EPS Surprise as an Investment Strategy

EPS surprises have long fascinated the financial media and investors. But can buying on favorable surprises and selling or avoiding shares of companies that disappoint really serve as the basis of a stock-market strategy? Maybe it’s just hot air. I explored the issues on Portfolio123, and came away with a solid Surprise-based strategy that now suggests $DKS, $EBF, $EVC, $HOFT, $HUN, $KIRK, $RS, $SCVL, $UVE and $ZAGG are buyable. 

Step One – Does Surprise Really Matter?

It was easy to get a “yes” answer to the question by comparing backtest results of two simple screens. 

In both cases, I worked with a Portfolio123 Universe that approximates the Russell 3000 constituent list and used, as a benchmark, the iShares Russell 3000 ETF (IWV). Each of these starter screens consisted of a single rule. The screens were rerun and the portfolios reconstituted (stocks no longer making a screen were deemed sold and replaced by any new names that appeared in the result set) each week.

• For the Positive screen, the rule was, simply, to include all stocks in the universe if the companies reported positive EPS surprises (percent of surprise must be above zero) in the most recently reported quarter.
• For the Negative screen the single rule called for inclusion of all stocks in the universe if the companies reported negative EPS surprises (percent of surprise must be below zero) in the most recently reported quarter.

Table 1 shows the results of 5-year backtests of each strategy.

Table 1:  5 Year Test, Positive v Negative EPS Surprise

	Weekly Surprises		Benchmark
	Positive	Negative	IWV
Annual % Return	13.42%	9.88%	13.39%
Standard Dev.	13.62%	15.67%	10.77%
Max Drawdown	-26.29%	-30.80%	-15.41%
Beta	1.12	1.23
Annual Alpha %	-1.39%	-5.39%
Avg. 1 wk periods
All	-0.02%	-0.08%	0.22%
Up markets	1.11%	1.11%	1.18%
Down Markets	-1.82%	-1.95%	-1.30%

The main portion of the table, the part that shows the results of a hypothetical portfolio over the course of five years, suggests that EPS surprise does matter. The results for the hypothetical positive-surprise portfolio are significantly better than those for the negative-surprise portfolio.

That said, however, we see several things that dampen our initial enthusiasm for the idea of investing (or, rather, trading) on earnings surprise.

• Even the positive-surprise strategy is not meaningfully better than the benchmark
• The risks are high enough to drive the Alpha, even for the positive-surprise portfolio, below zero.
• The results of the rolling tests (average of 1-week holding periods) show such modest differences between the positive and negative portfolios as to suggest the difference we saw for the five-year test may have been to the luck of the selection of a start date.

Finally, there’s another flaw not visible in the table. Each portfolio consists of hundreds of stocks, too many to own in total and too many to serve as a viable list of ideas for further study.

Step Two – Stretching EPS Surprise Into a Full-Blown Strategy

While anything can and does happen periodically in the real-world markets, I believe any serious stock-selection strategy should flow from the VQS (Value-Quality-Sentiment) theme which is explained below in Appendix A.

Earnings surprise falls under the category of S, Sentiment. Hence there never was a  logical reason to expect it, standing on its own, to have supported a stock-selection strategy. V (Value) and Q (Quality) were not considered. As to S, there is no law against expressing it in terms of just one factor, , as was implicitly done above, but there is something to be said for using other S-related factors along with it: Use of multiple factors to express an idea reduces the risk of being led astray by atypical influences on a singe factor similar to the way a diversified stock portfolio reduces the risk of being harmed by oddities impacting a single stock.

There are countless ways of combining V, Q and S, and one can elevate one theme over others as per one’s personal preferences and risk tolerance. It’s fair to assume that one who is inclined to invest or trade on the basis of Surprise will want to make S the primary theme, so that’s what I’ll do here. I’ll keep the basic screening rule that limits consideration of stocks within the Russell 3000-like universe to those for which the companies reported positive EPS surprises in the latest quarter. Also, I’ll bring V and Q into the mix by adding two additional screening rules:

• • The Stock must score above 80 on a scale of zero (worst) to 100 (best) in terms of the Portfolio123 “Basic: Value” ranking system (the details of all ranking systems used here are presented below in Appendix B).
  • The Stock must score above 80 on a scale of zero (worst) to 100 (best) in terms of the Portfolio123 “Basic: Quality” ranking system.

