Double Surprise into Higher Future Returns: 8/10

Investment Potential Rating:
8/10 (1 worst, 10 best)

============

OVERVIEW:
Double Surprise into Higher Future Returns
Alina Lerman
NYU
Joshua Livnat
NYU
Richard R. Mendenhall
Notre Dame

Financial Analysts Journal, Vol. 63, No. 4, pp. 63-71, July/August 2007

Abstract:
Post-earnings-announcement drift is the well-documented ability of earnings surprises to predict future stock returns. Despite nearly four decades of research, little has been written about the importance of how earnings surprise is actually measured. We compare the magnitude of the drift when historical time-series data are used to estimate earnings surprise with the magnitude when analyst forecasts are used. We show that the drift is significantly larger when analyst forecasts are used. Furthermore, we show that using the two models together does a better job of predicting future stock returns than using either model alone.


Data Source:
Accounting data is from Compustat Point in Time database, returns are from CRSP, and analysts' forecast data are from the I/B/E/S database. Moreover, the database is confined to companies that had at least 8 quarters of data and had at least one analyst following in the prior 90 day period (amongst other screens like market cap, US traded, sufficient accounting data, etc). The data series runs from Q2 1987 to Q2 2005.

Data Specification and Hypotheses:

Andy published a great summary of the Post-Earnings-Announcement-Drift literature
that outlines the history of this anomaly. For this work, the authors decided that using a random walk model with a drift to estimate unexpected earnings surprises was not going to cut it. Instead of estimating unexpected earnings by using the past earnings numbers, these authors decide to use actual analyst forecasts.

The reason the authors use analyst forecast is because, theoretically, analysts should provide better estimates of earnings than the estimates generated from some model that is solely based on past earnings.


Investment Strategy:

This trading strategy is similar to past PEAD trading strategies in the sense you form SUE (standardized unexpected earnings) portfolios of stocks. Traditionally, this would mean looking at the past 8 quarters of earnings data to come up with a model of expected earnings for this quarter. With this model you would estimate what the 'expected' earnings would be for this quarter and compare it to the actual results (all of this is scaled by the standard deviation of errors in the estimate). From there you sort all the stocks based on their SUE and go long those who have the highest SUE (beat expected earnings the most) and short those with the lowest SUE (lost to expected earnings the most).

For this paper they do something very similar to the above strategy; however, instead of using a model to estimate expected earnings, they use actual analyst forecasts and scale them by the dispersion of analyst forecasts. They call their measure SUEAF (SUE analyst forecast). So what you would do is get all the prior quarter analyst estimates for the universe of stocks that has estimates and compare that to their current quarter earnings announcements. Now, sort all the stocks based on their SUEAF, long the highest SUEAF (stocks that beat their analyst expectations the most) and short the lowest SUEAF (stocks that lost to analyst forecasts the most).

Using the SUEAF long and short portfolios generates around 4.22% (2.62% from long and 1.6% fro short per quarter--not bad for a fairly simple and intuitive strategy). The SUE strategy generates only a 2.56% quarterly return (which isn't bad either!).

The authors also provide evidence that using this strategy based on revenues and not earnings has even greater results.


Implementation Issues and Remarks:

PEAD strategies have always fascinated me. Even so, after doing a lot of back testing on various strategies I have found that these long short strategies are very unrealistic and it is really only the long side of the strategy that is implementable. Shorting is costly (funding spread, finding broker who has shares, etc), tax inefficient, and dangerous during bubble times like 98-00.

Using this strategy is also data intensive and requires access to analyst estimates, which can be expensive and time consuming.

Overall, I would cut 2% (per year) from the performance for short side constraints, and perhaps 1% (per year) for data acquisition and general pain in the ass factor. The strategy makes around 4.22% a quarter or roughly 17% per year. If we chop 3% from that we are left with 14%. This is a solid, although not awe-inspiring return. My guess is this 14% comes with a lot of volatility during wild markets.

While, I sound skeptical, I still believe PEAD is a profitable strategy and should be recognized by all professional investors. If anything, if you have a stock that has a good quarter, you may want to hold on to it for a while--it's got a good chance of going higher.
Investment potential rating 8/10

WRG