Firm-Level Political Risk

Download Data

We are pleased to host PRisk and PSentiment, firm-level measures of political risk and political sentiment developed by Tarek Hassan (Boston University), Stephan Hollander (Tilburg University), Laurence van Lent (Frankfurt School of Finance), and Ahmed Tahoun (London Business School).

Hassan, Hollander, van Lent, and Tahoun (HHLT) use textual analysis of quarterly earnings conference-call transcripts to construct a firm-level measure of the extent and type of political risk faced by individual firms listed in the United States. The vast majority of firms with a listing on a US stock exchange hold regular conference calls with their analysts and other interested parties, a forum where management gives its view on the firm's past and future performance and responds to questions by call participants about any challenges the firm may face. HHLT's approach to quantifying the extent of political risk faced by a given firm in a given quarter is simply to measure the share of the conversation between participants and firm management that centers on risks associated with politics.

HLLT adapt a simple pattern-based sequence-classification method developed in computational linguistics to distinguish between language associated with political versus non-political topics. For their baseline measure of overall exposure to political risk, they use a training library of political text (an undergraduate political science textbook and text from the political section of newspapers) and a training library of non-political text (an accounting textbook, text from non-political sections of newspapers, and transcripts of speeches on non-political topics) to identify two-word combinations ("bigrams") that are frequently used in political texts. They then count the number of instances in which conference-call participants use these bigrams in conjunction with synonyms for "risk" or "uncertainty," weight each bigram with its term frequency, and divide by the total length of the conference call to obtain PRisk- a measure of the share of the conversation that is concerned with risks associated with politics. PSentiment, similarly constructed by conditioning on positive and negative tone words, measures positive and negative news relating to politics.

For each of 9,481 firms listed on a US stock market between 2001 and 2016, HHLT's dataset gives the firm's name, its Compustat GVKey identifier, PRisk, PSentiment, as well as its logical components: Risk, calculated by counting only the number of synonyms for risk or uncertainty, without conditioning on political bigrams, and Sentiment, counting positive and negative tone words. In addition, the dataset lists NPRisk, a measure of non-political risk faced by each firm in each quarter, constructed by calculating the weighted sum on non-political (but not political) bigrams used in conjunction with a synonym for risk or uncertainty, as well as NPSentiment, a measure of non-political sentiment.

In addition, the authors also offer eight topic-based measures of political risks associated with "economic policy & budget," "environment," "trade," "institutions & political process," "health," "security & defense," "tax policy," and "technology & infrastructure". These measures are available from the authors upon request.

The dataset should be cited as: Tarek A. Hassan, Stephan Hollander, Laurence van Lent, Ahmed Tahoun, 2019. "Firm-Level Political Risk: Measurement and Effects," forthcoming in Quarterly Journal of Economics.