Frankfurt School launches real-time Twitter-/X-based Inflation Expectations Index

• The index evaluates German tweets starting from 2011 using Natural Language Processing
• It makes real-time short-term inflation expectations for Germany available for the first time
• The index responds to monetary policy shifts, rising post-easing and falling after unexpected tightening

Researchers from Frankfurt School of Finance & Management have developed a Twitter-/X-based Inflation Expectations Index. The index evaluates German tweets using advanced Natural Language Processing and creates a daily inflation expectations index – making real-time short-term inflation expectations for Germany available for the first time.

The research project developing the index is based at the new Centre for European Transformation at Frankfurt School. To create the index, in a first step, all German-language tweets that include keywords related to prices and inflation are downloaded. To clean them from bot activity that could potentially bias results and to remove further noise from the sample, the index uses machine learning techniques to split text corpus into different topics and select the tweets that belong to inflation topics. These tweets are classified as “up”, “down”, or “neutral”, using a pre-trained neural network language model which is fine-tuned on the specific classification task. To generate enough training data for this fine-tuning process, the index uses the ChatGPT API. For each day, all tweets labelled as up and down are aggregated to a daily inflation index. In addition to a German Inflation Expectations Index, the model is able to build regional indices, as well.

Based on German tweets starting from 2011, the new real-time inflation expectations index already proofed to strongly correlate with realised inflation rates over this period, for example from Bundesbank surveys. Furthermore, it provides additional explanatory value and improved forecasting accuracy beyond existing measures for quantitative inflation expectations. Notably, the index responds to monetary policy shifts, rising post-easing and falling after unexpected tightening. The influence is particularly pronounced from tweets by private users during recent periods of elevated inflation. Consequently, this Twitter-/X-centric index offers a valuable tool for stakeholders and policymakers to promptly assess prevailing inflation sentiments.

Inflation expectations are crucial in determining consumption, savings, and investment decisions of firms and households and, therefore, the development of the economy. Furthermore, they provide an important source for central banks to forecast future realised inflation. Knowing the expectations regarding future inflation is key for central bankers for adopting appropriate policy measures. However, inflation measurement so far was only available with a considerable time lag.

Professor Sascha Steffen, Professor of Finance and Director of the Centre for European Transformation at Frankfurt School, said: “Our Inflation Expectations Index is a strong proof how Frankfurt School and our Centre for European Transformation combine excellent research with a high level of practical relevance on challenges affecting the economy, politics, and society. We plan to extend our index analyses, for example by exploring the link between inflation and private consumption. In addition, we can develop indices for other countries, as well.”

Johannes Müller, Global Head of Research, DWS, added: “The newly developed Inflation Expectations Index is a great example how academia can add real value and practical relevance not only for financial institutions but the economy as a whole. This is important in the context of Europe’s ongoing transformation and challenges and the reason why DWS supports academic excellence with practical relevance in the Centre for European Transformation at Frankfurt School, addressing the challenges of the 21st century and fostering a sustainable transformation and growth in Europe.”