Each MacroLab project pairs a research paper with a working application. Below is what we are building — and the research behind it.
A real-time business-cycle indicator read directly from Norwegian news. Language models classify the daily news flow into demand and supply narratives, and the aggregated signal tracks GDP growth quarter by quarter.
Every reading is auditable: each quarter links back to the ranked articles that drive it, from the headline number down to the individual news stories.
Uncertainty indices built from about 500,000 historical text observations spanning 1988–2023. A topic model identifies 80 themes in the news flow; each series tracks how prominent one theme is over time.
The aggregate index summarises daily cross-topic uncertainty, with crisis episodes — from the banking crisis to the pandemic — standing out clearly.
An experiment in automated macroeconomic research: AI agents carry out complete research workflows — from data collection to estimation and reporting — under transparent, reproducible protocols.
The published application demonstrates the workflow end to end, and will grow as the research programme matures.