RESEARCH:
Using NLP News Signals to Forecast Volatility and Generate Positive Returns

With concerns over rising inflation and decisions at the Federal Reserve, market volatility was fueled to exacerbated highs in January 2022, spilling into February with no immediate sign of relief. Over three-fourths of companies in the S&P 500 cited inflation during Q4 earnings calls, and stocks had their worst month since the COVID-19 pandemic began. By leveraging contemporaneous signals from news and earnings calls, is there a better way to mitigate risk around inflation and manage your portfolio during these times of turbulent volatility?

Volatility Around the Announcement of Economic Indicators

As inflation has been at the forefront of the news this quarter, it should be intuitive that investors factor this information into the forecasts of portfolio and market risk accordingly. By doing this, investors can capture a real-time view of the market and risk factors and, in the case of January 2022 and beyond, inflation. There has been much discussion on adjusting risk for news at both the company and the factor level. Evidence shows news as a conditioner for forecasted volatility is a more efficient proxy than measures such as implied volatility and other contemporaneous signals.

Traditionally, the well-informed investor has relied upon economic indicators such as the Consumer Price Index (CPI), Producer Price Index (PPI), or announcements from the Federal Reserve as indication of what will happen with inflation and the effects on the markets. While they may have some idea of what actions to take once these announcements are made, it is often too late. Take, for example, a traditional Autoregressive Conditional Heteroskedasticity (ARCH)/Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model of volatility surrounding the announcement of an economic indicator. Investors know the announcement date; however, they are not certain whether the direction will be positive, neutral, or negative.

While strategies are almost certainly in place for action to be taken for each potential outcome, literature has shown that prior to an announcement of relevant economic and financial data, liquidity dries up. Based on this, our ARCH/GARCH process will forecast lower volatility. After an announcement, there will be heavy volume with market participants reacting and managing their portfolios as needed. This will cause our volatility forecast to increase, but only for a short period of time before the volume calms and returns to normal post-announcement levels. In essence, we’ll get our forecast for market volatility wrong twice.

Constructing a More Dynamic and Efficient Volatility Indicator

To remedy this, investors can look towards contemporaneous sentiment signals to adjust risk forecasts and dictate what trading strategies should be employed. As mentioned previously, concerns around inflation in the United States (and other countries around the globe) are shown by tracking management discussions in earnings calls and annual reports. By using this data alongside a macroeconomic sentiment indicator derived from the natural-language processing (NLP) of financial news, we can construct a more dynamic and efficient indicator of volatility.

To calculate the volatility adjustment factor, we use an exponentially weighted five-day aggregation of Alexandria Technology’s Macroeconomic news signal, which is scored for both volume and the polarity of the sentiment. While a study of the asymmetrical price reactions of positive vs. negative news warrants its own discussion, one of the earliest studies was published in the Journal of Financial Economics in 1988.

In the period leading up to the announcement of CPI data, there is a clear increase in forecasted volatility from the negative sentiment surrounding inflation during January with the market reacting negatively after most official indicators are released.

By using sentiment analysis alongside traditional economic indicators, we can forecast a more accurate view of the true market volatility and reposition our portfolios quicker than if we were using purely backward-looking data.

Creating Trading Strategies Around Periods of High Volatility

Investors can also leverage news sentiment data to create trading strategies that take advantage of these periods of high volatility to outperform the market. Using sentiment extracted from corporate earnings calls, there has been a clear divergence in the performance of companies that negatively mention topics such as inflation relative to the S&P 500. With analytics derived from financial news, investors can navigate their portfolios around negative trending names and position them towards those moving in a positive direction.

Our past research has shown that using a strategy that goes long for companies that exhibit positive sentiment and short for those with negative sentiment, we can generate returns over the benchmark, especially in times of crisis and high volatility.

Results From Implementing Investment Strategy Around Sentiment During Market Crises

Conclusion

As we await further data and indicators about the economy for this month and forward, investors should be proactive and use contemporaneous news signals derived using NLP that help us understand and mitigate the risk in our portfolios, in near real time. In doing so, active investors can also take advantage of these periods of high volatility and position their portfolios based on sentiment to generate positive returns.