The attached (“Equity Returns at the Turn of the Month”, Wei Xu and John McConnell) image demonstrates that the End of Month effect (8 days (4 days before and after the end of the month) is extremely strong over a lengthy period of time.
Do anyone believe this effect is significant enough to be used in a portfolio strategy?
For example, on the 25th of each month, rotate a portfolio. The new stocks are then purchased, and the ones to be sold are put in with a high limit price, with the aim that they will be sold within 8-10 days, i.e. by the 4-5th of following month.
If this effect is true for a part or the whole stock market, you could try to backtest it in a sim. In the best case you could get a few % outperformance over the market. But I have doubts, just because of the trading costs. With a screen it it not possible to backtest.
You could try to combine this effect with a strategy which does best with about 4 weeks rebalancing period.