A Long/Short Framework for Modeling Earnings Beats Given Non-Normality of Analyst Estimates

Abstract

Financial markets price equities as the sum of expected future cash flows, discounted back at their corresponding risk. Consequently, investors attempt to generate excess returns by identifying deltas between the market expectation for cash flows and eventual realized cash flows during the duration of an investment. Given a large number of market participants, the broker consensus earnings estimate is relied upon as a benchmark for market consensus. As markets increasingly focus on near-term earnings results, the distribution of analyst estimates has become more important. Currently, both academics and investors calculate the consensus estimate as the arithmetic mean of all earnings per share (EPS) projections published by brokers. This thesis analyzes the distribution of broker analyst estimates of S&P500 companies in three stages. First, this thesis argues that the distributions of broker analyst estimates are highly non-normal, and the use of the arithmetic mean EPS estimate induces significant error into academic and industry models. Second, potential parametric distributions are fitted to analyst estimates, before selecting a model for each set of analyst estimates for a certain date and security. The location parameters of this distribution is used to construct an "adjusted" mean estimate based on the model. When compared to actual earnings results, these adjusted mean estimates explain future stock price performance better than the arithmetic mean consensus estimate. Finally, using these "adjusted" mean estimates, a long/short trading strategy is constructed to exploit deltas between a fitted distribution of broker EPS estimates and realized actual EPS results. It compares this strategy against an identical one using the arithmetic mean estimate. A better fitting distribution for non-normal broker EPS estimates assists both academics using consensus estimates as an input to factor models and industry practitioners attempting to generate excess return from superior forecasting of key financial metrics.