I wish to replicate some of the findings of Heston and Sadka (2008)

General
1

I wish to replicate some of the findings of Heston and Sadka (2008) regarding the Seasonal Pattern: Stocks tend to exhibit similar relative performance in the same calendar month year after year.

The description of the signal is available here: CrossSection/Signals/Code/Predictors/MomSeasonShort.do at v1.4.0 · OpenSourceAP/CrossSection · GitHub

The study summarized the following findings:

  • Seasonal Pattern: Stocks tend to exhibit similar relative performance in the same calendar month year after year. This pattern persists for up to 20 annual lags.
  • Long-Term Effect: The seasonal effect is distinct from short-term momentum and extends far beyond previously documented momentum effects.
  • Off-Season Reversal: While stocks show positive returns in the same month across years, they exhibit negative returns in other months. This phenomenon is referred to as "off-season long-term reversal."

For testing this as a ranking system, the following formula should suffice:
100*(Close(220)-Close(240))/Close(240)

Thus, the stocks are ranked based on the return the stock had 11 months ago. A high return from 11 months prior also suggests better returns in the forthcoming month.

However, this does not yield very good results. Anyone tried something similar?

2

I have tried this for montly (5y history)
avg (isNA(Close (252) / Close (273),1), isNA(Close (504) / Close(525),1), isNA(Close(756) / Close (777),1), isNA(Close (1008) / Close (1029),1), isNA(Close (1260) / Close (1281),1))

and this for quarterly

avg (isNA(Close (252) / Close (315),1), isNA(Close (504) / Close(567),1), isNA(Close(756) / Close (819),1), isNA(Close (1008) / Close (1071),1), isNA(Close (1260) / Close (1323),1))