What it means
Data providers collect individual analyst models and publish the mean, and often the median, high and low. Companies, investors and journalists then treat that single number as what the market expects.
The number matters because prices already embed it. If a business is widely expected to earn $1.20 a share, a report of exactly $1.20 is not good news, it is simply the outcome already reflected in the share price.
Consensus is an average of opinions, not a measurement, and it inherits every bias of the people producing it. Analysts talk to the same management team, read the same guidance and cluster around each other, which is why estimates often bunch tightly and then move together.
Two extra statistics tell you how much to trust the average. The spread between the highest and lowest estimate shows how much disagreement exists, and the age of the estimates shows whether analysts have updated since the last piece of news.
Managements are well aware of all this. Guiding conservatively so the consensus settles slightly below what the company expects to deliver is a recognised habit, sometimes described as sandbagging, and it produces the small, repeated beats that some companies post quarter after quarter.
In practice
Real-world examples
Example 1
A semiconductor company reports revenue growth of 34% and the shares fall 6%. Consensus had assumed 40%, so a headline that reads well in the press release is a miss in the only comparison the market makes.
Example 2
A retailer with only three covering analysts posts a large earnings surprise. Investors discount it heavily, because with such thin coverage the consensus is barely more than one analyst's opinion plus two followers.
Example 3
A pharmaceutical group's consensus revenue estimate barely moves for six months while its pipeline news changes twice. A fund manager checks the estimate dates, finds most are stale, and builds a position on the view that the published consensus is out of date.
Think of it
Think of the consensus estimate like a weather forecast before a major outdoor event. It is the average prediction of multiple meteorologists, and people use it to decide whether to bring an umbrella.
Formula
Calculation
Consensus estimate = sum of individual analyst estimates / number of estimates
Earnings surprise % = (actual - consensus) / consensus x 100
Eight analysts publish quarterly earnings per share estimates for a distribution business: $1.10, $1.12, $1.15, $1.15, $1.18, $1.20, $1.22 and $1.28. These add to $9.40, so the consensus is $9.40 / 8 = $1.175, usually quoted as $1.18.
The median sits between the fourth and fifth values, ($1.15 + $1.18) / 2 = $1.165, slightly below the mean because the $1.28 outlier pulls the average up. The dispersion is $1.28 - $1.10 = $0.18, which is $0.18 / $1.175 = 15.3% of the consensus, a wide spread that signals real disagreement about the quarter.
The company reports $1.24. The surprise is ($1.24 - $1.175) / $1.175 x 100 = 5.5%, a beat against the average, though it lands below the most optimistic analyst's $1.28 and would disappoint any investor anchored to that view.Case study
Seen in the real world
The following is an illustrative and entirely fictional example. Verrall Fieldwork, an invented survey and inspection company, listed with six covering analysts whose first-year earnings per share estimates ranged from $0.72 to $0.96, a consensus of $0.84. The finance director noticed that the spread of $0.24 was almost 29% of the consensus, which is unusually wide, and traced it to different assumptions about how quickly a new contract would ramp up.
Rather than let the disagreement persist, the fictional company began disclosing contract start dates and expected first-year contract values in its quarterly release. Within three quarters the spread had narrowed to $0.07 while the consensus itself barely moved, from $0.84 to $0.86.
The invented management team regarded the narrowing as the win, not the level. Wide dispersion had been producing sharp share price moves on ordinary results, and once the range tightened the shares behaved far more calmly around each announcement.
Watch out
Common mistakes
- Treating the consensus as a forecast of what will happen, when it is an average of opinions that tend to cluster and lag events.
- Ignoring how many analysts are behind the number, so a consensus built on two estimates is treated with the same confidence as one built on twenty.
- Comparing a company's own adjusted earnings figure with a consensus built on a different definition, which manufactures a beat or a miss that is not really there.
Questions
People also ask
Is the consensus the mean or the median?
Most providers publish the mean, though the median is more resistant to a single extreme estimate and is worth checking whenever the spread is wide.
Why do shares sometimes fall on a beat?
Because the market may have been positioned for a larger beat than the published consensus, or the guidance issued alongside the result was weaker than expected.
How often is the consensus updated?
Continuously, as individual analysts revise, which is why the date range of the underlying estimates matters as much as the number itself.