What Is Earnings Consensus? Four Things to Check Before Using Analyst Estimates
Consensus combines analyst estimates on a comparable basis. Learn to check the mean or median, collection date, accounting basis, and contributors before comparing earnings.
Consensus is a representative value calculated by combining multiple analysts' earnings forecasts according to specified criteria. It differs from a company's actual results or management's own targets. Its meaning becomes clear only when you know the period and metric being forecast, who made the estimates, and when they were prepared.
The same label can conceal different numbers
When opening a consensus data screen, first check the company, fiscal year, quarter, currency, and units. Operating profit and net income attributable to owners of the parent are different metrics, while consolidated and separate financial statements cover different scopes. Comparing a US company's adjusted EPS with diluted EPS under accounting standards can create a gap that looks larger than it really is. FactSet's official methodology also describes a structure that distinguishes estimates prepared on different accounting bases. Before calculating a representative value, you must first assemble a comparable group.
The mean and median answer different questions
The mean divides the sum of all estimates by the number of institutions, making it sensitive to extreme forecasts. The median is the middle value in the ordered set, so a few unusually large figures have less influence. Neither is always more accurate. If only the mean is high, it helps to read the sales-volume or margin assumptions behind the most optimistic forecasts. If a screen shows just one number, check the service's documentation to find out whether it is a mean or median. If you cannot establish that, record the limitation.
Estimates before and after an earnings release are different
To evaluate an earnings release, preserve the estimates that were available before the announcement. A screen updated after results are released may contain both new and older forecasts. Using the value currently displayed as though it represented past expectations will misstate the size of the surprise. This is why point-in-time data is preserved. Record the retrieval time, latest revision date, and contributor count beside the number so you can make the same comparison at the next release.
A company without consensus coverage may simply lack publicly aggregated estimates, rather than be impossible to forecast. Do not replace blanks with zero or call one institution's forecast a market-wide consensus. If you prepare your own forecast, separately record the sales-volume and pricing assumptions and the date, and identify it as a different kind of information from market consensus.
Worked example
Hypothetical example: Four institutions forecast operating profit for the same quarter at KRW 8 billion, KRW 9 billion, KRW 10 billion, and KRW 17 billion. The mean is KRW 11 billion and the median is KRW 9.5 billion. Actual profit of KRW 10.5 billion is KRW 500 million below the mean but KRW 1 billion above the median. Rather than simply calling it a beat or miss, specify which representative value you used. If the KRW 17 billion forecast is outdated, do not arbitrarily delete it; distinguish the original aggregate from an aggregate limited to institutions with current estimates.
What to check
- Have you aligned the forecast period and consolidation scope?
- Have you checked whether the figure is a mean or median?
- Have you preserved the pre-announcement reference point?
- Have you checked each institution's latest revision date?
Sources and primary materials
This is educational material about research methods. For an actual company decision, verify the filings and primary sources for the relevant period.