How accurate are analyst earnings estimates?

Accuracy varies with the company and the quarter. Forecasts tend to be closer for businesses with steady results and guidance to anchor on, and further off for cyclical or fast-changing ones. Companies often land near the consensus, and some make a habit of clearing it by a small margin.

How often are earnings estimates updated?

Whenever an analyst changes a model, which can happen after company news, a peer's report, an industry data release or a conference appearance. Data providers pick up those changes on their own schedules, so the consensus can move several times in the weeks before a report.

How many analysts cover a stock?

It ranges widely. Large, heavily traded companies can have many analysts publishing forecasts, while small companies may have one or two, or none at all. Most estimate screens show the count next to the consensus, and a figure built from very few forecasts deserves less trust.

Where Do Earnings Estimates Come From?

Q: Where Do Earnings Estimates Come From?

Sell-side analysts at brokerages and banks publish their own forecasts for a company's earnings and revenue. Data providers collect those forecasts and average them, by mean or median, into a consensus estimate. Each provider tracks a different set of analysts, so consensus figures differ between sources.

“Raising next-quarter EPS estimate to $1.20 from $1.15 on stronger pricing.” That’s a line from a research note. A consensus estimate starts there: one analyst at a hypothetical brokerage changes one number in a model, the change flows to the data providers that track the stock, and when the same thing happens across everyone who covers the company, the average moves and the figure in the estimate column on your earnings calendar changes with it.

Who makes the forecasts?

Sell-side analysts. They work at brokerages and investment banks, cover a list of companies in one industry, and publish models with forecasts for revenue, earnings per share and other lines, usually for the next few quarters and years. Their research goes to the firm’s clients. The individual forecasts also flow to data providers, who collect them from many firms. Analysts publish ratings and price targets as well. Those are separate from the EPS consensus, and most research screens give them their own columns.

Buy-side analysts, at funds and asset managers, make forecasts too. Those stay in-house and don’t show up in the published consensus.

How does a consensus get built?

The provider takes the forecasts it has for a given quarter and averages them. Some use the mean, some the median, and some show both. The choice matters when one analyst sits far from the rest.

Same company, same quarter, three different consensus figures. If the company then reports $1.19, it missed on one calendar, beat on another and beat on a third, and all three are telling the truth about their own inputs.

Why do different sources show different numbers?

Coverage. Each provider has its own list of contributing analysts, its own rules for dropping stale forecasts, and its own timing for updates, so a forecast that’s months old might still sit in one consensus after another provider has already dropped it. The earnings surprise on one screen won’t always match the next. Pick one source. Use it for both the estimate and the result.

What role does company guidance play?

A large one. Analysts build their models around earnings guidance when a company gives it. Forecasts cluster near the guided range. Guide to $1.15 to $1.25 for the quarter and a consensus of $1.20 is no coincidence. When a company doesn’t guide, forecasts usually spread wider, because each analyst is working from their own read of the business, and the consensus becomes a rougher average of guesses that may not agree on much at all.

What are estimate revisions?

Forecasts change. Analysts update after the company speaks at a conference, after a peer reports, or when new industry data comes out, and the consensus moves with them. Research screens often chart it.

Traders watch the direction. Rising estimates into a report mean analysts have been getting more optimistic. Falling ones mean the bar is being lowered, and a beat against a lowered bar is weaker than it looks, which is one of the explanations worked through in why a stock can fall after beating earnings alongside weak guidance and a run-up into the date. Some screens also count how many analysts raised their numbers and how many cut them over the past month, which tells you whether a move in the average came from one big change or from broad agreement.

What’s a whisper number?

An unofficial expectation. Traders talk about the number they think the company “really” needs to post, usually above the published consensus. Nobody publishes it. There’s no method. Still, the idea behind it holds: if the market’s hopes sit above consensus, a small beat can disappoint. Treat the whisper as a reminder of that risk, and keep it out of your arithmetic.

Adjusted or GAAP?

Most consensus EPS estimates are on an adjusted basis, meaning they leave out items the company excludes from its non-GAAP figure, such as restructuring charges or stock compensation. Compare the consensus with the company’s adjusted EPS, and check the release’s reconciliation table if the adjusted and GAAP figures are far apart.

What to check before a report

Open the estimate for a stock you hold and look at four things: the consensus, how many analysts sit behind it, the high and low forecasts, and which way the number has moved over the past month. Then check the date and time on the earnings calendar. A consensus that has been drifting down on thin coverage tells you something quite different from one that has held steady across a large group, and the difference is worth knowing before the release lands. More on reporting season is collected under the earnings topic.