Previous: The Underlying Method

Up: The Underlying Method

Next: The Gelman-King Model

Sources of Error in Electoral Data

In principle, the source of error in an experimental outcome can be divided into systematic error, in which the apparatus for conducting an experiment skews the result in the same way each time it is run, and random error, which is different every time an experiment is conducted. In many cases in political science, the estimated error in an experiment cannot be divided into its two component types.

In the application of two-party elections, however, the ``experimental apparatus'' to be considered is an electoral system that changes minimally between elections in its construction. This gives us a method for estimating what fraction of an election's variation is caused by properties of the system, and what changes can be considered to be random in nature.

Under this assumption, two different types of elections can be simulated: those in which the existing system is no longer in place, such as the prediction of future elections, and those in which we consider what would happen whether a particular election was run again using the same systematic conditions, two fundamentally different sets of analyses with a common method of generation.

Quantities of interest in the political science literature can often be explained in terms of the outcomes of hypothetical elections or properties of the election system.