Big 'C' or Small 'c'? Conservatism, Risk, Late Swing and Polling Errors

The 1992 and 2015 general elections in the UK share three common features: a close race, a Conservative incumbent government, and a major polling understatement of that incumbent government's eventual vote share. While there are sampling and response bias issues at play, too, late swing looks a plausible explanation for at least part of these hefty polling errors. Such late swing is often supposed to be to the 'safe' option. But the particular understatement of incumbent support when the Conservatives were in power raises the "big 'c' or small 'c'" question. Do voters also see Conservative, and right-wing parties in general, as somehow safer? In this Paper, we address that question in two ways. First, we draw on Jennings and Wlezien's (2015) enormous dataset of poll results, isolate the final polls at each election, and calculate how far the polls systematically understate the eventual vote share of a) the incumbent, and b) the main right-wing contender for power. Second, for another angle on causal inference, we conduct a survey experiment on British voters, manipulating the ideological position as well as the incumbency of parties and observing the effects on evaluations of that party as a 'safe pair of hands'. Together, the two projects will give us an insight into the nature of risk as perceived by voters.