Explain how behavioural biases distort analysts' forecasts and ways to offset them
The idea
Analysts are subject to the same behavioural biases as other people, and forecasts show it. Overconfidence produces ranges that are too narrow; the remedy is to keep a record of past forecasts against outcomes and to use scenario analysis. The illusion of control makes a more detailed model feel more accurate than it is; the remedy is to focus on the few drivers that matter. Conservatism bias leaves forecasts anchored to old views after new information arrives; the remedy is to update deliberately, in the spirit of Bayes' rule. Representativeness leads analysts to extrapolate a vivid story while ignoring base rates, so the remedy is to check forecasts against what happens to similar companies. Confirmation bias leads them to seek out supporting evidence; the remedy is to look actively for the case against, for example by assigning a colleague to argue it.
A practice question
An analyst gives a 90% confidence range for next year's earnings that past outcomes have fallen outside of far more than 10% of the time. This most likely reflects:
- conservatism biasWhy not: Conservatism is slow updating after news. The problem here is ranges that are too narrow.
- overconfidence — correct
- confirmation biasWhy not: Confirmation bias is seeking supporting evidence. It does not by itself describe narrow ranges.
Ranges that are missed more often than their stated confidence implies are the signature of overconfidence. Recording forecasts against outcomes makes the problem visible.
A harder one
A company reports a clear, lasting improvement in margins, but an analyst revises the forecast only slightly and keeps it close to the old figure. This best illustrates:
- conservatism bias — correct
- representativenessWhy not: Representativeness is over-reacting to a story that looks familiar. Here the analyst under-reacts.
- illusion of controlWhy not: That bias is overrating one's grip on outcomes, not failing to update for news.
Under-reacting to new information is conservatism bias. The remedy is to update deliberately, weighing the new evidence as Bayes' rule would.
The hardest tier
An analyst expands a model from 40 to 400 line items and becomes more confident in the forecast, though its accuracy does not improve. This is best described as:
- conservatism biasWhy not: Conservatism concerns failing to update. Here confidence rose from effort, not from new evidence.
- anchoring on base ratesWhy not: Using base rates is a remedy for bias, not a bias that inflates confidence.
- illusion of control — correct
Believing that more detail gives more control over the result is the illusion of control. The remedy is to simplify and focus on the few drivers that really move earnings.
In the app: 6 practice questions on this outcome, a narrated walkthrough with diagrams, spaced review that brings it back before you forget it.
Take the free diagnostic →