Despite the fact it has been recognised since Aristotle that ethos and credibility play a critical role in many types of communication, these facts are rarely studied in linguistically oriented AI which has enjoyed such success in processing complex features as sentiment, opinion, and most recently arguments. This paper shows how a text analysis pipeline of structural and statistical approaches to natural language processing (NLP) can be deployed to tackle ethos by mining linguistic resources from the political domain. We summarise a coding scheme for annotating ethotic expressions; present the first openly available corpus to support further, comparative research in the area; and report results from a system for automatically recognising the presence and polarity of ethotic expressions. Finally, we hypothesise that in the political sphere, ethos analytics – including recognising who trusts whom and who is attacking whose reputation – might act as a powerful toolset for understanding and even anticipating the dynamics of governments. By exploring several examples of correspondence between ethos analytics in political discourse and major events and dynamics in the political landscape, we uncover tantalising evidence in support of this hypothesis.