The quantitative active investment strategy has a well-defined process:
- Define the market opportunity.
- Acquire and process data.
- Back-test the strategy.
- Evaluate the strategy.
- Construct the portfolio.
Define Market Oppurtunity
Quantitative managers use publicly available information to predict future returns of stocks, using factors to build their return-forecasting models.
Acquire and Process Data
The categories of most commonly used data are:
- Company mapping: Tracking many companies over time and across data vendors.
- Company fundamentals: Collect company demographic, price, and other financial data from vendors such as Bloomberg.
- Survey data: Details on corporate earnings and forecasts, macroeconomic variables, sentiment indicators, and information on fund flows.
- Unconventional data: Unstructured data including satellite images, measures of news sentiment, customer-supplier chains, corporate events, and many other types of information.
This involves applying the strategy to historical data to assess performance. The correlation between factor exposures and subsequent portfolio returns for a cross section of securities is used as a measure of factor performance in back-tests.
The idea is that if there is a strong relationship between factor exposure and subsequent performance then the factor has high predictive power. This correlation coefficient is known as the factor’s information coefficient (IC).
Evaluate the Strategy
Out-of-sample testing, where the model is applied to data different to those that were used to build the model, is conducted to confirm model robustness. Managers would look at both returns generated and risk measures such as VaR and maximum drawdown.
Construct the Portfolio
The following aspects are particularly relevant to quantitative investing when constructing portfolios:
- Risk models: Used to estimate the risk of the portfolio by considering individual variance of positions and correlation across positions. Managers generally rely on commercial risk model vendors for these data.
- Trading costs: Both explicit (e.g., commissions) and implicit (e.g., market impact cost) costs are considered. If two stocks have the same expected returns, the one with the lower trading costs will be selected.
Pitfalls in Quantitative Investment Processes
Pitfalls in quantitative investing include the following:
- Survivorship bias: If back-tests are only applied to existing companies, then they will overlook companies that have failed in the past, and this will make the strategy look better than it actually is.
- Look-ahead bias: Results from using information in the model to give trading signals at a time when the information was not available. A
- Data-mining/overfitting: Excessive search analysis of past financial data to find data that shows a strategy working.
- Turnover: Constraints on turnover may constrain the manager’s ability to follow a strategy.
- Lack of availability of stock to borrow: For short selling, this may also constrain a manager’s ability to follow a strategy.
- Transaction costs: This can quickly erode the returns of a strategy that looked good in backtesting.
- Quant overcrowding: Can occur if many quantitative managers are following similar strategies