Each quarter, the LSEG Proprietary Research team publishes an earnings season forecast, where we identify five companies we expect to beat earnings expectations and five companies we predict will miss. Our analysis focuses on companies within the Russell 1000 index, leveraging analytics from the StarMine SmartEstimate and Predicted Surprise available in LSEG Workspace. These tools help us refine consensus estimates and highlight potential market surprises by overweighting the more timely and accurate analysts, allowing us to pinpoint discrepancies and deliver sharper predictions for the upcoming earnings season.
Our Russell 1000 selections for 2024 Q2 were 70% accurate, which can be viewed
here.
Positive surprise predictions for 2024 Q3
Negative surprise predictions for 2024 Q3
How to Create Your Own Earnings Season Forecast using LSEG Workspace
LSEG Workspace offers the ability to streamline complex data analysis, allowing users to harness cutting-edge tools like the ‘Screener’ to identify potential earnings surprises.
Users of LSEG Workspace can create their own earnings season forecast by going to the ‘Screener’ tool and setting their desired universe. Next, they can add the following criteria as shown in the image below.
Step 1: Screen for Earnings Surprises
Note: It’s important to set the predicted surprise filter to greater than or equal to 2%. If users wish to screen for negative candidates, set the filter to less than or equal to -2%. When choosing these thresholds, research finds that StarMine correctly predicts the direction of the earnings surprise 70% of the time.
Once the screen is run, a subset of companies that meet the specified criteria will be displayed.
Step 2: Apply a Light Human Overlay
When selecting candidates for an earnings season forecast, a light human overlay can be applied to increase confidence in predicting an earnings surprise. We share tips and tricks that we use when creating our forecast:
1. Broad disagreement among analysts and recent estimates can decrease confidence, especially when two highly rated analysts have differing estimates—one above the mean and the other below.
2. Relying on a single estimate, or a limited number of estimates, particularly when only one or two are included in the SmartEstimate due to stale data, can reduce forecast reliability.
3. Ensure a minimum dollar difference exists between the SmartEstimate and the Consensus estimate to avoid instances where a Predicted Surprise percentage greater than 2% or less than -2% is generated solely due to rounding discrepancies.
By combining advanced analytics with intuitive human input, users can refine their earnings forecasts, improving the likely hood of identifying actionable insights during earnings season.
