Recent years have shown a growing interest in the analysis of audio recordings to assess nonverbal characteristics of political speech. Currently, this development is hampered by the very limited availability of processed and ready-to-use audio data. This currently leaves audio-as-data approaches inaccessible to many applied researchers. We introduce a new data set of processed audio recordings of more than 20,000 floor speeches in the German Bundestag spanning three legislative terms between 2011 and 2018. Following existing approaches, the data set includes vocal pitch as a measure of legislators' emotional intensity during a speech, but allows the extraction of other quantities as well. The data is harmonized with existing text of speech corpora and thus enables researchers to connect text-as-data approaches with audio data.

Because previously existing sources of structured audio data exclusively focus on the US, our data facilitates the testing of theories about legislative speechmaking under an alternative institutional setting. It further paves the way for the integration of audio data into more comparative research programs on legislative speech.

Demonstrating the utility of this new source, we show that women legislators in the German Bundestag speak with greater emotional intensity about women-related issues as compared to their male colleagues and other issues. This replicates an earlier result from the US House of Representatives (Dietrich et al. 2019) in a new context. Taking emotional intensity as an indicator of issue commitment, this adds confidence to the finding that an increase in womens' descriptive representation fosters their substantive representation.