\citet{NBERw24029} subject this measure to a battery of stringent validity checks, which includes (1) a human audit of whether the algorithm correctly identifies conversations about risk associated with political topics in the transcript, (2) an inspection of how the measure aligns with political events over time and with sectors that have high versus low exposure to political risk, (3) a set of tests of the correlation between political risk and firm-level outcomes that are a priori likely impacted by  political risk, (4) a set of tests to ensure the measure does not reflect news about the mean, i.e., about the sentiment about political events in a firm’s conference call, and (5) a set of tests to establish that PRiskit is different from non-political risk. 
In this way, \citet{NBERw24029} show that PRiskit is positively associated with implied and realized stock price volatility, and negatively correlated with investments, planned capital expenditures and employment growth. They also show that larger companies with higher Priskit actively manage their exposure by donating more money to the election campaign of politicians and spend more on lobbying. 
Perhaps the most relevant finding for our study pertains to the variance decomposition of Priskit. In contrast with received wisdom that political and regulatory decisions have relative uniform impacts across firms in a developed economy \citep{pastorVeronesi2013}, the political system appears to be a major source of “idiosyncratic risk”. Only 0.81 percent if the variation in PRiskit is explained by time fixed effects (i.e., by aggregate shocks), whereas sector fixed effects and sector-by-time fixed effects explain another 4.38 percent and 3.12 percent, respectively. The remaining 91.69 percent is labeled “firm-level” and consists of 19.87 percent of permanent differences across firms (i.e., firm fixed effects) and 71.82 percent of changes over time in assignment across firms within the sector.