To wind up with a manageable number of stocks (whether one wants to own all the stocks or research each of them in more detail), all that pass this screen, now expanded to three rules, are sorted on the basis of the Portfolio123 “Basic: Sentiment” ranking system, and the top 10 are selected for the final list. 

We now have a full-blown VQS strategy with EPS Surprise as the centerpiece. 

Step 3 – Testing the Strategy

While the theoretical underpinnings are sound (we don’t test for this, we rely on the financial logic explained in Appendix A), we cannot be sure that the specific factors we chose to express it are workable. That is the proper subject for testing.

We also need to consider holding periods. I initially assumed the screen would be refreshed and the portfolios would be reconstituted each week. I made that choice because of the headline-oriented act-now flavor of the hoopla that tends to surround EPS Surprise.

But is that really a sound choice? Information travels at nano-speed today and it is tempting to assume extreme data freshness is essential. But ultimately, the market is made up of humans who take in information at human speed. And stocks are associated with companies tied to humans (within the companies and among customers and clients) who also act at human speed, meaning investment cases, stories, often take time to develop notwithstanding that the machines wish they’d move faster. So I also test a slowed-down version of the strategy, one where we refresh the screen and the portfolio every four weeks. 

Tables 2 and 3 show the results of tests run over five- and one-year time horizons respectively.

Table 2:  5 Year Test, VQS Strategy Based on EPS Surprise

	4 Week Holding Period		1 Week Holding Period
	Strategy	Benchmark IVW	Strategy	Benchmark IVW
Annual % Return	18.89%	13.39%	22.18%	13.39%
Standard Dev.	16.50%	10.77%	18.67%	10.77%
Max Drawdown	-22.08%	-15.41%	-21.37%	-15.41%
Beta	0.95		1.16
Annual Alpha %	5.22%		6.66%
Avg. 4 wk periods
All	1.64%	0.96%
Up markets	3.19%	2.52%
Down Markets	-1.34%	-2.05%
Avg. 1 wk periods
All			0.30%	0.22%
Up markets			1.31%	1.18%
Down Markets			-1.30%	-1.30%

 

Table 3:  1 Year Test, VQS Strategy Based on EPS Surprise

	4 Week Holding Period		1 Week Holding Period
	Strategy	Benchmark IVW	Strategy	Benchmark IVW
Annual % Return	29.30%	15.78%	8.09%	15.78%
Standard Dev.	13.48%	9.24%	16.20%	9.24%
Max Drawdown	-9.61%	-9.36%	-13.46%	-9.36%
Beta	0.42		0.54
Annual Alpha %	23.26%		-1.55%
Avg. 4 wk periods
All	1.22%	1.11%
Up markets	2.50%	2.37%
Down Markets	-2.32%	-2.38%
Avg. 1 wk periods
All			0.15%	0.35%
Up markets			0.76%	1.15%
Down Markets			-1.01%	-1.18%

Step Four – Assessing the Strategy

In contrast to what we saw in Table 1, where we tested a one-trick pony consisting of EPS Surprise and nothing more, Tables 2 and 3 show that the EPS Surprise as the featured theme of a full-fledged VQS strategy has the potential to significantly outperform the benchmark and do so on a risk-adjusted basis. (Note, too, that trading costs are accounted for through an assumption of a 0.25% “slippage” penalty on the prices of all transactions that add stocks to or delete stock from the portfolio. So these 10-stock portfolios are real-world investable.)

We also see that the four-week refresh period is the better choice. 

In the five-year test (Table 2), either refresh period seems fine; in fact, we could say the one-year refresh period looks better in the regular start-to finish five-year test.

But the rolling tests at the bottom of Table 2 suggest a lesser degree of superiority for the one-week refresh intervals, thus raising the prospect that the superior start-to-end result we saw was influenced by good luck (and that a five year test with different start and end dates might not look as good). Table 3, however, suggests, however that there are more likely to be times when jumping like act like jackrabbits and re-doing things every week has greater potential for bad outcomes (likely as stocks are sold off before the VQS elements are able to properly develop).

Conclusion

The VQS strategy featuring EPS Surprise developed here is usable assuming a four-week refresh policy. It can serve as a fully-automated automated own-everything strategy, or as a very manageable idea generator consisting of 10 stocks that can be researched in more detail.

