Mathematical model ignores analyst track record to draw bead on unexpected results
In the earnings estimate fandango, IROs have long contended with factors beyond their control that can drive trading activity – analysts straying outside of stated guidance, outdated or sloppy estimates, whisper numbers.
Soon, IROs will also face a new mathematical model that promises to identify ‘significant’ – meaning 5 percent or greater – annual earnings surprises 62 percent of the time in the US and 58 percent of the time internationally.
Standard and Poor’s Capital IQ unit has introduced Intelligent Estimate, a new analytical tool that improves on the accuracy of naïve consensus numbers.
It does so by weighting some analysts’ estimates more heavily than others based on several factors, but not by looking at an individual analyst’s previous track record for accuracy.
S&P says its intelligent estimate tool reduces the median fiscal year forecast error by 14 percent in the US and 10 percent internationally.
‘Investment professionals will be provided with a quick snapshot of where earnings surprises are likely to occur so they can position themselves accordingly,’ explains the firm in a statement.
For subscribers only
The catch for IROs is that only companies that subscribe to the service will be able to see when their company is flagged for a potential earnings surprise that may be driving activity in their stock.
‘For now, the product is available to subscribers’ of S&P’s earnings estimate service, says Cara Kiewel, director of communications for S&P Capital IQ.
The product is designed to ‘help clients identify opportunities and position themselves for trading and investing ahead of earnings,’ she adds.
When asked about the potential impact on issuing companies flagged for potential earnings surprises, Kiewel says ‘there are currently no discussions’ about informing issuers or making the intelligent earnings forecasts available on a stand-alone basis.
Which factors make analysts accurate?
David Pope, managing director at S&P Capital IQ, and one of the model’s authors, explains that most investors ‘keep an eye on which analyst has historically been the most accurate.’
The S&P model improves on that intuitive approach, he says, by weighting individual estimates based on the tenure of individual analysts (longer is better), the size of the brokerage firm (assuming bigger firms have better analysts), how recent the estimate is and how close to the earnings period the estimate is.
‘Rather than ask which analyst is best, we ask which factors make the analyst’s forecast best,’ comments Pope, adding that this multi-factor approach does a better job of separating skill from luck in outlier forecasts.
‘There is a tremendous amount of academic underpinning [behind developing the model]’, and S&P is ‘in a unique position’ to develop such a model because of the depth of its database, Pope adds.
Who’s covered?
The S&P Capital IQ Intelligent Estimate collects data from more than 700 contributors and covers more than 21,000 companies in over 100 countries.
Only companies with at least two estimates are included in the forecasting tool. Current coverage includes all of the S&P 500, the MSCI Asia Mid Cap, the FTSE indices and 99.6 percent of the MSCI World index.
The Intelligent Estimate model is strongest in North America, S&P says, because analysts are timelier in updating their estimates in North America. Analysts outside the US tend to herd to a greater degree as guidance takes on a more important role in countries such as Japan.
S&P back tested the model against historical estimates and earnings from 1999 in the US and from 1996 internationally.
