Curated measures

The Curated Measures Library offers research-ready datasets with firm-level measures, maintained by NL Analytics — an alternative to building your own measure when an established construct fits the design.

What the library contains

Each curated measure is a maintained, refreshed dataset built on the same earnings-call corpus as your own searches. Entries state the research the measure is based on, how many measures the dataset contains, and the date of its last data refresh. Examples include:

• The NLA Corporate Country Risk Index — perceived risk for 45 major economies and its transmission across borders, based on research published in The Review of Economic Studies.
• Political Risk and Sentiment — firm-level political risk (PRisk) and sentiment (PSentiment) measures in the spirit of Hassan et al. (2019), based on research published in The Quarterly Journal of Economics.
• Thematic indices such as a Macro Attention Index, an AI Investment Barometer, cyber and geopolitical risk indices, and Brexit exposure measures.

Open an entry in the library to see its description, source research, and access details.

When to use a curated measure instead of your own search

Use a curated measure when the construct is already established in the literature and you want the maintained, citable version: the dictionaries and method follow the source papers, and the dataset refreshes with the corpus.

Build your own measure when the construct is novel, needs different terms or exclusions, or must be defended on your own validation. The two approaches combine well: a curated measure can serve as a benchmark for a custom measure of the same theme.

The political-risk and country-risk families are also available as weighted dictionary searches, which run the underlying weighted dictionaries through the regular search workflow.

Interpreting curated measures

Curated measures are text-derived, like everything on the platform: they measure discussion in earnings calls, not outcomes. The interpretation caveats in What the metrics do not mean apply. For research use, cite the source paper named in the library entry and record the dataset's refresh date alongside your export.