Stress testing or scenario analysis is a tool that is in every portfolio manager, financial risk manager or strategist’s toolbox. Additionally regulators nowadays require all financials to perform stress tests. The main critique on scenario analysis is that it always remains a bit vague and not concrete in as to whether which scenario will play out as mainly scenarios are defined as: “if this will happen, then this scenario will play out and it will probably has this effect on our portfolio.”. The problem here is that it is hard to assess which is going to play out, scenario analysis only tells you what will happen if it plays out.

In comes quantitative narrative analysis. We build an engine that can deduce a topic from financial messages or text and to assign a positive or negative connotation to it (in financial context, so similar to sentiment analysis but it can cope with financial terminology which standard sentiment packages or analysis cannot). In this way I can go through vast amounts of data and see whether the narrative in financial markets is trending towards a specific scenario.
An example of this is the Covid situation most of us will be able to remember. When markets tanked financial analysts started to make scenarios which all looked a bit like, scenario 1: extreme pandemic scenario, scenario 2: containment scenario, scenario 3: recovery scenario. It does not take a rocket scientist to determine these scenarios and their impact, but in order to make these scenarios really useful one has to assess which scenario is going to play out.

The “recovery” scenario that played out was not the result of rational behaviour or exact calculations. Rather it was glimmer of hope that started small when some people started to that the sentiment around Covid wasn’t getting any worse but started improving.

This is exactly where our model offers additional insight. First of all we monitor emotions and fear in headlines, which was getting increasingly worse pre covid crash. Following the covid crash we added some Covid scenarios to our narrative monitor in order to see if sentiment started to get worse or started improving. The “recovery” scenario that played out was not the result of rational behaviour or exact calculations. Rather it was glimmer of hope that started small when some people started to that the sentiment around Covid wasn’t getting any worse but started improving. As a small teaser have a look at when our quantitative sentiment engine recognized a shift in sentiment towards the recovery narrative relative to the fear narrative vs SP500. Note that “Emotions” scenario is a scenario that was established pre Covid (and therefore proved a suitable warning indicator). The “Covid” scenario was established in the midst of crisis in order to get a feeling for recovery.

The problem with traditional scenario tooling is that it does not tell you anything on which scenario is gathering momentum and is increasingly likely to become reality. In other words, you don’t know what narrative is gaining traction. Fortunately, with the recent progress in AI and big data we can get insight in to this. AI helps in “understanding” messages. We can train AI to recognize whether a sentence, tweet or headline is positive or negative. Moreover we can also train AI to recognize the topic the text is about. Additionally, big data helps in providing the “fuel” for AI. We have millions of opinions, headlines and texts on financial markets that we can feed to AI. Where a human can at most dozens of headlines in a day, AI can do millions and therefore can spot changing narratives with far greater accuracy.

We can bring insight in to which scenarios are at risk (or opportunity) to play out.

So we now we can enhance traditional scenario analysis with some really valuable metrics. Namely we can offer insight in to which scenarios are at risk (or opportunity) to play out. To give another example of this we put some scenarios of well know financial media outlets (JP Morgan, the Economist, Amundi and some other sources that published their scenarios) into our model. These scenarios are (amongst others):
• Value_comeback – After a significant increase in tech stock many scenarios envision a rotation back to value stocks. This narrative captures the presence in financial media and opinions on this scenario.
• Footloose_world – Airlines and travel industry took a big hit during covid. A frequently mentioned scenario envisages a rebound of a travel and leisure.
• Smaller_caps – The recovery of small cap companies as the economy reopens
• Recovery – Again a Covid related scenario. Will people be a as positive on a strong recovery or will this falter?
• Inflation – A hot topic after monetary policy. And especially interesting as our model has been picking up inflation fears already for months
This results in the following dashboard, containing all the stress scenarios. From this dashboard we can get insight in to which scenario or narrative is gaining traction and this gives us an indication in to which scenario the market is increasingly fearing or anticipating.

Complete interactive dashboard --> Narrative dashboard

This makes scenario investing more useful as one can adjust their portfolio or time rebalancing according to which scenario or narrative is more likely to play out. Our engine is able to provide insight in to any scenario or narrative a human can throw at it, so if you are interested in any additional scenarios let us know