How to Detect Confirmation Bias in an Investment Thesis
BIZ CPAs · BEHAVIORAL FINANCE SERIES 01
Conviction is necessary. The discipline is keeping conviction open to evidence that could prove it wrong.
Once an investor buys a security, a subtle change often occurs: research that began as an open question becomes a defense of a decision already made. Positive developments feel more relevant. Contradictory evidence appears temporary, misunderstood, or less credible. The portfolio position may be unchanged, but the decision process has changed.
This is confirmation bias. It does not mean the investment thesis is necessarily wrong. It means the method used to evaluate the thesis may no longer be neutral. For an individual investor, family, investment committee, or advisor, that distinction matters because a sound investment program should be capable of reaching the same decision standard repeatedly—especially when new information is uncomfortable.
A repeatable investment process is not designed to eliminate judgment. It is designed to prevent judgment from quietly changing the rules after capital is committed.
A formal definition
Confirmation bias is a cognitive tendency to seek, interpret, emphasize, and remember information in a way that supports an existing belief while giving insufficient weight to evidence that contradicts it. In the CFA Institute behavioral-finance framework, confirmation bias is classified as a belief-perseverance bias: an investor maintains a prior view through errors in information selection or processing.
The bias can appear at several stages. An investor may choose only supportive sources, interpret ambiguous facts in the thesis’s favor, demand a higher standard of proof from bearish evidence, or remember the correct parts of an earlier forecast while forgetting the assumptions that failed. The common thread is asymmetric treatment of evidence.
Confirmation bias is different from having conviction. Conviction follows a conclusion supported by evidence and remains conditional on the facts. Confirmation bias protects the conclusion from the evidence. A fiduciary-oriented process must preserve that distinction because the purpose of research is to improve the decision made for the investor—not to protect the analyst’s pride, reputation, or prior recommendation.
Why intelligent investors are still vulnerable
Expertise does not immunize an investor from bias. In some cases, greater knowledge provides more material with which to build a sophisticated defense of the original position. The analyst may generate plausible explanations for every disappointment while continuously postponing the moment when the thesis must be reconsidered.
Three conditions make confirmation bias particularly dangerous:
Ambiguity. Business outcomes have many possible explanations. When revenue slows, management may cite seasonality, execution, competition, pricing, the economy, or customer timing. Ambiguity gives the mind room to select the explanation it prefers.
Identity. A position can become associated with the investor’s intelligence or professional credibility. Changing the thesis then feels like admitting personal failure instead of updating a forecast.
Commitment. The larger the financial, reputational, or emotional commitment, the stronger the temptation to defend the original decision. Concentrated positions can therefore create both portfolio risk and decision-making risk.
Two practical examples
EXAMPLE ONE · THE GROWTH COMPANY
An investor buys a software company because customer growth is strong, recurring revenue is expanding, and management expects operating leverage. Over the next three quarters, revenue continues to grow, but customer acquisition costs rise, renewal rates weaken, stock-based compensation increases, and free-cash-flow conversion falls.
A confirmation-biased review highlights the revenue growth and dismisses the other changes as temporary investment for the future. A disciplined review asks whether the original economics are still intact. Was the thesis about growth at any cost, or profitable growth with improving retention and cash conversion? If the latter, several critical assumptions may have deteriorated even though the headline revenue number remains positive.
EXAMPLE TWO · THE FAMILIAR DIVIDEND STOCK
A long-time shareholder owns a company known for paying and increasing its dividend. The investor interprets the dividend record as evidence of safety. Meanwhile, leverage rises, interest coverage declines, the payout ratio expands, and the company funds part of its distribution through asset sales or new borrowing.
A confirmation-biased investor searches for articles praising the dividend history and treats the current yield as an opportunity. A more complete process separates past policy from present capacity. The relevant question is not whether management values the dividend; it is whether future cash generation and the balance sheet can support it without impairing the enterprise.
The strongest evidence for a position may explain why it was purchased. The strongest evidence against it may determine whether it should still be owned.
How confirmation bias damages a repeatable investment process
An investment program becomes repeatable when the same architecture governs idea generation, research, valuation, portfolio construction, monitoring, and exit decisions. Confirmation bias is an impediment because it introduces an undocumented exception: evidence is treated differently depending on whether it supports an existing position.
Process stage | What bias changes | Why repeatability suffers |
Research | Supportive sources receive attention; disconfirming sources receive less scrutiny. | The information set changes according to the desired conclusion. |
Forecasting | Positive assumptions are updated quickly while negative evidence is labeled temporary. | Forecast rules become asymmetric and difficult to audit. |
Valuation | The investor adjusts the multiple, terminal assumptions, or scenario weights to preserve upside. | Valuation becomes a defense of price rather than an estimate of value. |
Monitoring | Good news confirms skill; bad news is explained away or given more time. | Review thresholds move after the fact. |
Selling | Thesis-breakers are redefined, delayed, or ignored. | Capital remains committed under standards that were never approved. |
Learning | Memory preserves the correct elements of the thesis and minimizes failed assumptions. | The feedback loop cannot improve because errors are not accurately recorded. |
The most reliable controls are procedural
Awareness is useful, but simply telling oneself to be objective is rarely enough. Controls work better when they are defined before the position is emotionally or financially important. The goal is not to create permanent indecision. It is to make conviction earn the right to persist.
