How has systematic investment, with trades generated by rules or algorithms, evolved? What are its strengths and weaknesses? In his February 2021 paper entitled “Why Is Systematic Investing Important?”, Campbell Harvey summarizes the history, advantages and disadvantage of systematic (algorithmic) investing. Based on the body of research and personal experience, he concludes that:
- Historical stages of algorithmic investing are:
- Early algorithms automated a century-old investing approach called technical analysis, with identification and extrapolation of trends as cornerstone.
- Next came algorithms that specified stocks to buy or sell (or, overweight or underweight) based on price data and fundamentals such as value, growth, profitability and quality.
- Next smart beta strategies arrived, typically focusing on a particular stock return model factor, such as value.
- Next emerged high-frequency trading.
- Most recent is ascendance of machine learning and big data, enabled by computing advances and open source software, to exploit vast amounts of structured and unstructured financial information.
- Advantages of algorithmic investing are:
- Discipline to avoid investor emotional reactions and even to learn and exploit emotional errors of others.
- Processing of large datasets and quick responses to market news.
- Disadvantages of algorithmic investing are:
- Insufficient flexibility to respond to market adaptations/competitors.
- Suborning of researcher tendencies to overfit models during development, ignoring economic foundation.
- Lack of transparency, with fund managers limiting explanation because their algorithms are “proprietary information.”
- Increasing sophistication of algorithmic investing may force industry consolidation as large firms with both experience and technology drive out smaller competitors.
In summary, systematic (algorithmic) investing is increasingly dependent on specialized skills and technology required to implement machine learning-based strategies.
Cautions regarding conclusions include:
- The paper does not demonstrate/quantify the superiority of algorithmic investing outcomes.
- Development of algorithms based on machine learning is beyond the reach of most investors, who bear fees for use of algorithmic strategies of others.
