Behavioural biases in forecasting

Behavioural biases in forecasting

Published September 28, 2026

Behavioural biases in forecasting

On this page

Learning Outcome Statement

Explain how behavioral factors affect analyst forecasts and recommend remedial actions for analyst biases.

On the exam: You get a short analyst-behavior vignette and must identify the bias, state its forecast effect, or choose a remedial action, usually as a multiple-choice question or a bias-to-remedy matching grid.

Forecasting is not just a search for facts. The analyst’s process decides which facts receive weight and how much uncertainty reaches the final estimate. You care because the exam wants a bias label tied to a forecast effect and a remedy that changes the process.

BiasForecast effectProcess remedy
AnchoringInsufficient adjustment after new evidenceIndependent base case, update triggers
Confirmation biasSupporting evidence selected, contrary discountedDisconfirming log, devil’s-advocate review
OverconfidenceExcessive certainty, narrow forecast rangesCalibrated probabilities, scenario ranges, forecast-error tracking
HerdingConfident peers pull estimate toward consensusBlind independent estimate before group discussion
Availability and representativenessVivid events dominate, familiar patterns repeatBase rates, structured evidence checklist

Anchoring and confirmation bias change how new evidence enters the estimate, while overconfidence and herding change how uncertainty and independence are expressed. Availability and representativeness change how evidence is weighted, so a remedy must insert a step the biased process would otherwise skip.

To test whether confidence matches results, track forecast error after the outcome arrives. The actual outcome, $A_t$, is the result for period $t$. The forecast made for that period, $F_t$, is the estimate recorded before the result. Call the difference $e_t$ and calculate it this way.

$$ e_t = A_t - F_t $$

A positive $e_t$ means the actual outcome exceeded the forecast, so the forecast was too low. A negative $e_t$ means the forecast was too high. Tracking the sign and size of $e_t$ shows whether narrow confidence is earned. Calibration means stated probabilities match the frequency of outcomes.

Keep the forecast separate from evidence and group influence. A blind estimate and a contrary-evidence check protect the process.

An independent estimate before group discussion protects independence. A base rate and contrary-evidence check keep vivid facts from doing all the work.

Worked example

You are handed a vignette about an analyst who issued a bullish earnings forecast. New evidence weakens the outlook, but the analyst makes only a small change because the original estimate still dominates. You need to identify the bias, state its forecast effect, and recommend a process remedy.

ObservationForecast implication
Original estimate dominatesNew evidence is underweighted
Outlook weakensForecast changes too little
Process remedy requiredIndependent base case, update triggers

The corrected sequence starts with new evidence, not a defense of the old view. Make the independent base case before reviewing the prior estimate. Then compare that case with the old forecast and apply the stated update triggers. This order makes the analyst explain any difference using evidence, instead of shrinking the revision simply to preserve the original number.

Loading diagram…
  1. Identify the bias as anchoring by comparing the old estimate with the contrary evidence. This is the right test because the old estimate remains too influential after the evidence changes. Step 1 produces the bias label anchoring.
  2. State the forecast effect as underreaction by comparing the revised forecast with the old bullish view. This comparison matters because a small change leaves the forecast too close to the original estimate. Step 2 produces the forecast effect, underreaction, shown by an insufficient downward revision.
  3. Recommend an independent base case and explicit update triggers by specifying the process before the analyst revisits the old estimate. This process matters because it separates fresh judgment from the anchor and identifies evidence that should change the forecast. Step 3 produces the remedy of an independent base case and explicit update triggers.

The exam answer is anchoring with an under-revised forecast, so the analyst should use an independent base case and explicit update triggers before deciding whether the outlook changes.

Changed condition

The same analyst faces the same weakening evidence. The only change is that the analyst prepares an independent base case before seeing the prior estimate.

FeatureBeforeAfter
Starting pointPrior estimate firstIndependent base case first
New evidenceWeakening evidenceWeakening evidence
Expected responseSmall adjustmentEvidence-led revision

Your prediction before looking should have been a more independent response to contrary evidence. The unchanged evidence means the movement comes from the process, not from a new fact. Before reading on, write the evidence that would force your revised forecast.

Common error

A candidate sees an optimistic forecast and calls every such view overconfidence. It feels right because overconfidence sounds like strong belief in the upside, whereas a well-supported bullish view can remain wide-ranging and evidence-balanced. What gives it away is unjustifiably narrow certainty or precision despite weak calibration and repeated misses.

