2-24. Optimization with Transaction Costs

Now that you have some intuition of how transaction costs arise, let’s discuss how to model transaction costs as a function of the trade. Well, think about what’s called a Linear Impact Model. So we’re looking for an independent variable that has an effect on a dependent variable. The dependent variable is some measure that … Read more

Recall that the mid-point price is the middle of the bid-ask spread. Before we begin to execute a parent order, the mid-point price can serve as the benchmark from which we estimate the transaction cost. For each child order that is traded, the difference between the price paid and this benchmark is a measure of … Read more

To get some intuition on transaction costs and the reason that trading impacts prices, let’s look at what’s called a limit order book which is a list of all the bids and asks for an asset that are made by market participants. Bids are commitments to buy a certain quantity of an asset at a … Read more

An important reality of investment management is that transaction costs are incurred when making trades. For institutional traders, the cost is due to the impact of trade has on the market price of the asset being bought or sold. When you buy an asset in large enough amounts, this tends to push the price upward. … Read more

Note that the terms we’re calculating for the objective function will be in dollar units. So the expected portfolio return defined as the dot product of the alpha vector with the holding vector, will be in dollar units. The vector of portfolio holdings will be in dollar units. So now, let’s think about the vector … Read more

Now that we’ve moved beyond the Alpha research phase and into backtesting, we’re going to refer to the set of positions on assets and the portfolio as holdings as opposed to weights. In the research stage, the weights are considered percentages of the portfolios notional and the notional doesn’t need to be specified at this … Read more

An important thing to remember about backtesting is that we’re attempting to simulate live trading. So let’s say we have some data from which we construct an Alpha, and we also have the daily returns data that we use to realize gains or losses for our portfolio. If we look at the returns data recorded … Read more

Barra data, we’ll be using factor data that is generated by Barra. This will be good practice because Barra data is used throughout the industry. Note that we’ve preprocessed the raw Barra data files and sort the data into pickle files. The alternative would be to load the original data, and perform the parsing each … Read more

In previous projects, you’ve learned about various constraints that you might apply to an optimization in order to adhere to certain requirements. A constraint in an optimization falls into one of two categories. Either the constraint does nothing to the solution, or the constraint makes the solution sub-optimal. Constraints can also cause the optimizer to … Read more

Hi. In this lesson, we’ll show you how to design the portfolio optimizer for a back-test. The major theme of this lesson is how to design the back-test, so that it’s realistic enough that it’s results are meaningful. You’ll get familiar with data provided by Bara, which is widely used in the quant industry. You’ll … Read more