Smart beta is a relatively new term that has became ubiquitous in finance in the last few years, but the concepts behind it have been known for decades. It has its roots in factor investing, going back to the 1960s when risk factors were identified as the primary drivers of equity returns. Essentially, factors are risks that cannot be diversified and by bearing those risks one is expected to be compensated with a higher return. Therefore, factor investing seeks to capture these persistent and proven drivers of returns through exposure to macroeconomic factors such as economic growth or inflation and through micro-factors (also named style factors) that are investment characteristics, such as seeking inexpensive companies, or companies with higher yields. The idea behind smart beta is to take the concepts of factor investing that have been used in proprietary, discretionary ways in active management for years, and apply them in a transparent, systematic, rules-based approach. In doing so, the costs of smart beta drop dramatically compared to active management and moves towards the realm of passive investment.
After the economic crisis of 2007-08 there was an increasing demand from investors for more transparent, systematic strategies, and a decrease in management fees after a loss in confidence. This has led to a new demand in the financial sector for research on allocation of these risk factors in systematic, multi-factor, index-based strategies. Following this trend, the goal of our project is to improve smart beta strategies using data-driven techniques based on machine learning and big data. This means that model construction and investment decisions will be derived solely from the analysis of big data and not from fundamental analysis (insights into specific companies or economic conditions).
While smart beta strategies have shown a strong performance in the long run, they can often suffer from severe short-term drawdown (peak-to-trough decline) with fluctuating performance across cycles. These fluctuations can arise from extreme macroeconomic conditions, elevated volatility, heightened correlations across multiple markets and uncertainty monetary and fiscal policy responses. To address this we wish to build a factor timing framework, either through building a dynamic strategy using forecasted factor returns or by building a regime switching model.
Finally, the main question remains: where can big data have an impact in the allocation strategy? That is, which factors can be better predicted to get an early read of the factor, and therefore to get a sense of future smart beta performance. We can divide the sources of data in two groups
1. Sentiment indicators – social media, trading volumes, trader sentiment, bid ask spreads, volatility (VIX), etc.
2. Economic indicators – higher frequency economic data such as PriceStats daily inflation, predictors of unemployment, interest rates, inflation, PMI, GDP, etc.
With the increasing adoption of smart beta strategies and the demand for new research on signals coming from big data analysis, the next few years foresee a lot of activity in this interdisciplinary field, bringing together research insight from finance, econometrics, and computer science.
Elizabeth Fons is based at AllianceBernstein 2016-2019, and her research project is Smart Beta Investing – A Data-Driven Strategy to Exploit Systematic Risk Factors in the Financial Markets (WP4)