Scotland Polling Error Analysis
Click the button below to run 1000 simulations of the Scottish Parliament election using randomised vote shares centred around the current polling average. It then applies further layers of variation by tweaking every party's swing in every constituency based on local factors, including geographical area, urban/rural characteristics, and support for independence in the 2014 referendum.
We use the Monte Carlo method to demonstrate the degree of uncertainty that's inevitable in election projections and show how even small polling errors, and local factors can lead to significantly different outcomes. In many cases the difference between an excellent year for a party and a terrible year is well within a normal polling error.
Click the button below to run the analysis. Results are calculated using the latest polling data.
This simulation uses a normal distribution of errors from the polling average which means that parties are just as likely to overperform as they are to underperform their average. The vast majority of misses will have each party's vote share within 2.5% of the Nowcast but there will be some rare instances where the misses are much larger. The bigger the miss, the less likely it is to occur. We also randomise the degree of tactical voting, sometime certain parties might benefit by considerably more than is expected, in other cases no tactical voting will happen at all.