidence base is confirmation-biased from the start.

Mistake 2: Asymmetric burden of proof. Bull-case evidence is accepted readily; bear-case evidence requires ironclad proof. A management assertion of "best-in-class operations" is taken at face value. But a competitor's claim of superior technology is dismissed as hype. The standards are different because confirmation bias lowers the bar for confirming evidence.

Mistake 3: Building models to confirm the thesis. Assumptions are chosen to lead to a bullish outcome: high growth, margin expansion, low cost of capital. The analyst does not realize the assumptions are biased because she believes they are justified. But they would be different if the thesis were bear-case.

Mistake 4: Ignoring or downplaying counterarguments. The analyst acknowledges a potential weakness—say, customer concentration—but does not quantify its risk. If one customer is 15% of revenue and churns, EBITDA falls 15%+. But the analyst writes, "concentration is not a concern" without adequately analyzing it.

Mistake 5: Misinterpreting management commentary. Management says, "We expect to maintain guidance despite near-term headwinds." Confirmation bias reads this as "We're confident despite headwinds." Bear-case readers hear, "We can't commit to growth because headwinds are real." Confirmation bias chosen the charitable interpretation.

FAQ

How can an analyst detect her own confirmation bias?

Ask: If I believed the opposite thesis, what evidence would I highlight? Then check whether the bullish or bearish evidence you presented is balanced. Also, when you revise a recommendation, quantify how much your assumptions changed. If you upgrade from hold to buy with only modest assumption changes, confirmation bias may be at work—you are reinterpreting facts rather than updating beliefs.

Is confirmation bias worse for bull or bear cases?

Both suffer equally. Bull-case analysts ignore deterioration and reframe negative data. Bear-case analysts ignore improvement and assume the cycle will inevitably turn. The bias is direction-agnostic; it serves whatever thesis the analyst adopts.

Can structured frameworks prevent confirmation bias?

They can reduce it. Checklists, pre-mortems, and explicit counterargument development are helpful. So is requiring analysts to assign a probability to each assumption and update that probability as evidence arrives. But frameworks cannot eliminate confirmation bias entirely because the bias operates partially beneath conscious awareness.

Should analysts avoid having a thesis?

Not avoiding a thesis is impractical and undesirable. A thesis focuses analysis and decision-making. But analysts should treat the thesis as provisional, not sacred. Assign it a probability. Actively search for evidence that would lower that probability. Commit in advance to downgrading if specific events occur, then honor that commitment.

Does confirmation bias make analyst research worthless?

No, but it means analyst research should be consumed with skepticism. Look for: (1) fair representation of bear and bull cases; (2) explicit assumptions clearly stated; (3) symmetrical sensitivity analysis; (4) evidence that the analyst has engaged with contrary viewpoints. Research that exhibits these traits is more reliable.

How should investors respond to confirmation-biased research?

Seek research from multiple analysts with opposite theses. Read short-seller reports and bull-case notes side-by-side; they confirm different biases. Build your own model from first principles. Interview multiple customers and competitors yourself rather than relying on the analyst's selective quotes. Treat analyst research as input, not gospel.

Can an analyst be bullish and still avoid confirmation bias?

Yes. A bullish analyst who actively seeks evidence against her thesis, who fairly represents the bear case with quantification, who updates assumptions when evidence warrants, is fighting confirmation bias. Bullishness plus intellectual humility is possible. The problem arises when bullishness combines with overconfidence and dismissal of contrary views.

Real-world examples

Enron and analyst confirmation bias, 1990s–2001. Sell-side analysts were bullish on Enron for years despite warning signs. Confirmation bias played a clear role. Analysts were committed to Enron as a "best-of-breed" energy company. When executives boasted of off-balance-sheet structures, analysts accepted the explanations. When the accounting became questionable, analysts trusted management. When insiders were selling shares, it was dismissed as "tax planning." Confirmation bias defended the bull thesis until it collapsed.

Facebook's "it's about engagement, not revenue" era. In the early-to-mid 2010s, bullish analysts on Facebook faced a thesis challenge: the company wasn't monetizing mobile traffic quickly. Rather than downgrade, many reframed the narrative: "Engagement is the key metric; monetization will follow." This reframe (confirmation bias defending the thesis) turned out to be right, but it delayed realistic assessment of risks.

Intel's process leadership thesis, 2015–2021. Analysts were committed to a thesis that Intel would maintain process leadership over TSMC (Taiwan Semiconductor). As TSMC gained ground, analysts reframed: "Intel still has advantages in overall solution, power efficiency, and customer relationships." The thesis was defended through reframe until Intel's public admission that it had lost process leadership became unavoidable in 2021.

Related concepts

• Narrative fallacy: The tendency to construct coherent stories from random or incomplete data, then accept the story as explaining reality. Confirmation bias is a mechanism that enables narrative fallacy.
• Belief perseverance: Once a belief is formed, it resists change even in the face of contradictory evidence. Confirmation bias is a driver of belief perseverance.
• Backfire effect: In some cases, exposure to contradictory evidence strengthens a person's original belief. This can occur when confirmation bias is very strong.
• Motivated reasoning: Using logic selectively to arrive at a predetermined conclusion. Confirmation bias enables motivated reasoning.
• Hindsight bias: After an outcome is known, the tendency to believe you knew it would happen. Related to confirmation bias in that both distort memory and interpretation of evidence.

Summary

Confirmation bias is perhaps the most pervasive distortion in fundamental analysis. Once an analyst adopts a thesis—bullish or bearish—she unconsciously orients all subsequent analysis toward confirming it. She gathers evidence selectively, interprets ambiguous data to align with her view, and frames counterarguments as irrelevant. The bias is strongest after a public recommendation, when the analyst's reputation is at stake. It manifests across the entire chain of analysis: evidence selection, assumption-building, sensitivity testing, and interpretation of incoming data.

Investors and fellow analysts can detect confirmation bias by checking whether research presents bull and bear cases fairly, whether assumptions are explicit and justified, whether sensitivity analyses are symmetric, and whether the analyst actively engages with contrary viewpoints or dismisses them. The strongest defense against confirmation bias is intellectual humility: treating your thesis as provisional, assigning it a probability rather than certainty, and actively seeking evidence that would lower that probability. Analysts who fight confirmation bias write better research and make better decisions.

Next

Even with disciplined analysis and emotional distance from the thesis, analysts overestimate the precision and reliability of their models: Overconfidence in DCF models