“…When polls are conducted at regular intervals, it seems natural to use a statistical model to extract the full potential of the information that is contained in these time series of poll results by using them to forecast public opinion beyond the most recent poll date. However, long time series of poll data are scarce, and, to the best of our knowledge, all previous studies that have analysed time series of political opinion polls have used data that were observed at the monthly frequency or lower; see, for example, Box-Steffensmeier and Smith (1996), Byers et al (1997Byers et al ( , 2000Byers et al ( , 2002, Dolado et al (2002Dolado et al ( , 2003 and Jones et al (2014). Authors of these studies have noted that an ideal data set would have all observations contained within a single government regime spanning only one political cycle, while providing a sufficiently large sample to conduct meaningful statistical analysis.…”