Our estimates data is sourced from Thomson Reuters who have a data collection team with more than 20 years of experience. They, in turn, monitor the research output of over 600 contributing analyst/brokerage firms, including:
- Research Notes
- Non-research documentation contributed directly from brokers
- Web based electronic contributions which allow Analysts to contribute real-time revisions to the database direct from their desktop.
Thomson Reuters’ trained staff use strict, uniform methodologies to collect estimates and provide consistent, reliable detailed revisions and consensus data. It is Reuters' policy not to keep stale estimates or estimates from brokers who have dropped coverage to ensure that the data for a company is as current as possible in the database. All estimates older than 100 days are reviewed.
The estimates database is designed with quantitative / screening usage in mind and is used by academics and quantitative professionals worldwide. Nevertheless, it's important to understand the limitations of estimates data.
The consensus estimate is the aggregation of all the estimate of individual analysts' tracking a particular stock to provide a common view of the company. Depending on the level of coverage, this number could be the average of 30 different forecasts. On the other hand, if it's a smaller company, the estimate could be the average of just one or two analyst forecasts.
Thomson Reuters compute both a mean and a median for their consensus estimates. The mean is computed as the sum of the different estimates divided by the number of estimates provided. The median is obtained by ranking estimates from highest to lowest and then the middle estimate is taken as the median the median. If the number of estimates is even, then the average of the middle two figures will be the median. Note that all Thomson Reuters excluded all ’Suppressed Estimates’ (see below) before computing means and medians.
Because so many investors watch consensus numbers, the difference between actual and consensus earnings (known as positive or negative Earnings Surprise) is one of the most important factors driving share-price performance over the short term.
Not all forecasts will be included in the consensus calculation. In certain situations like profit warnings, improved company guidance, unexpected results or after a merger or acquisition, analysts may upgrade or downgrade their estimates to reflect the newly available information. Old estimates are excluded from consensus once analysts start to upgrade or downgrade after a significant event has occurred. Some analysts may have anticipated this significant event and have already set their estimate to take account of the event. These estimates are not excluded.
Reuters will also suppress estimates from consensus that are on a different basis to the majority of other estimates. There are certain situations where analysts are not in agreement as to what should be classed as an exceptional item, for example, restructuring costs for some companies can at times be looked upon as a recurring item.