I am getting differences in the results for ETF dividend treatment in the screener. A test screen is here . A daily backtest for 11/15/09 to 12/1/09 shows a gain of 22.2% on the trade with screen date of 11/19/09 buying at the open on 11/20 and selling at the open on 11/23, after the weekend. Looking at the results using As Of = 11/19/09 and End Date of 11/23/09 I get a 6.84% gain for the TNA trade.
There was a dividend of $4.658 (here) according to Direxion. The Record date was 11/24/2009, the Ex-Date was 11/20/2009 and the Pay Date was 11/30/09. Since the buy on 11/20 was on the ex-date and the sell on 11/23 was before the record date it would seem that the dividend would not accrue to this trade although the backtest must be including it. Unfortunately, I think the 6.84% is the correct gain, not the 22.2% in the backtest. I wonder if this is a problem with other short term hold screens or just a data issue.
Glenn
----------------------------------------
[Edit 1 times,
last edit by gfagerlin at Jul 25, 2010 8:46:04 PM]
Glenn, you are correct. For the ETFs we simply adjust the price data with dividends using the exdate. It's a simplistic approach that generally works well, except for short term trading.
A comprehensive dividend handling using the appropriate dates for stocks and ETFs is on the to-do list.
Note that the adjusting the price data with dividends is only done for ETFs for now.
----------------------------------------
Portfolio123 Staff
Thanks Marco. I don't think it is a big deal. I checked other dividend dates in my screen and found no other instances when the short term trades crossed the ex/record dates. This was one trade out of about 200. Also, an unusually big dividend, I think. I doubt it happens that much.
