Random errors are not necessarily politically neutral

TL;DR

This paper investigates how random errors in STV elections can introduce systematic bias, disproportionately invalidating certain ballots and potentially altering election outcomes, challenging the assumption of neutrality.

Contribution

It reveals that random errors can cause systematic bias in STV elections due to ballot invalidation patterns, a novel insight into election robustness.

Findings

Random errors can bias election results by invalidating specific ballots.

Ballot invalidation disproportionately affects votes with many preferences.

Systematic bias from errors can change close election outcomes.

Abstract

Errors are inevitable in the implementation of any complex process. Here we examine the effect of random errors on Single Transferable Vote (STV) elections, a common approach to deciding multi-seat elections. It is usually expected that random errors should have nearly equal effects on all candidates, and thus be fair. We find to the contrary that random errors can introduce systematic bias into election results. This is because, even if the errors are random, votes for different candidates occur in different patterns that are affected differently by random errors. In the STV context, the most important effect of random errors is to invalidate the ballot. This removes far more votes for those candidates whose supporters tend to list a lot of preferences, because their ballots are much more likely to be invalidated by random error. Different validity rules for different voting styles mean that errors are much more likely to penalise some types of votes than others. For close elections this systematic bias can change the result of the election.