Whenever one talks about trading strategies, the issue of trading frictions raises its ugly head. Here are some considerations and rules of thumb that I think put this into perspective.
If a manager owns an amount of an issue that represents a decent sized fraction of the average daily trading volume, then an attempt to see that issue is likely to cause the bib price to start sinking. This is a very difficult cost to quantify, and it is one of the hazards of very large positions. A manager with a position that is say 15 times as large as the average daily volume is bound to hurt himself getting out.
Market impact costs are a serious concern to institutional investors, but they are seldom a concern to individual investors. Only in absurdly illiquid shares such as micro-caps in the pink sheets would an individual have to worry about market impact costs --- and in that case it could still be the least of the worries.
Typically traders and institutional investors do not worry about the taxes unless that is specifically part of their charge. For individuals, trades within IRAs and other sheltered accounts can be done without regard for taxes.
In the real-world, taxes are an important part of an individual investor's decision making, but in the happy land of trading strategy analysis we simply acknowledge that our gains or losses are short term capital gains or loses and we leave the tax analysis for another day.
The next cost that anyone has to deal with is the bid-ask spread. If I immediately buy stock A and sell stock A, it is clear that I will have lost the cost of the spread. By the same logic, if I sell stock A and buy stock B, I have lost one spread (assuming that both A and B trade with the same, or comparable, spreads).
Today, at the retail level, the spread is typically one cent on liquid shares that are from $10 to $100.
Now here is a new wrinkle. If I simply buy at the Asking price and sell at the Bid, everything I have said above is true, but what if I place a limit order at the current Asking price. If my order is executed --- and the Bid-Ask has not moved, I have in essence been paid (at least part of) the spread.
Thus, paying one spread is an upper bound on this particular cost. In favorable circumstances it is feasible to ignore this cost. In perfect circumstances (which really do occur), you can consider yourself to be paid the spread!
If you are a retail investor with a full service broker, then you are paying huge costs on every trade and you are too dumb (or lazy) to learn anything from this page. We can safely ignore this case, and just think about accounts at places like Vanguard, Fidelity, Schwab, TD, IB, etc.
At Vanguard in an Admiral Account you pay a flat $8 per trade for any quantity --- limit or market. Thus, for a $10K trade your fee is 8bp and for a $20K trade your cost is 4bp.
If you are an institutional investor, your costs are more subtle to compute. There are even many situations where you "get paid" to trade, say when portfolio strategies are combined with market making.
I think that it is reasonable for an individual investor who makes trades in the $10K to $100K range to take 10bp to be a typical "transaction cost" for moving from asset A to asset B. Once could move this up or down a little, but, if I have to pick a figure, this is the one I would pick.
Here are some applications:
It is easiest and most fun, to work out the returns to a strategy without paying any attention to transaction costs. If you don't find cheese there, you won't find it anywhere!
Simplest Method
Given a strategy with cheese, look at your total amount of purchases. Take 10bp of that volume and subtract it from your terminal wealth. This is very crude, but but it is still a useful way to approximate total return after expenses.
More Precise Method
For each of your purchases add 5bp to the cost and for each of your sales subtract 5bp from the proceeds. This method deals with the issue of compounding more precisely that the simple method.
This also shows that unless the expected return to an individual trade is more than 10bp, then the trade is not likely to be worth doing. Since stocks in aggregate have historically paid about 4bp per day, you have to win about 2.5 days base-rate return to justify a trade.
What to Do?
Recalling the "weakest link" argument from our discussion about modeling, I think that it is perfectly sensible just to use the simple method. In the case when the transaction cost kill the strategy returns, this method will root out the truth with a lot less work than the more precise method.
Still, t he differences that you find between he simplest method and the more precise method do build up over time. We would not be happy applying the simple method on 100 years of data, but then again we would never dream of using such rusty old data anyway.
Navigation: Steele's Home Page, 434 Home Page, 956 Home Page