Recent failures to predict landmark elections such as the 2016 U.S. Presidential race and the Brexit vote have, rightly or wrongly, rattled public confidence in election polls and the survey method at large. Why polling errors occurred in a given election but not in others often remains a matter of speculation. The aim of this paper is to develop a contextual understanding of polling errors and their triggers. To do so, we cover a broad range of features of the electoral contest which pundits and scholars have suspected of encouraging polling errors, including candidacies, mobilization, polarization, and electoral conduct. Whereas most previous studies of U.S. Senate elections confound poll bias and sampling variance, we adopt a Bayesian hierarchical modelling approach that allows us to disentangle systematic and random error at the election level, and then to link election features to specific error components.