Political skew insights show how a particular podcast's content stands in terms of political leanings. This feature helps our API users and Pro users identify shows based on political bias, helpful for creating targeted campaigns.
How do we define different political skews?
We define political skew along a spectrum of “Left” and “Right.” “Left” represents progressive or liberal views, often associated with the Democratic Party in the USA, while “Right” represents conservative or traditional views, typically aligned with the Republican Party.
Each podcast is also assigned a degree of skew: low, medium, or high. This reflects how strongly the podcast aligns with the political skew. However, it is not indicative of how far or extreme the political views are.
Here are the different skew tags:
Neutral/Mixed: The podcast's content is either politically neutral or there is no clear preference for one party over another.
Low Left Skew/Low Right Skew: The podcast's content slightly leans to the left or right.
Medium Left/ Medium Right Skew: The podcast's content moderately leans left or right.
High Left/High Right Skew: The audience podcast's content leans left or right.
Unavailable or "–": This means we do not have enough data about the podcast to accurately predict political skew.
Which podcasts have political skews?
Currently, political skew data is available for podcasts in the English-language and mainly focuses on U.S. audiences. If a podcast audience or content is mostly non-U.S.-based, we do not display political skew data to prevent providing incorrect predictions.
How do we calculate political skews?
At Podchaser, we’ve developed a sophisticated approach to predict political skews for podcasts, combining advanced models, unique data, and proprietary techniques. Here’s how it works:
Our prediction process begins with large language models, trained with a mix of extensive datasets from known entities and our own curated data. To refine these models, we used a targeted fine-tuning process, feeding them a dataset focused on political news. This helped the models get better at understanding political bias.
Once we developed an accurate bias-predicting model, we applied it to a specialized set of labeled podcast data. By experimenting, we discovered that analyzing the last five episodes of a show gave us the best insights. Finally, we used a multi-step prediction approach, allowing the model to evaluate skews in a way that’s thoughtful and nuanced.
This combination of custom data, fine-tuning, and a layered prediction process gives Podchaser unique accuracy in capturing political leanings across podcasts.
How accurate is political skew?
As of October 2024, overall accuracy of our political skew data was found to be 96% when compared to validation datasets, reflecting a 57% improvement over standard predictions from general models like GPT.
While our skews offer a strong predictive insight, they’re meant as indicators rather than absolute values. For the most definitive view, we suggest contacting the podcast directly.