Written By: Dr Markus Ebner
Dr Markus Ebner from Quoniam Asset Management considers how companies that can use unstructured data to generate reliable models will have the edge in producing long-term risk-adjusted returns
Data is sometimes referred to as the new oil or the new gold, and it is hugely valuable for a wide range of companies. Asset managers are among the firms that can benefit from the explosion in the amount of data being produced by using it to inform their investment decisions.
The data age
The amount of data being produced has skyrocketed in recent years. Google receives around 5.5 billion search queries every day, while there are roughly 6,000 tweets per second – adding up to around 500 million per day. The total amount of data being created, captured, copied and consumed is expected to hit 97 zettabytes in 2022, up from just two zettabytes in 2010 (a zettabyte is a unit of information equivalent to 1021 bytes). All this information clearly provides rich pickings for asset managers looking for input to give them an edge over their competitors and help them outperform the broad market.
Structured and unstructured data
Evaluating and analysing data in the financial sector is, of course, not new, but until recently much of the data that asset managers used was structured – generally in the form of quantitative information that is highly organised, fits nicely in spreadsheets, and can readily be interpreted by computer programs. Asset managers often use structured data from sources such as Bloomberg or Datastream to help them forecast the returns of individual stocks, sectors and markets.
Yet much of the information out there today is in the form of unstructured, alternative data. Examples include news articles, videos, audio files, social media posts and images. This is qualitative information, and it is more difficult to process using traditional data analysis tools. The good news for asset managers though is that more and more data providers are doing the preliminary work for them, for example, by filtering news articles according to certain terms. Once it is systematically aggregated and analysed, portfolio managers can build time series to show how sentiment develops over time. This kind of information can provide valuable insights into the growth prospects of individual companies, sectors, countries and investment themes, and as such it can be used as a trading signal.
A powerful tool
Unlike conventional macroeconomic data and widely publicised financial figures, unstructured data can provide indications of the underlying mood of a message. For example, how does the number of words with positive connotations in a news article or company outlook compare with the number of words with negative undertones? In other words, is the article positive or negative overall? Sentiment indicators that capture such dynamics and thus provide important signals for investment managers can be captured from a wide range of texts.
To back this up, studies conducted by Quoniam suggest that there is a causal link between media reports and the financial markets. In other words, we found that it is not only facts, but also the classification of those facts by investors, that determine the development of the prices of financial assets. Incorporating news sentiment into trading strategies therefore has the potential to enhance a portfolio’s return.
Another big advantage is that unstructured data is often available at much more frequent intervals than the structured data contained in, say, a company’s quarterly report or an analyst’s assessment of a stock’s prospects. This enables asset managers to capture market sentiment almost immediately – after all, their goal is to find and use information that has not yet been priced in by the markets. At Quoniam, we have found that forecasts for certain markets, such as Spain, improve when we include sentiment factors based on unstructured data. This gives us a better understanding of the current economic situation and its likely future development, and this forms the basis for our medium-term investment decisions.
Making use of unstructured data, however, isn’t always straightforward. There are two important prerequisites for it to be useful: first, the data must contain useful information about the security or market of interest; and second, this information must be reliable. This, of course, is not always the case.
For example, at Quoniam, we consider text from newspaper articles, but we do not currently use information from social media. That’s because we believe in adopting an academic approach, so we only use data that we know to be from a robust source. One study suggests that well over 30% of tweets are made by bots rather than people, clearly making Twitter a data source that can be manipulated. Yet that’s not to say that we won’t use data from social media in the coming years.
An unstoppable trend
Over the long term, the ability to exploit the possibilities provided by new sources of data could become the factor that determines whether an asset management firm is successful or falls by the wayside. Those companies that are able to use the flood of information to generate reliable models that make systematic investment decisions will in our view be best-placed to produce attractive long-term risk-adjusted returns.
That’s not to say that all fund managers will choose to adopt such an approach. There will always be contrarian investors who swim against the tide by, for example, betting against a certain market despite positive sentiment towards it.
On the whole, however, we expect more and more unstructured data to find its way into active managers’ investment processes, such that the challenge for any asset manager using it will be to find new data sources to keep ahead of the pack and maintain an information edge. This will involve the use of new technologies to process an ever-increasing volume of data covering not only text but also audio, images and videos, with the aim of making well-informed investment decisions. Of course, it will require considerable expertise and resources to process all this information, as well as in-depth financial knowledge to determine which data is useful and has a bearing on the returns of financial assets.