automated trading - How to account for transaction costs in a simulated market environment?

I am simulating a market for my trading system. I have no ask-bid prices in my dataset and use adjusted close for both buy and sell price. To account for this I plan to use a relative transaction cost.

The question is how large such a relative transaction cost should be to be realistic?

• 
  

  I trade stocks only from Dow Jones Industrial average (read that it might be around 0.2% in this paper)

  
  



• 
  

  I trade only on day-to-day (no intraday) thus I can assume to buy at close prices (adjusted close?)

  
  


• 
  

  I simulate trades in the time period 2000-2012

  
  


• 
  

  By transaction cost I am interested in any cost related to a real trade, e.g., brokerage fee, spread, slippage (thanks @chrisaycock)

  
  

Answer

We cannot give you a relative bid-ask spread that would make sense. The reason for that is that it really depends on several parameters:

• The type of financial asset you invest in (futures, funds, index, options, ...)


• The period during which you're trading (I think the liquidity in markets hasn't been the same over time).


• If you trade intraday, it depends on the time you trade at.


• If you assume you trade close price, you would submit MOC (market on close) orders. Your broker is most likely to be able to sometimes execute your orders at a better price than the close, and sometimes the opposite. I think that some brokers will have some terms that allow you to make sure that your trade will be exactly the close price, I've seen this for some commodity futures.

Finally, bid-ask spreads are not transaction cost per-say; they are the market price of liquidity in a sense. Your transaction costs would be more a fee that is charged by the broker for executing the order. It depends more on the quantity you trade and the terms you manage to negotiate with your broker.