How large are transaction costs in practice?

External question — Quantitative Finance Stack Exchange Author: Stefan Voigt Original post: https://quant.stackexchange.com/questions/30816 License: CC BY-SA 3.0 — https://creativecommons.org/licenses/by-sa/3.0/ Adaptation: HTML converted to plain text; contact email addresses removed. I am wondering, what kind of transaction costs practitioners (institutional investors) are faced to. Portfolio optimization literature often evaluates portfolio performance after adjusting for a value taking the form $T(\Delta) = 50/10.000 \sum_{i=1}^N |\Delta_i|$ where $\Delta_i$ is rebalancing of wealth in asset $i$. I do not expect this number to reflect what is really going on in the markets, however, due to the lack of better approximations the functional form above is used frequently in academia. Does the proportionality constant of $50$ bp has some reliability in the industry? Do investors in reality face a fixed fee as soon as they touch a single asset? In other words, given I rebalance a small amount $\varepsilon>0$, am I going to by a fee anyway just because I called my broker? The form above also assumes that transaction costs keep smooth even I rebalance a lot of my wealth. Does this approximation hold in reality, for example by performing some form of smart order routing? I am happy for every reply or reference coming up with new ideas on how one could take into consideration transaction costs!

External answer — Quantitative Finance Stack Exchange Author: user18489 Original post: https://quant.stackexchange.com/a/30819 License: CC BY-SA 3.0 — https://creativecommons.org/licenses/by-sa/3.0/ Adaptation: HTML converted to plain text; contact email addresses removed. There are two types of costs one occurs while trading: commission fees and slippage. Commission fees (fixed amount paid to broker) are a based on a per share basis (amount per share x shares traded) while slippage (% worse/better execution than market) is a function of participation to total trading volume as well as the order type. Commission fee also depends on execution i.e. algorithm vs. discretionary. If you trade very liquid stocks,using algorithmic execution (i.e. VWAP) and trade below 3-5% of daily volume then commission fees can be $0.0015 per share and slippage <20bps. Think of this as your lower bound. If you trade illiquid stocks or bigger part of daily volume then slippage will increase. 50bps seems arbitrary, so unless you know the above conditions (trading volume, order type, execution venue) or you are able get some numbers from the brokers in order to model slippage, it would be more useful to find pnl decay to various costs assumptions and then think whether the strategy makes sense or not.