How to spot the next investing hype, narrative or trend using sentiment analysis
In a previous article I explained how behavioural flaws leads to predictive behaviour of investors and how to measure that. This lead to an innovative stock market indicator. In this article we will dig one level deeper and look at how sentiment can be used to get insight in niches of investing with actual examples. This article explains on which domains to focus and why, what metrics are important and finally some real-word examples. For all domains interactive dashboards are available at: Trend reports — Alternative Analytics (alternative-analytics.eu)
Which domains?
One could argue that some domains are more sensitive to sentiment than others. These are generally domains where it is hard to get a fundamental valuation. A large industrial company for example can be hyped, but one can (relatively) easy assess whether the hype is real. For companies or sectors that are harder to fundamentally value it is harder to assess this and sentiment plays a larger role in its sensitivity to price. Therefore we have identified some domains, which would be interesting to look at from a sentiment perspective:
- Trends — e.g. crypto, cannabis, solar power etc. Which trend is getting more attention, either positive or negative?
- Narratives — e.g. Covid-19 news, stimulus news, trade war news. 2020 has shown how a narrative can keep stock markets in its hold. Covid-19 comes to mind. However narratives on stimulus (will it pass or not) might be even more interesting due to the binary nature and direct effect on the stock market.
- SPACs — SPACs are currently the new interesting kid on the block. They are hard to value especially when there isn't an acquisition target but sometimes rumours move a SPAC value away from its intrinsic value. Sentiment and media exposure without a doubt plays a part in that and keeping track of it can offer interesting opportunities.
- Sectors — Tech, Industrials, Materials etc. Maybe not the first that domain that comes to mind. However capital flowing to a certain sector can also be influenced by sentiment. Moreover sector rotation is also visible in sentiment.
What metrics do we look at?
The first set of metrics that is looked at concerns exposure metrics. This is done by having an algorithm assign a topic to an article. This in turn is recalculated to show:
- Exposure of one topic compared to another topic — Eg the exposure of Tech sector vs Industrials sector
- Exposure of a topic over time — E.g. is the US-Chinese trade ware getting more or less exposure?
We can take this analysis one step further by looking at sentiment of these topics. Determining sentiment is done with AI as described in a previous article (How AI, alternative data and insights from psychology resulted in an innovative stock market indicator | by AlternativeAnalytics | Dec, 2020 | Medium). This gives us two similar metrics to the exposure metrics, namely:
- Sentiment of one topic compared to another topic — Eg the sentiment level of one SPAC vs a comparable SPAC
- Sentiment of a topic over time — Eg is Covid-19 news getting more positive?
The usage of these metrics becomes most evident by looking at some examples, which is done next.
Some real life examples
Now it is time for the most interesting part. See how this works in real life. All plots that are displayed can be found on: Trend reports — Alternative Analytics (alternative-analytics.eu) if you would like to play with the data yourself. The plot below shows series in the Narratives domain.
Covid-19 and the financial policy
Select “Covid” and “Policy” in the plots above. What you find with Covid-19 is what you expect, sharp decrease in sentiment followed by improvement. However we also included the narrative for financial policy, or put differently, actions by the FED. What you see is that after a sharp decrease following the Covid-19 news, sentiment with respect to the FED started picking up. More interestingly, if you plot it against the SP500 you see that an increase in policy sentiment preceded the rally in the SP500.
Trend investing — Crypto and Cannabis
Another interesting use case is looking which trends are gaining an increase in exposure. Put simply, Trend investing means buying an ETF that has exposure to a specific trend. Since these trend companies generally are still small and largely growth companies, they are sensitive to sentiment. Therefore we added also a Trend investing dashboard which follows current trends. The plot below shows the exposure of trends over time. Select “Cryptocurrency” and “Cannabis” for example and see how closely these terms lead Bitcoin price or a Cannabis ETF.
Some final words
In this article I tried to explain how quantitative sentiment can be used to get a feeling for the market and how it can give insight as to what the market is focussing on. If your are interested in the full dashboards you can find them on: Alternative Analytics — Viewing the stock market through a different lens (alternative-analytics.eu) . In a future articles we will show more novel use cases for quantitative sentiment.
