Markus Ebner, head of multi-asset at Quoniam Asset Management, says trading desks are in store for a busy week of earnings, with potential for volatility and high volumes across the semiconductor and the financial sector.
Ebner discusses how the buy-side firm derives sentiment signals from 1.5 million news articles daily and how it is used to inform investment and trading decisions. He unpacks the firm’s methodology for filtering out the noise, particularly during busy news cycles such as earnings, the US election frenzy, and central bank policy decisions.
In this episode, Ebner also looks at the challenges and limitations of using unstructured data sets like news and how his team hopes to overcome them.
Interview transcript
Jo Gallagher Welcome to Trader TV This Week your insight to how trading desks can prepare for the week ahead, I’m Jo Gallagher. Today I’m joined by Markus Ebner at Quoniam Asset Management, to discuss the main topics and events leading this week. Markus, welcome to the show.
Markus Ebner Thank you.
Jo Gallagher To start us off, what will be driving activity this week and what should trading desks be paying attention to?
Markus Ebner So we will definitely have a very interesting week ahead. The earnings season picks up, with more than 30% of S&P 500 announcing earnings this week, including Amazon and including Tesla. What we saw last week was on the one side, very positive earnings announcements from US banks like Morgan Stanley, for example. And we have seen a lot of flows into banks. On the other side, we have seen very disappointing figures that were published by ASML, with bookings, 50% less than expected. And this is very significant and that [is what] we have seen in the share price. If other semiconductor companies will be affected this week, this will be very interesting.
The sentiment we are measuring here currently regarding the earnings is still positive. However, sentiment from a macroeconomic perspective is weakened because of the geopolitical tensions with the likes of China doing military drills close to the border of Taiwan on one side, on the other side, there’s still the Israel-Iran conflict going on. We are all waiting for the response of Israel and the harder they strike back, the more the movement will be out of risky assets into the US dollar, into government bonds, and into safe-haven assets.
Jo Gallagher How do you derive signals from news feeds and then use that to inform your sentiment-based strategy?
Markus Ebner We believe it’s not only a hard fact driving the markets, but even more sometimes it’s the sentiment. And this strategy is one of the new strategies we launched at Quoniam three and a half years ago, and it’s purely based on unstructured data, on news article data. So what we do is we are collecting more than 1.5 million news articles every day, applying machine learning methods.
We are filtering them down to those news articles that are really driving the markets. We derive the sentiment, the underlying sentiment. And then on base of this, we are going long and short equity index futures, fixed-income futures, and currency forwards. So we are talking about more than 30 assets we are trading on. We are not trading the news, but we are trading the underlying sentiment of the news.
Jo Gallagher We are also in a really busy news cycle this Q4 with the likes of earnings, a US election coming up and also central bank decisions. How do you drown out the noise and make sure that the data in your models are reliable?
Markus Ebner So filtering out the noise is the most essential part of our approach. Finding the needle in the haystack is a highly complex process, and it took many years to develop. The first stage of this process is to filter about the topics. We don’t include, like yellow press topics, but topics which are really driving the markets like CPI announcements, earnings announcement, M&A activities and so on. This is a first filter stage. The second is about the quality of articles and our analysis have shown as an example, it is an article about inflation. If this article’s also mentioning central bank members, it is of higher quality than if this article is not mentioning central bank members. So we apply natural language processing methods to get the highest quality articles. And the third step is about the sheer amount of articles published because we have seen that journalists, of course, have to write something on a daily basis, even if nothing happens at the markets.
But if there is an event, we see the amount of articles jump significantly. As an example, ASML, when they published their poor bookings last week, the amount of articles about the Netherlands jumped more than 300%. So we are focusing on these events because otherwise we just get a signal of small, positive or small negative sentiment. In the end, nothing happened. But we have just a noise and if we would trade on this, we would just create trading costs, which we try to avoid.
Jo Gallagher And what are some of the challenges and limitations to using unstructured forms of data?
Markus Ebner There are so many data providers out there, and to find the data providers which not only give access to a high quantity of data but also to a high quality of articles, bringing us behind the paywalls on the Internet, was a strong challenge. So we ended up working with several data providers that complement each other.
Limitations are, we need a proper amount of news articles per day. We are talking about 100 plus news articles per market is necessary. And we have seen in some emerging market countries we don’t get this. High-quality articles are not available in this amount. And so we are restricting ourselves at the moment to develop market currencies. However, we are still on the search for new data providers to close this gap and also to include more emerging market countries.
Jo Gallagher Thank you, Markus, for your insight and of course you for watching, this has been Trader TV This Week.
Catch the show each week:
To watch other Trader TV This Week shows go to – Video
Or make sure to follow us on LinkedIn – Trader TV’s LinkedIn
Available as a podcast – here
