Working on sentometrics

Sentometrics is a small field with a low barrier to entry: the methods are documented, the software is on CRAN, and several indices are published here for anyone to use. What follows is what you need to start.

Start here

  1. Read the survey — it maps the whole chain from corpus to econometric model, and tells you which step your question actually belongs to.
  2. Install sentometrics and work through the JSS vignette.
  3. Pick an index from Indices & data and reproduce a result from the paper behind it. This is the fastest way to find out whether the field suits you.
  4. When you want to go wider, the survey maps the literature and its open problems.

Data you can build on

Every index we publish is documented, versioned, and free for academic use. They are deliberately reusable as starting points:

  • MCCC — daily US media climate change concerns, 2003 to present, with topical and thematic subindices. The most reused of our datasets.
  • EPU Quebec — a century of economic policy uncertainty from French-Canadian sources.
  • EPU Belgium — daily, French- and Dutch-language, from 2001.
  • US Topical Economic Sentiment — daily sentiment across economic topics.

Topics that are open

These are directions we think are worth a thesis or a master’s project. They are not assignments — bring your own if you have one.

  • Sentiment under large language models. What survives when the lexicon is replaced by a generative model? The aggregation and identification problems do not disappear; they change shape.
  • Concern indices beyond climate. The MCCC recipe — risk plus sentiment, weighted by attention — has been extended to social risk. Other domains are open.
  • Text selection and coverage. Which sources enter a corpus, and how coverage shifts over time, can move an index more than the sentiment method does.
  • Indices in languages other than English. Our EPU work in French and Dutch showed the standard recipes do not transfer unchanged.
  • Monetising sentiment. From index to implementable strategy, net of costs.

Get in touch

  • PhD — HEC Montréal runs a PhD in Financial Engineering. Doctoral supervision also happens at Université de Sherbrooke and Vrije Universiteit Brussel.
  • Master’s theses and supervised projects — at HEC Montréal, for students enrolled there, including projects with industry partners.

Write a short note describing what interests you and your timeline. See the team for who works on what.