Markets explained · 2019-02-28

Correlations or insights

Illustrated story by Maurice van Tilburg – Glimne

< English translation of my De Telegraaf column with additional examples >

Have you seen recent articles in press and social media about 'Alternative data', 'Machine learning', 'quants' and 'artificial Intelligence'? These are difficult terms for sometimes complicated processes that are also becoming increasingly prominent in the world of trading and investing.

Listed companies periodically report on financial and non-financial results. The annual reports are filled with facts and figures. If you collect all those facts and figures from different companies for several years and combine them with, for example, economic and political information, you can come to valuable insights. Consider how strong a company is impacted by a change in interest rates, a decrease in the dollar or an increase in oil prices.

This information can be very important for an investor or asset manager. They could, in fact, use it to estimate the risks of his or her investment, but more importantly, hedge these risks. And in order to come to such insights, you first need to analyse a vast amount of collected data to great depth. When starting the analysis, however, the more complex terms and processes come into play.

Because of the large amount and availability of data, you need smart computers to analyze the data, interpret it and recognize patterns. The possible outcome of that process can be very valuable, but also misleading. I give a random example to illustrate: research shows that black cars are more likely to be involved in an accident than cars with any other colour. This pattern has been discovered during the analysis of data on car accidents. At first sight this information can be valuable for an insurance company. But we do not really know why this correlation exists, why is it just black cars? Probably you can think of a logical explanation, perhaps because black cars are less 'visible'?

Deeper analysis of the related data shows that especially people with high testosterone levels buy black cars. And those people also happen to cause accidents more frequently. So it was not the black colour of the car in itself. And seeing this we can conclude that the value for insurance companies is more in the actual cause and effect than the ‘correlation’ between accident trends and black cars.

It also works this way with stock prices and investments. By gathering, combining and analysing a material amount of data, you can come to surprising insights. An interesting example of such an analysis can be found for Royal Dutch Shell (RDS): If I ask you what the largest economic dependence of Royal Dutch Shell would be, then you are most likely to think of the price of a barrel of oil. But from deeper data analysis we can see that an actual change in the price of oil only has a weight of 11% in the share price of Royal Dutch Shell. Much greater impact on the RDS stock price can be found in an alteration of the inflation forecast, which weighs in at 24%. That's pretty surprising.

By looking carefully at all this data, we can make predictions on how a share price will react in a given situation. This makes data very valuable and indispensable for investing in any sort of company, large or small. So, take advantage of correlations and insights, but don't conclude too quickly and prevent being mislead!

EXAMPLES

Below I provide some examples of trend analysis and quant in depth analysis that could help investors take the right decisions and limit risk.

The chart of the BMW shows a detailed macro risk map based an a quantative data analysis for the BMW share price. The share price shows large correlation with Global sovereign risks (30%) the FX exchange rate evolutions (15%) and metals (13%). Based on the actual macro-evolutions a level confidence can be calculated.

Chart: QuantInsight / Euronext



Other quantitative data analysis of the of shares can do predictions based on deeper data research and early warning signals for changes. For example in the chart on the right trend B, when it signalled clear 'change upwards' and trend A when a 'confirmed increase' trend has persisted. In the same way it can highlight when certain shares enter a downward trend.

Chart: TrendRating / Euronext

Originally published on LinkedIn. View original article.