1 · Start with a question, not a conclusion
Frame the work as “What must be true for this security to offer an attractive risk-adjusted return?” Avoid beginning with “Why this stock is a buy.” The first formulation invites testing; the second invites advocacy.
2 · Write the thesis before or at initiation
Record the expected return drivers, critical operating factors, valuation assumptions, time horizon, catalysts, material risks, and evidence that would weaken or invalidate the thesis. Date the document. Do not rewrite the original after the outcome is known; add a new version and preserve the prior one.
3 · Define disconfirming evidence in advance
For each critical factor, identify what adverse observation would matter and over what period. Examples might include renewal rates below a stated level, leverage above a stated range, failure to reach a regulatory milestone, or a competitive response that compresses unit economics.
4 · Build the strongest credible bear case
Do not create a weak opposing argument merely to check a box. Use the most capable skeptical source available: a lower forecast, a competitor, an industry specialist, a customer complaint pattern, a short thesis, or an independent colleague. Then state what the bear case gets right.
5 · Use scenarios without changing the arithmetic
Maintain base, upside, and downside cases with explicit probabilities. When facts change, update the affected operating assumptions first and allow valuation to respond. Avoid changing the discount rate, multiple, or scenario weights solely to recover the preferred target price.
6 · Separate thesis review from price movement
A falling price is not proof the thesis is wrong, and a rising price is not proof it is right. Review business evidence, valuation, and market price as related but distinct inputs. This reduces the tendency to use price action as selective validation.
7 · Require independent challenge for material decisions
For concentrated positions or major allocation changes, assign someone to test the thesis. The challenger should identify omitted evidence, fragile assumptions, correlated risks, and the strongest reason the expected return may be overstated. The decision owner remains accountable for the final judgment.
8 · Conduct a decision post-mortem
Compare the original thesis with what occurred. Distinguish process quality from outcome quality: a sound process can produce a loss, and a weak process can produce a gain. Record forecasting errors, information that was missed, and rules that should change before the next decision.
A five-question bias check before the next decision
Before buying more, holding through a material change, or declining to sell, an investor or committee should be able to answer five questions in writing:
What is the strongest current evidence that our thesis is correct?
What is the strongest current evidence that our thesis is wrong?
Which original assumption has changed the most, and did we update the model consistently?
What would we conclude if we did not already own the position?
What action does the pre-established decision rule require—hold, resize, exit, or investigate further?
If the second question is difficult to answer, the research is probably incomplete. If the fourth answer differs materially from the current action, ownership may be influencing judgment. If the fifth question has no answer, the investment program needs a clearer monitoring and sell discipline.
Technology can accelerate the same human bias
Search engines, social-media feeds, financial platforms, and AI tools can make research faster while also making selective research easier. An investor who asks only bullish questions will often receive an increasingly coherent bullish narrative. Speed and polish do not make the information set balanced.
Recent CFA Institute Research and Policy Center work on AI-supported investment workflows found that explicitly presenting both positive and negative perspectives, together with human oversight, was more effective at reducing tested framing effects than merely instructing a model to avoid bias. Framing bias is distinct from confirmation bias, but the process lesson carries over: balanced inputs, defined analytical steps, and accountable review are stronger safeguards than a general request to “be objective.”
What a fiduciary posture requires
A fiduciary posture is not demonstrated by never changing one’s mind. It is demonstrated by placing the investor’s objectives and circumstances ahead of the comfort of defending a prior view. That requires documentation, consistency, appropriate diversification, attention to risk capacity and liquidity, and a willingness to act when the evidence no longer supports the original allocation.
The practical standard is not perfect objectivity; no person or process can guarantee that. The standard is a decision system that makes bias visible, creates a credible challenge, preserves an audit trail, and applies the same rules to favorable and unfavorable evidence.
Good investing is not the absence of conviction. It is conviction that remains accountable to disconfirming evidence.
Next in the series
ARTICLE TWO
When conviction becomes certainty
The next article will examine two related biases from the CFA behavioral-finance framework: overconfidence and illusion of control. Overconfidence can lead investors to overestimate the accuracy of their forecasts or the quality of their information. Illusion of control can lead them to believe that greater involvement, more data, or frequent trading gives them more influence over outcomes than they actually possess. Together, these biases can encourage concentrated positions, narrow scenario ranges, excessive turnover, and insufficient respect for uncertainty.
Future articles can then connect the full decision chain: anchoring and adjustment when an investor remains attached to a prior price target; availability bias when recent or vivid information dominates the evidence set; loss aversion and the endowment effect when selling feels more painful because the position is already owned; and hindsight bias when the past appears more predictable after the outcome is known.
This sequence is intentional. Confirmation bias shapes what we look for. Overconfidence shapes how certain we feel. Anchoring shapes how slowly estimates change. Loss aversion shapes whether we act. Hindsight bias shapes what we believe we learned. A disciplined investment program must manage the entire chain.
Sources and further reading
Educational disclaimer
This material is provided solely for general educational and informational purposes. It does not constitute individualized investment advice, an offer or solicitation, or a recommendation to purchase, sell, or hold any security or adopt any investment strategy. Investment decisions involve risk, including possible loss of principal, and should be evaluated in light of the investor’s objectives, time horizon, liquidity needs, tax circumstances, risk tolerance, and risk capacity. References to CFA Institute materials describe educational concepts and do not imply endorsement of BIZ CPAs by CFA Institute.
Comments