Predicting economic development with massive amounts of news data

Leif Anders Thorsrud

The variation in topics covered by the media tells us what is happening, and how important it is to the economy.

Imagine you are head of the Norwegian central bank just before the 2008 financial crisis. All seems well. The economy is growing, unemployment is low and inflation is under control. Suddenly, US financial markets crash and drag other markets down with them.

You cannot find updated information about the state of the economy in official statistics. The most commonly used indicator is the Gross Domestic Product (GDP), which tells us the value of what is produced by every company in Norway. The problem is that GDP is announced only every three months. The same applies to unemployment numbers. If you want to know GDP or unemployment numbers for the end of 2019, you have to wait until February 2020.

Financial markets produce a constant stream of data, but they are not credible when it comes to telling us what is going on with the economy. For example, the stock market predicts many more recessions than actually happen.

What is happening?

To track the economic stories told in the news, BI Centre for Applied Macroeconomics and Commodity Prices (CAMP) have created the Financial News Index (FNI) together with the media analytics company Retriever. FNI uses advanced computer programmes to search thousands of news articles and identify which topics they write about. A topic is a series of words that often appear together. For example ‘funding’.


Media coverage of the topic 'funding', 1989-2014

In the 2008 financial crisis much was written about words like ‘loans’, credit’, and ‘risk’. FNI monitors more than 20 different topics which we combine to learn about the state of the Norwegian economy. Because news are produced daily, and because news outlets cover all parts of the economy (not only the financial markets) FNI provides daily updates on the general economic development.

How does it affect the economy

The figure below shows two things. Firstly, the development of FNI (black line) is connected with the development of GDP (blue line). This is because studying topics covered by the news enable us to predict among other things economic production, consumption, prices and expectations. Zero indicates the economy is in a normal state with few or no changes. Positive numbers indicate an increase in activity, and negative numbers a decrease in activity. For example, economic activity declined after the oil price collapse in 2014.

FNI 2015-2020.jpg

Financial News Index with confidence bands and GDP, 2015-2020.

Secondly, the figure shows that towards the end of the period (inside the orange circle) GDP numbers are not available. This is because they had not been published for the end of 2019 at the time this article was written (January 2020). Nevertheless FNI is available and can tell us something about the state of the economy.

FNI is the result of years of research, where we have used news data to predict the economic development in Norway and other countries. Using news data in this way is a relatively new innovation in economics. FNI can be valuable to heads of central banks and others who need to make decisions in real time, as a supplement to statistics like GDP, unemployment data and data from the financial markets.



Leif Anders Thorsrud (2018) Words are the New Numbers: A Newsy Coincident Index of the Business Cycle, Journal of Business & Economic Statistics, DOI: 10.1080/07350015.2018.1506344

Vegard H. Larsen, Leif A. Thorsrud (2019), The value of news for economic developments, Journal of Econometrics, https://doi.org/10.1016/j.jeconom.2018.11.013

Vegard H. Larsen, Leif A. Thorsrud (2018), Business Cycle Narratives, Working Papers 6/2018, Centre for Applied Macroeconomics and Commodity Prices (CAMP), BI Norwegian Business School, https://biopen.bi.no/bi-xmlui/handle/11250/2492713

Vegard H. Larsen, Leif A. Thorsrud, Julia Zhulanova (2019), News-driven inflation expectations and information rigidities, Working Papers 3/2019, Centre for Applied Macroeconomics and Commodity Prices (CAMP), BI Norwegian Business School, https://biopen.bi.no/bi-xmlui/handle/11250/2598089

Published 30. March 2020

You can also see all news here.