Aggregate Operating Earnings Forecast

Over the long term, corporate earnings are a principal driver of stock valuations. Operating earnings, more than as-reported (Generally Accepted Accounting Principles) earnings, convey prospects for future earnings. Therefore, aggregate S&P 500 operating earnings are a critical input for both our Real Earnings Yield (REY) Model and our Reversion-to-Value (RTV) Model of the U.S. stock market.

For several years, we experimented with publicly available aggregate S&P 500 earnings forecasts from Standard and Poor’s and Reuters (prior to their merger with Thomson Corporation). As described in our blog entry of 1//07/09, these forecasts have often been highly variable and substantially inaccurate. The underlying methodologies may incorporate too much trend following and not enough mean reversion. They may also incorporate known analyst biases. Based on this experimentation, we developed a simple technical model for aggregate S&P 500 operating earnings.

The following discussion provides the CXO Advisory Group LLC forecast for aggregate S&P 500 operating earnings over the next several quarters, along with a description of the methodology used to develop the forecast.

We update this forecast as new actual aggregate quarterly earnings data become available.

See Blog Synthesis: Valuation Based on Fundamentals for research on the usefulness of operating earnings versus other accounting measures (such as cash flows), and actuals versus estimates, as stock valuation drivers.

Forecast  -  Methodology  -  Analyst Biases

FORECAST

The following chart shows our forecast for aggregate one-year trailing S&P 500 index operating earnings over the next four quarters, along with a one standard deviation error range based on backtested forecasts since 1960. We have not tested the degree to which the distribution of errors is normal; if it is not, then the "normal" interpretation of standard deviation as a measure of potential variability does not apply.

This forecast drives REY Model and RTV) Model projections (in combination with the Inflation Forecast for the REY Model).

METHODOLOGY

We rely on simple technical analysis to forecast aggregate earnings. By technical analysis, we mean prediction based only on past aggregate S&P 500 operating earnings data rather than prediction based on fundamental economic data such as trends in employment gross domestic product. Key guiding beliefs for this analysis are:

1. Corporate earnings are mean reverting to an approximately constant growth rate. The longer/further earnings have deviated from this baseline growth rate, the stronger the tendency to revert. (See our blog entries of 1/12/06 and 1/12/07, and for supporting research.)
2. The political cycle, and attendant economic policy, may be significant for earnings growth. In other words, earnings trends should consider at least two or four years of history (congressional or presidential cycle) and should consider history in two-year increments. (See Blog Synthesis: Politics and the Stock Market for research on the connection between the political cycle and stock market behavior. See also our blog entry of 8/26/08 on a two-year reversion effect for stock returns.)

Relevant to the first point, the following chart plots the log of actual trailing annual aggregate S&P 500 earnings by quarter, as tabulated by Robert Shiller and by Standard and Poor’s (with the former dovetailed to the operating earnings tabulation of the latter at the end of 1988), from 1911 through 2008 (97 years). We assume that earnings become "actual" as reporting of the S&P 500 companies exceeds 50% according to Standard & Poor's. The chart also shows an extrapolated earnings growth trend line generated by a rolling linear extrapolation of log earnings for the immediate past 40 years. Since about 1960, this rolling extrapolation produces a good trend line for actual log earnings.

Based on the beliefs and research noted above, we hypothesize that (since 1960) the two-year cumulative deviation of actual earnings from the modeled trend line has some ability to predict the future change in earnings. We execute that prediction using rolling regressions of the two-year cumulative deviation and actual changes in earnings for the next one, two, three and four quarters using the latest known data.

We generate error ranges for the earnings forecast by applying this methodology to each quarter since 1960 and calculating the standard deviations of the percentage differences between forecasted earnings and actual earnings for each of forecasted quarters one through four.

In summary, as indicated by research and empirical tendencies, we use mean reversion from a two-year cumulative trend deviation to forecast changes in annualized earnings for the next few quarters.

ANALYST BIASES

The research listed below describes biases that expert analysts may have with respect to forecasting corporate earnings.

1//07/09, for an update of how S&P 500 aggregate quarterly operating earnings forecasts evolve from initiation through earnings season.

7//22/08, for perspectives on the valuation implications and sustainability of earnings growth.

5/20/08, for a study of the impacts of Regulation FD and the Global Analyst Research Settlements on sell-side analyst earnings forecasts.

2/21/08, for a comprehensive and organized overview of past research on equity analyst inputs, processing and outputs.

4/9/07, on the conservatism bias of investors with respect to both downside and upside extremes.

2/20/07, on the tendency of earnings forecasts to shift from optimistic to pessimistic across reporting periods as release of actuals draws near;

1/15/07, regarding the effects of the 2002 Sarbanes-Oxley Act (SOX) on management-analyst earnings dynamics;

11/16/05, for evidence that Regulation FD has leveled the playing field for analysts at large and small firms;

9/16/05, regarding the earnings estimate biases of analysts who work for brokers;

8/8/05, summarizing patterns and informativeness of management earnings guidance;

7/26/05, on the tendency of analysts to underreact to good company news when revising their earnings forecasts; and,

7/12/05, regarding the effect of Regulation FD in making analyst earnings forecasts less accurate (with some follow-up commentary in our blog entry of 7/17/05).