What is machine learning?

  • Machine learning is a method where computers learn to identify patterns in data and make predictions. 
  • It works by having algorithms analyse large volumes of data, identify relationships and adjust themselves over time to become more accurate.  
  • Machine learning is used for everything from economic forecasting to facial recognition in smartphones and traffic analysis in self-driving cars. 

The Method in brief

To predict Austria’s GDP more accurately, the project uses machine learning, specifically techniques such as ‘Random Forest’, also known as decision trees, and Neural Networks.  

The Random Forest technique works by having a computer generate a series of ‘trees’, each making a ‘qualified guess’ about what the data suggests. The final prediction is based on the average of all tree outputs. 

Neural Networks is inspired by the way the human brain functions. Put simply, they can be used to recognise patterns in data. 

These methods allow the analysis of large amounts of economic data and the identification of patterns that traditional models often overlook.  

The model builds on historical data from the Austrian central bank and combines it with forward-looking indicators to provide more reliable forecasts. 

Project partners

Oesterreichische Nationalbank (OeNB):  The Austrian central bank, which uses the model for economic forecasting. 

OECD: International organisation which has integrated the project into a working group on economic forecasting. 

Schweizerische Nationalbank: Is part of an international working group investigating the use of the model in other countries. 

WU Vienna, Vienna University of Technology and University of Innsbruck: Collaborating on educational projects where students work with the model in their theses. 

News

See all news

Events

See all events

CBS research improves economic forecasting and makes fiscal policy more effective

Research project uses machine learning to improve GDP forecasts for the Austrian central bank.

05/12/2025

In a time of economic uncertainty and shifting trade agreements, the ability to predict economic trends has never been more important.  

A research project led by Associate Professor Thomas Lindner from the Department of International Economics, Government and Business at Copenhagen Business School has successfully used machine learning to improve Austria’s GDP forecasts.  

By applying advanced algorithms fed with data on foreign transactions, the project has significantly increased the accuracy of economic predictions. The method could, in principle, be applied to many other countries, according to Thomas Lindner. 

“In principle, nothing prevents us from using machine learning to improve GDP forecasts globally,” he says. 

More precise methods almost halve uncertainty 

Traditional forecasting models often struggle to account for the complexity of foreign investments, which make up more than half of Austria’s GDP.  

”Historically, the central bank has used very simple methods,” Thomas Lindner explains.  

“For example, they have looked at the growth of German subsidiaries of Austrian companies and applied the same growth rate to the entire sector. Our model improves this by using more detailed data and better methods.” 

By applying machine learning techniques, Thomas Lindner and his team have developed a system that has almost halved the error margin in GDP predictions.  

”Previously, margin of error was around 50%, but we have reduced it substantially. Even if it proves difficult to put an exact number on this before this method has been put into production, it is safe to claim that this is a significant improvement,” he says. 

Economic precision makes fiscal policy more effective 

One of the most significant results of the project is the direct application to fiscal policy.  

In times of economic downturn, accurate early warning signals allow governments to respond with countermeasures such as targeted financial support or stimulus programmes.  

Conversely, improved forecasting during periods of rapid growth and increase in the money supply helps prevent inflation and ensures sustainable economic development. 

”GDP is perhaps the most important economic indicator for a government, Thomas Lindner underlines.  

”Budgets and fiscal policies depend on GDP forecasts because governments must comply with deficit limits, such as the 3% threshold in the EU Maastricht criteria. If GDP is not measured accurately, public spending cannot be planned effectively.” 

Growing international interest 

Thomas Lindner started the project in collaboration with the Austrian central bank, OeNB. Since then, other European central banks and international organisations have also shown interest.  

”There is now an OECD working group exploring broader applications of our method. The Swiss central bank is leading the work with Austria’s central bank as one of the partners,” says Thomas Lindner. 

The potential to expand the model to other countries is significant, but there are challenges.  

”Data is structured differently from country to country, so the model needs to be adapted before it can be applied more widely, but in principle, nothing prevents us from using machine learning to improve GDP forecasts globally,” he says. 

The future of economic forecasting

Going forward, the task is to refine the machine learning model by integrating additional real-time data and a wider range of alternative economic indicators.  

Expanding the model to industry-specific forecasts could provide even more detailed insights, benefiting both businesses and policymakers in the long run. 

The model also plays a role in education.  

”We have partnered with Vienna University of Technology, where students use this method in their dissertations,” says Thomas Lindner.  

”This gives them practical experience with data analysis while helping us develop the model further.” 

By using machine learning to identify patterns in vast amounts of economic data, Thomas Lindner’s work is not just an academic achievement – it is actively shaping the future of economic policymaking.  

The page was last edited by: Sekretariat for Ledelse og Kommunikation // 05/12/2025