Thursday, August 24, 2017

Efficiently Inefficient #6 : On the costs of trading.

Some folks wrote to me concerning whether my back-testing models incorporate transaction costs. As of today, I kept the settings on transactions mostly default so I expect my models to over-estimate returns. To keep things safe, I always create a benchmark portfolio of equal-weighted STI stocks to compare investment return so that I do not get too carried away when I observe good back-test results.

In today's article, we will look at transaction costs. And as it turns out, it is hard to perfectly simulate transaction costs into a back-testing exercise.

One of the most objective definitions of transaction costs is the difference between the cost of one share upon execution and the average between the bid price and ask price of a stock, so generally speaking, any of the following can increase your transaction costs.

a) Brokerage costs

Not all brokers are the same. For investors who really want to pinch pennies, they should consider opening an account with FSMOne. Casual observation is that trading costs can be as low as $10 and there are no platform fees for stock trades.

b) Bid-ask spreads

The next element of transaction fees are bid-ask rates. Take for instance a counter like CEI, you can buy it from SGX at $1.055 but a seller will only accept $1.015 per share making this a really illiquid counter. Bid-ask rates are also insidious as the transaction fees tends to become larger the more units you buy.

This cost is what torpedoes back-tested dividends portfolios which tend to flag out thinly traded counters that generate a lot of free cash flow. Which brings us to the next point.

c)  Illiquid counters

Illiquid counters also increase transaction costs. Suppose you want to buy 100,000 of Global Testing. There are only 40 lots on sale at $1.10. If you insist on 100,000 in one trade, you might put in a Limit order of $1.15 and pray that you will get 100,000 shares with an average price of maybe $1.135.

This happens to me quite a lot because I trade during lunch breaks and just want to get my trades over and done with. As I seldom sell and look forward to years of dividend flows, I do not have patience to stretch my purchases over a couple of days.

d) There is an opportunity cost to optimizing transaction costs.

The is a limit to optimizing your transaction costs because you would need to change the way you trade to keep it at a minimum. The trade-off is against your opportunity cost. Perhaps your strategy exploited a trading opportunity that would disappear too quickly.

Where your transaction costs are too high, perhaps you need to put more focus on implementation measures. Trade more slowly and go after more liquid counters and find cheaper brokers. If this results in high opportunity costs, you might want to execute with faster speed.

I guess my only cold comfort is that all my back-tests so far tend to rebalance a portfolio only once a year.


No comments: