While investments in robotrading have grown considerably in recent years, other aspects of the trading process remain a more manual affair.  For instance, analyses of firms’ earnings are often still performed by analysts, and there can be a tendency for them to bias upwards.  New research from Wharton highlights how a machine learning-based approach can provide more realistic earnings reports.

“[With the machine-learning model], we can predict how the prices of the stocks will behave based on whether or not the analyst forecast is too optimistic or too pessimistic,” the researchers explain.

Their analysis revealed that analysts’ biases are most pronounced when the earnings announcement is not due for a little while.  These are then reviewed steadily downwards as the earnings announcement date approaches.  These revisions have the consequence of inducing negative cross-sectional stock predictability, however.

Cashing in

The researchers compared analysts’ earnings expectations with the benchmark figures provided by the algorithm.  This revealed the extent of the analysts’ biases and the window of opportunity thus created.  This provides investors with the opportunity to profit.

“With our machine-learning model, we can measure the mistakes that the analysts are making by taking the difference between what they’re forecasting and what our machine-learning forecast estimates,” the researchers explain.

For instance, investors could utilize this arbitrage opportunity to short-sell stocks that have received an overly optimistic report from analysts, with profits then pocketed as prices come down to a more realistic level as the earnings announcement looms.  Similarly, they could take advantage of overly pessimistic reports and then sell them for a profit then the earnings are higher than originally forecast.

A key finding from the research is that there tends to be a lot of variance between stocks in terms of how biased analysts are.  This means it’s hard to make a single aggregate statement that analysts are too optimistic across the board.

Varied use cases

The researchers also believe that their algorithm could be equally valuable for companies as it could be for individual investors.

“If you are a manager of a firm who is aware of those biases, then in fact you can benefit from that,” they explain. “If the price is high, you can issue stocks and raise money.”

Similarly, if the negative biases of an analyst push the stock price down, this would serve as a signal for firms to avoid issuing new stock at that moment in time.