Quantitative investing is often perceived as an opaque and intimidating realm of finance with complex algorithms, mathematical models, and reams of data converging to make investment decisions. In this article, the aim is to demystify it for you. The way it may be implemented may be systematic, but we believe that at its core, quantitative investing is very logical and easy to understand. This will be a two-part article, and as a start we explain what factors are and how we identify those that work because quantitative investing is also known as factor-based investing. In the second part, we will go into how we translate these factors into investible portfolios. Our focus is equity specific, but these concepts can be extended to other asset classes too.

We talk about factors a lot in the quant space, but what are these factors?

Each publicly listed company generates information every day like the number of shares traded, changes in stock prices and so on. Factors are the information about a company that can be measured. Factors can change over time as the company reports new information like sales figures, revenue growth or profit/loss. This updated information is typically accompanied by a movement in the company’s stock price. In quantitative investing, factors are seen as possible predictors of stock prices and they form the core of a systematic investing strategy.

Factors can be grouped into styles like value (dividend yield or price-to-earnings ratio), momentum (three- or six-month price change), profitability (return on assets or return on equity) and so on. Some factors are related to the past information of the company while others may be based on analyst’s predictions on the future of the company. Usually, factors in a style behave similarly to one another. Depending on what is driving the market, some factors may be better predictors of stock price movements than others, at different points in time. This makes selecting factors that work complicated and creates a need for an evidence-based and objective framework to test and assess factors.

So how do we do factor testing?

We can define many factors, but we also need to define what we mean by if they ‘work’ or not. In simple terms, we want to select factors that have meaning, stability and significance. Meaningful factors have economic, financial, or behavioural reasons for predicting stock returns. Stability means the factors have performed well through different market conditions like recessions or interest rate cycles and have been persistent across different regions and countries. Statistical significance shows how a factor has performed in the past, and we want to select factors that have outperformed historically. We don’t want factors to have worked by accident or coincidence in the past, so three properties of meaning, stability and significance protect us against picking factors that have worked by chance. For example, the postal code of a company’s headquarters does not pass the ‘meaning’ test because there is no reason why the postal code should explain the movement in the company’s stock price consistently over time.

While meaning is an easy sense check, we have developed statistical tests to measure the significance and stability of factors. We conduct tests to inspect how these factors have performed historically. We examine the factor data for each stock from one period (possible cause) and the subsequent returns of the next period (possible effect). The battery of statistical tests that check for stability and significance, form a complete assessment of any given factor. The process is totally objective, and evidence-based.

We would then combine multiple factors that behave differently from each other, to make a composite factor. We send the composite factor through the same battery of tests to ensure that it tests better than its individual components, hence amplifying the predictive power of those components. The different methods that can be used to combine individual factors to make composite factors will be discussed in a separate article.

Pitfalls in calculating factors

A lot of care and consideration goes into calculating factors based on raw data. Making certain mistakes will impact the input data we use for testing factors and will give us incorrect results. For example, various corporate events can change the number of shares in issue and if we don’t adjust for this, our results will be distorted. Another possible source for error is the look-ahead bias – the reporting of data by a company could be delayed so the report date and the date when the data was available could be different. Survivorship bias is another common mistake where certain companies that have either delisted or gone bankrupt are not included when testing a factor. These mistakes will make the factor appear to be more or less predictive than what they have been.

 Understanding factors is the basics of quantitative investing and how to select factors that work. But what happens if trading costs are high or if very few parties want to trade the stock, so liquidity is low? What happens if, due to regulation, we can’t hold certain stocks, or we don’t want to take a large amount of risk? These are extremely relevant questions when we want to convert factors into investible portfolios. We will cover this in the next article in this series of when theory meets practice.