Core Earnings – corporate earnings which have been adjusted, using a combination of human and machine learning inputs, to account for transitory shocks and for earnings from activities which are not central to a company’s business activities – represent a more accurate and persistent measure of a firm’s profitability than traditional metrics. The difference between Core Earnings and reported net income – the Earnings Distortion – can significantly explain future net income even after we have considered analyst consensus forecasts and accounting accruals.
We find that Core Earnings is more persistent over time compared to net income; its autocorrelation with next year’s value is 48%, which is noticeably higher than the 31% for net income. In addition, the autocorrelation with values further into the future is also stronger for core earnings.
We further show that a signal based on Core Earnings, which is long stocks with large negative Earnings Distortion and/or large positive growth expectations from the sell-side relative to our Core Earnings forecast (short stocks with large positive Earnings Distortion and/or large negative growth expectations), generates an annualized return of 10.1% and Sharpe ratio of 1.44 from 2015 to 2021 (long/short top/bottom decile, non-compounded, equally weighted monthly rebalanced). Most of the return is due to stock idiosyncratic returns (‘alpha’) rather than factor or sector tilts; after accounting for Fama-French 5 factors, momentum, short-term reversal, and 12 sectors, the signal’s residual return is 9.3%.