Here are the stocks that currently pass muster:

Dick’s Sporting Goods (DKS)

Ennis (EBF)

Entravision Communications (EVC)

Hooker Furniture (HOFT)

Huntsman (HUN)

Kirkland’s (KIRK)

Reliance Steel & Aluminum (RS)

Shoe Carnival (SCVL)

Universal Insurance Holdings (UVE)

Zagg (ZAGG)

 

APPENDIX A – THE VQS STRATEGIC THEME

• Starting Point: Stock should be priced at the present value of expected future cash flows
• Formulation: Gordon Dividend Growth Model, P = D/(R – G)
  • • P = Stock Price
    • D = Next Expected Dividend
    • R = Required Rate of Return which is based on market interest rates qand premiums for assuming (a) the risk of the equity market in general and (b) the unique risks associated with this company
    • G = Expected future (infinite) rate of dividend growth
• Practical adaptation to address the reality that dividends are often minor or non-existent; treat all of Earnings as accruing to shareholders with shareholders implicitly choosing to reinvest all or most earnings back into the business. Substituting E (Earnings) for D, we get:
  • • P = E/(R – G)
  • Re-arranging the equation (dividing both sides by E), we get a formula for the ideal P/E
    • P/E = 1/(R-G)
      • This tells us that as G (growth) rises, P/E should rise
      • This also tells us that as R (required return) falls, P/E should rise. Therefore, considering the components of R, we know that as Interest Rates fall, P/E should rise, as equity market risk falls, P/E should rise, and as company-specific risk falls, P/E should rise. Expressing the latter a different way, as Company Quality rises, P/E should rise.
• It’s not practical to work as if this was a specific formula into which we can plug numbers because real-world estimation (especially the notion of infinite growth) is too difficult
• But we can use these ideas as a theoretical roadmap for developing a stock selection strategy
  • • Favor stocks that show such data characteristics as to justify a credible assumption that V (Value; P/E and/or other valuation ratios) is too low relative to G (expected future growth) and/or Q (company quality)
    • Future growth cannot be quantified; instead, use S (Sentiment) as a proxy for the investment community’s broad expectations regarding the future)
    • Hence the overall theme of the strategy: a combination of Value, Quality and Sentiment, or VQS

APPENDIX B – PORTFOLIO123 RANKING SYSTEMS USED IN THIS POST

NOTE: TTM = Trailing 12 Months

Basic: Sentiment

• Estimate Revision (50% of Ranking System)
  • • Revision of EPS for Current Fiscal Quarter in past 4 weeks, higher is better (33.3% of Category)
    • Revision of EPS for Current Fiscal Year in past 4 weeks, higher is better (33.3% of Category)
    • Variability of Current-Quarter EPS Estimates, lower is better (33.3% of Category) 
• Surprise (30% of Ranking System)
  • • EPS Surprise in latest Quarter, higher is better (65% of Category)
    • EPS Surprise in preceding quarter, higher is better (35% of Category)
• Recommendations (20% of Ranking System)
  • • Change in Average of Analysts Recommendations in past 4 weeks, more bullish is better (75% of Category)
    • Average of Analysts Recommendations, more bullish is better (25% of Category)

 

Basic: Value

• Factors Based on Income Stream (65% of Ranking System)
  • Earnings (50% of Category)
    • • Price/TTM EPS, lower is better (33.3% of subcategory)
      • Price/Estimated EPS for Current fiscal year, lower is better (33.3% of subcategory)
      • PE to Growth, lower is better (33.3% of subcategory)
  • Other (50% of Category)
    • • Price to TTM Sales, lower is better (50% of subcategory)
      • Price to TTM Free Cash Flow lower is better (50% of subcategory)
• Factors Based on Assets (35% of Ranking System)
  • • Latest Price/Book Value, lower is better (100% of subcategory)

Basic Quality

• Margins (25% of Raking System)
  • • TTM Operating Margin, higher is better (75% of Category)
    • 5 Yr. Avg. Operating Margin, higher is better (25% of Category)
• Turnover (25% of Raking System)
  • • TTM Asset Turnover, higher is better (100% of Category)
• Return on Capital (25% of Raking System)
  • • TTM Return on Invested Capital, higher is better (30% of Category)
    • 5 Yr. Avg. Return on Invested Capital, higher is better (40% of Category)
    • TTM Return on Equity, higher is better (20% of Category)
    • 5 Yr. Avg. Return on Equity, higher is better (10% of Category)
• Finances (25% of Raking System)
  • • Latest Current Ratio, lower is better (30% of Category)
    • TTM Interest Coverage Ratio, higher is better (45% of Category)
    • Latest Debt to Total Capital Ratio, lower is better (25% of Category)