Long-time correlations in company profit fluctuations: Presence of extreme events

The accuracy of earnings predictions is hampered by the several predominantly unpredictable effects due to the complex evolution of economy. Finding out which are the dominant market features embracing uncertainty is therefore the key to get beyond present state-of-art earnings forecasts. The analysis of annual revenues and earnings data (1954-2008) from the 500 largest-revenue U.S. companies suggests a linear relation between company expected mean profit and revenue. Annual profit fluctuations are then obtained as difference between actual annual profits and expected mean values. It is found that the temporal evolution of profit fluctuations for a single company displays a slowly decaying autocorrelation, yielding Hurst exponents in the range H=0.75±0.17. The study of profits cross correlations between companies suggests a way to distinguish typical earnings years from anomalous ones by looking at minimal information structures contained within the space defined by the associated covariance metric. © 2009 The American Physical Society.