Go deeper

Bias labels describe an action in the vignette, not a personality diagnosis. Read what the analyst did with evidence, uncertainty, or consensus before naming the bias. Telling yourself to be less biased is not enough because it adds no check. A correct diagnosis can still lead to a wrong conclusion because bias concerns the process used at the time.

An independent base case is formed before the old estimate or consensus can set the answer. Consensus may still be right, so disagreement with it is not proof of independence. A disconfirming-evidence log and devil’s-advocate review force the preferred story to face contrary facts.

Calibrated probabilities and scenario ranges expose overconfidence by making uncertainty visible. Forecast-error tracking then tests whether the stated certainty matches outcomes. A single miss does not prove overconfidence because calibration is judged across many forecasts, not one result.

For vivid events or familiar patterns, start with the base rate and use a structured evidence checklist. This slows the jump from an attention-grabbing fact to a forecast that assumes the pattern will repeat. A rare event can justify departing from the base rate when reliable, case-specific evidence explains why this case differs. The departure needs a stated reason, not just confidence.

Check yourself

Practice

An analyst keeps an old estimate after contrary evidence. Which choice best explains the bias and forecast effect, and recommends a process remedy?

An analyst sees a vivid recent event and assumes a familiar pattern will repeat, while ignoring broader frequencies. Which remedial action is most appropriate?

Revision snapshot

Track forecast error with $e_t = A_t - F_t$. A positive $e_t$ means actual exceeds the forecast. A negative $e_t$ means the forecast is too high. A remedy must change the forecasting process, not merely tell the analyst to be less biased.

BiasForecast effectRemedy
AnchoringInsufficient adjustment after new evidenceIndependent base case, update triggers
Confirmation biasSupporting evidence selected, contrary discountedDisconfirming log, devil’s-advocate review
OverconfidenceExcessive certainty or narrow rangesCalibrated probabilities, scenario ranges, error tracking
HerdingConfident peers pull estimate toward consensusBlind independent estimate before discussion
Availability and representativenessVivid events overweighted, familiar patterns repeatedBase rates, structured evidence checklist

On the exam, identify the behavior, state its forecast effect, and choose the matching process remedy. Do not call every optimistic forecast overconfidence. Narrow certainty, weak calibration, and repeated misses are the stronger evidence.

Unlock the full lesson and continue reading

Signup to continue reading this lesson and unlimited access questions, flashcards, AI notes, and more

Scan to download app

Scan to download
UNLOCK FREE ACCESS

Flashcards: Behavioural biases in forecasting

  1. 1. In financial statement modeling, what are the four main forecast approaches?

    Show answer

    Historical results, historical base rate and convergence, management guidance, and analyst discretionary forecast.

  2. 2. In financial statement modeling, how are maintenance and growth capital expenditures forecast differently?

    Show answer

    Maintenance Capex is forecast starting from historical depreciation + inflation. Growth Capex requires knowing management's future business strategies.

  3. 3. In forecasting capital expenditures, what are the two main categories that analysts split Capex into?

    Show answer

    Maintenance capex and growth capex.

  4. 4. In financial statement modeling, what formula calculates operating profit?

    Show answer

    \[\text{Operating Profit} = [Q \times (P - VC)] - FC\] where \(Q\) is quantity, \(P\) is price, \(VC\) is variable cost, \(FC\) is fixed cost.

  5. 5. In expense forecasting, what formula estimates COGS using the gross margin approach?

    Show answer

    \[\text{Forecast COGS} = (1 - \text{Gross Margin}) \times \text{Estimate of Future Revenue}\]

  6. 6. In financial statement modeling, what formula calculates the Degree of Operating Leverage (DOL)?

    Show answer

    \[DOL = \frac{\% \Delta \text{ Operating Profit}}{\% \Delta \text{ Sales}}\]

  7. 7. In financial statement modeling, what are the four key forecast objects?

    Show answer

    Drivers of financial statement lines, individual financial statement lines, summary measures, and ad hoc objects.

  8. 8. In financial statement modeling, how should nonrecurring items be treated in forecasting?

    Show answer

    Excluded from base forecasts and analyzed on a stand-alone basis.

  9. 9. In financial statement modeling, what are normalized earnings?

    Show answer

    Expected midcycle earnings, absent the temporary effects of business cyclicality or one-time events.

Everything you need for CFA prep

Get full Oncourse access with lessons, practice questions, flashcards and AI study tools.

GET STARTED FOR FREE
75% OFFLimited time offer
GET 75% OFF