Confirmation Bias: Testing the Case Against Your Thesis
Test competing explanations before turning a useful fact into a convenient story.
Start Investing ClearlyBeginner4 min
Test the explanation
Plain English
A true fact can support an incomplete story
Confirmation bias can appear without anyone inventing a fact. A researcher may search harder for supportive material, interpret ambiguous evidence favourably or demand stronger proof from a contrary argument. The resulting story can contain accurate observations while giving them uneven treatment. The problem is the process of evaluation, not simply whether one sentence is true.
An investment thesis is a reasoned claim with assumptions and evidence. It is reasonable to have a provisional view; research would be difficult without questions or expectations. The useful distinction is between a view that can be tested and a view that absorbs every observation as confirmation. A stated condition that could change the interpretation makes the difference easier to see.
Classic research by Lord, Ross and Lepper examined how people with opposing views evaluated mixed evidence about capital punishment. It supports discussion of asymmetric scrutiny, but it was not a stock-selection experiment. This article uses an original fictional company example and makes no claim that a checklist or a change of research habit improves investment returns.
Worked Example
Revenue rises while unit sales fall
A fictional company sells 100 identical units at £10 each in one period. In the next comparable period it sells 95 units at £12 each. Assume every unit is sold at the stated price, with no returns, discounts, product-mix changes or currency effects. Revenue rises from £1,000 to £1,140, an increase of 14%.
One interpretation is that the company can charge more while retaining most sales. Another is that the higher price is accompanied by weaker demand that could become more pronounced. Both acknowledge the same revenue arithmetic. Neither is established by it. The example reports realised prices and volumes but does not identify why customers bought fewer units or whether the pattern will continue.
Price rises 20% and volume falls 5%; multiplying 1.20 by 0.95 gives 1.14. Adding the percentages would overstate the revenue change because the higher price applies to fewer units. The calculation says nothing about costs. Profit could rise or fall depending on production, distribution and other expenses that have deliberately been left unspecified.
First-period price per unit
£10
Second-period price per unit
£12
First-period units
100
Second-period units
95
First-period revenue
£1,000
Second-period revenue
£1,140
Revenue growth
14%
Testing Explanations
Test explanations with observations that distinguish them
A useful next question is what evidence would help distinguish durable pricing power from a temporary response or deteriorating demand. Comparable repeat purchases, customer retention, competitor prices and unit costs could bear on different parts of those explanations. Their usefulness depends on measurement quality and timing; naming more data does not automatically resolve the uncertainty.
Record what each observation supports, what it challenges and what remains unresolved. For example, a revenue increase supports the narrow statement that sales value rose under the example’s assumptions. It does not alone support higher profit or future customer loyalty. Keeping those claims separate prevents a small verified fact from silently expanding into a broad conclusion.
Evaluating Evidence
Apply the same scrutiny in both directions
Challenging a thesis does not mean automatically believing its opposite. A negative anecdote can be unrepresentative, just as a positive one can. Examine sample size, source incentives, dates and alternative causes consistently. Asking what would change a view is useful only if the answer is specific enough that later evidence can actually meet it.
A dated research note can preserve the original assumption and the reason for a revision. That makes changes in reasoning inspectable, without implying that consistency itself is desirable or that every new observation requires a reversal. The purpose is to connect conclusions to evidence and uncertainty, rather than to reward confidence or defend an earlier statement.
Common Mistake
Counting only the evidence that agrees
A list with ten supportive items and one challenge does not establish the balance of the case. Several items may repeat the same underlying source, and one well-supported contradiction may matter more than many weak confirmations. Judge relevance and independence before counting observations. The goal is a testable explanation, not a larger collection of agreeable notes.
Self-check
Check your understanding
Does 14% revenue growth prove pricing power?
No. It follows from the stated prices and volumes, but other explanations and future demand remain unresolved.
Does lower volume prove the higher price failed?
No. Costs, customer behaviour and alternative causes are unspecified; the opposing conclusion also needs evidence.
Why preserve the original dated assumption?
It allows later readers to see what changed and why, instead of rewriting the earlier view to make new evidence appear inevitable.
Continue learning
Connect the ideas
Follow the related articles below to examine these assumptions in another setting.
Disclaimer
Educational Use Only
This article is for informational and educational purposes only. It does not provide personalised investment advice.
