Curated Measures

Political Risk and Sentiment

Data through Sep 30, 2026 Updated Oct 19, 2026 9 series

Firm-level political risk (PRisk), political sentiment (PSentiment), and eight topic-specific political risk measures extending Hassan et al. (2019).

Based on peer-reviewed research The Quarterly Journal of Economics, 2019

We’ll email you when the data is updated. No account required.

At a glance

Coverage
About 16,800 companies in 94 countries
Period
Jan 2002 – Sep 30, 2026
Updates
Quarterly
Files

CSV: time series, firm panel

Source

English earnings calls

Overview

Political Risk and Sentiment is the maintained continuation of the call-level political risk (PRisk), political sentiment (PSentiment), and eight topic-specific political risk (PRiskT) measures introduced in Hassan, Hollander, van Lent, and Tahoun (2019, The Quarterly Journal of Economics). Recomputed under a single methodology from 2002 forward, it provides one continuous panel so you do not need to splice the historical firmlevelrisk.com archive (which ends in 2021-Q2) onto newer updates.

Use it to

  • Measure firm-level exposure to overall political uncertainty (PRisk) and directional political tone (PSentiment).
  • Decompose political risk into eight policy topics: Economy, Environment, Health, Institutions, Security, Tax, Technology, and Trade.
  • Extend empirical panels built on the firmlevelrisk.com archive past 2021-Q2 without a methodological break.
  • Study how firm investment, hiring, lobbying, and stock volatility respond to idiosyncratic political risk.

What it does not measure

  • Measures tf-idf-weighted political language near risk or sentiment words in earnings calls, not objective policy outcomes or lobbying expenditures.
  • Compare changes and episode timing within a series rather than raw levels across different topic dictionaries.
  • Call-level CSV exports ship unscaled weighted counts alongside transcript word counts (nr_of_words); apply word-count scaling and firm-quarter averaging downstream to match the published panel's conventions.

Explore the data

What to look for

In the featured PRisk series, weighted political-risk discussion reaches its pre-GFC low in Q2 2007 (6.4), rises sharply during the global financial crisis (9.3 in Q3 2009) and US fiscal-cliff debates (8.3 in Q2 2013), surges to its all-time peak during the COVID-19 pandemic and US election cycle (12.2 in Q1 2021), and climbs again in Q4 2025 (10.0).

Click the chart to open the series, where you can filter it by headquarters country.

PRisk and PSentiment (v2.1) over time

Date range: January 1, 2002 to September 30, 2026.

Time-series chart for PRisk and PSentiment (v2.1) over time.
Loading figure...

Mean weighted risk matches per earnings call, measured in Weighted matches, shown quarterly using a 4-quarter moving average. Q1 2002 to Q3 2026, 99 data points. Higher values mean more weighted risk discussion, not a probability or comprehensive risk rating. These values reflect weighted risk discussion, not a probability or a comprehensive risk rating.

Explore series

Series

9 total Data through Sep 30, 2026

Browse the related series and open any series for its full query, filters, and methodology notes.

No series match your filter.

Methodology and data

How closely it matches the published series

Computed on the firm-quarters that overlap with the firmlevelrisk.com release (2002-Q1 through 2021-Q2; $N = 310,303$ firm-quarters and 81 quarters):

Series Firm-quarter (gvkey × quarter) Aggregate (quarter)
PRisk 0.91 0.98
PSentiment 0.86 0.95
Topic-specific PRisk (8 topics) 0.81 – 0.91 0.95 – 0.99

Topic measures span Trade (lowest) to Economy (highest) at both levels. Every number in this table is reproducible using the script in our public prisk-correlations GitHub repository, which takes the firm-level panel from this page and the historical CSV from firmlevelrisk.com.

What is the same, and what is different

Verified directly against the original implementation, the following are identical:

  • Scoring rules: The scoring formulas for PRisk, PSentiment, and the eight topic-based PRiskT series (Economy, Trade, Tax, Security, Institutions, Health, Environment, Technology).
  • Counting windows: ±10 words around a risk or uncertainty term for PRisk (±20 words for the topic-specific PRiskT measures), with the center bigram's own tokens excluded, as in the paper.
  • Lexicons: The political bigram library and its tf-idf weights, the Oxford Thesaurus risk/uncertainty synonym set, and the Loughran-McDonald sentiment dictionaries.

Remaining differences affect tokenization, sample mapping, and downstream scale rather than the scoring formula:

  • Tokenization: Punctuation-split word boundaries with digits retained; a small number of multi-word phrases do not match under the current tokenizer.
  • Sample composition: Updated mapping to gvkey accounts for most of the remaining firm-quarter discrepancy and averages out in quarterly aggregates.
  • Transcript length and safe-harbor rules: No minimum-length filter on very short transcripts, and minor differences in one boilerplate safe-harbor rule.
  • Scale and panel structure: Each CSV row is one earnings call carrying earningscallID, calendar date, UTC start time, company identifiers (gvkey, isin, tickersymbol, headquarters country), raw weighted scores, and transcript word count (nr_of_words). Division by word count rather than bigram count, library-weight normalization for PRiskT, and scaling by 100,000 are scale-only and do not affect correlations or firm rankings.

References

Hassan, T. A. & Hollander, S. & van Lent, L. & Tahoun, A., "Firm-Level Political Risk: Measurement and Effects," The Quarterly Journal of Economics, 134(4):2135-2202 (2019)

NL Analytics. (2026). Political Risk and Sentiment [Data set]. NL Analytics. https://apps.nlanalytics.tech/curated-measures/political-risk/

Browse Curated Measures