Dividends ignored in Simulation Testing of Trading Systems

[gahobbsau]

Do the usual methods for Simulation Testing of Trading Systems all ignore the contribution of Dividends recieved during the holding period of a stock?

OBSERVATION: I am using BullCharts and TradeSim and as far as I can see:
a) The usual methods and software for simulation testing on historical data, ignore the profit contributions of Dividends received during holding periods;
b) The dividends show up (usually) as decline in share price on ex-Div date, thus lowering profit results and ignoring the cash received.

QUESTION: Is this observation of mine about dividends in system testing results correct, or is there another way?

This situation seems to apply with any of the method descriptions for such testing that I have so fare read and for other software platforms - including AmiBroker, Metastock, TradeStation(?) - but I have not reviewed many.

I am aware of the following which can be left aside in discussion of this topic:
a) There are manual methods available to account for the dividends, such as during Paper Trading - whether live or on historical data such as using the Training Mode in BullCharts.
b) Historical dividend data is readily available and that the BullCharts data feed provides dividend data in its reports.
c) Franking Credits increase the contribution - but testing methods developed in the USA would not even consider these since they do not apply there.
d) In the US market, dividends yields are typically smaller than in Australia - for a variety of reasons. Locals will know that dividend yields (on share price) in AUS can be as much as 9% for Telstra, 6% for Com Bank.

 

[TabJockey]

Well most trading systems are focused on pretty short time frames for trades, so div payments are not a huge factor. If you have a system that incorporates cum and ex div dates into the strategy its going to be tough to back test that out by computer, I have never seen asx data with dividend details.

If you have a system that holds through allot of dividend collections, why dont you just work out that average amount you will collect per year? its probably not going to change a large amount from that. It might be worth on average an extra 4% account gain per year, and just tack it on to your estimates.

Do the usual methods for Simulation Testing of Trading Systems all ignore the contribution of Dividends received during the holding period of a stock?

OBSERVATION: I am using BullCharts and TradeSim and as far as I can see:
a) The usual methods and software for simulation testing on historical data, ignore the profit contributions of Dividends received during holding periods;
b) The dividends show up (usually) as decline in share price on ex-Div date, thus lowering profit results and ignoring the cash received.

QUESTION: Is this observation of mine about dividends in system testing results correct, or is there another way?

This situation seems to apply with any of the method descriptions for such testing that I have so fare read and for other software platforms - including AmiBroker, Metastock, TradeStation(?) - but I have not reviewed many.

I am aware of the following which can be left aside in discussion of this topic:
a) There are manual methods available to account for the dividends, such as during Paper Trading - whether live or on historical data such as using the Training Mode in BullCharts.
b) Historical dividend data is readily available and that the BullCharts data feed provides dividend data in its reports.
c) Franking Credits increase the contribution - but testing methods developed in the USA would not even consider these since they do not apply there.
d) In the US market, dividends yields are typically smaller than in Australia - for a variety of reasons. Locals will know that dividend yields (on share price) in AUS can be as much as 9% for Telstra, 6% for Com Bank.

 

[gahobbsau]

Thanks TabJockey. I will take that as vote for "I haven't missed something and its down to manual methods".

In response to your suggestions, no, I not developing or testing a system to catch dividends.

My obseration is about the negative impact of falls at ex-div dates.
a) It occurs in short holding periods as well as long. It only takes a day.
b) The share price impact is negative, though the impact on equity may be neutral.
c) In backtesting, such negative ex-div moves could be enough to cross a stop threshold and exit, even though the total equity from the share may not have changed.

 

[TabJockey]

Yeah there is no easy way around it, getting stopped out at ex div is going to affect your back test results. The only thing we can really do is look at a sample of the results, calculate the effect that dividends had and add that on to our system performance results.

Backtest results are pretty shakey indicators of future performance anyway, so you know that if you get great results even though you are not allowing for dividends you have an even greater chance of realising the sort of gains you saw in backtest.