The Backtesting Process

There are three steps in backtesting: strategy design, historical investment simulation, and analysis of backtesting.

Step 1: Strategy Design

The first step is to formulate the investment hypothesis and goals. An active strategy would aim to achieve excess returns above the relevant benchmark. Note that an investment hypothesis is a security selection criterion, trading rule, and a portfolio whose objective is to achieve the investment goal. A procedure is then conducted by translating the hypothesis into specific key parameters, rules, and processes that need to be tested. The key parameters are:

• Investment universe: An investment universe refers to all potential securities that we can invest in. Most investors use broad market indexes as their universe. For this reading, we will use, S&P/TSX Composite Index for Canada, MSCI China A for mainland China, Russell 3000 Index for the USA, and S&P Global Broad Market Index (BMI) for all other markets.
• Return definition: The currency in which the return will be computed needs to be determined. A manager can either use one currency or denominate reruns in the local currency. If a portfolio manager does not hedge their currency exposure, they will backtest using single-currency-denominated returns. A benchmark must be specified if the aim of the investment strategy is an excess return. It is imperative to note that the benchmark must relate to the investment universe.
• Rebalancing frequency and transaction cost: Managers often rebalance their portfolios on a monthly frequency. Higher or daily frequency rebalancing will often result in higher transaction costs. Consequently, bid-ask spreads will cause price data to become biased. Many market anomalies are included once transaction costs are included. It is, therefore, imperative to consider them.
• Start and end date: Other factors held constant, investment managers will always prefer backtesting investment strategies with a long history. This is because a larger sample implies a much higher level of statistical confidence in the results. However, long period performance should be supplemented with the examination of discrete regimes within the long history.

Step 2: Historical Investment Simulation

During this step, the portfolio will be constructed. The portfolio must be rebalanced based on a pre-determined frequency. It is noteworthy that the construction process will depend on the investment hypothesis considered in step 1. Equally noteworthy is the fact that the investment manager’s styles and capabilities, and the client’s investment mandate are relevant to the potential strategy.

Analysts typically use rolling windows to simulate rebalancing. In rolling windows, a strategy or portfolio is constructed initially using historical in-sample period data. The strategy or portfolio is then tested with the out-of-sample period. As time goes by, the process becomes iterative and replicates the live investing process since managers adjust their positions based on newly available information.

Step 3: Analysis of Backtesting Output

During this step, backtesting results are presented and interpreted. Here, we are concerned about the risk profile as well as the average return of the portfolio. Therefore, it is common for analysts to use the Sortino ratio, Sharpe ratio, volatility, and maximum drawdown. Apart from measures, graphical representations are also key performance outputs. Graphical representations are a good way of summarizing multiple data points that reveal more than one number summary measure.

It is useful to assess the backtested cumulative performance of an investment strategy over an extended history. Performance should be plotted using a logarithmic scale where percentage changes are presented as the same vertical distance on the y-axis. One can readily identify downside risk, structural breaks, and performance decay using the cumulative performance graphs. Regime breaks or structural changes result from external factors, making the past an unreliable prediction of the future.

Question

Which of the following is likely to occur when an investment manager rebalances their portfolio every week?

1. The transaction costs will increase.
2. The transaction cost will reduce. 
3. Transaction costs will remain the same.

Solution

The correct answer is A.

The consequence of increased frequency in rebalancing is higher transaction costs.

Reading 42: Backtesting and Simulation

LOS 42 (b) Describe and contrast steps and procedures in backtesting an investment strategy.