Voter interaction with a candidate-based electoral system: the varying operation of STV in party-based elections

Although STV is probably the oldest form of proportional representation, having been part of the by-laws of the Society for Literary and Scientific Improvement of Birmingham in 1819, and used in a public election (of the Adelaide City Council) in 1840, it is not widely studied outside of the three countries (Ireland, Australia, Malta) that use it for national parliamentary elections - notwithstanding its somewhat wider use in sub-national elections in other places (e.g. across the Australian states, parts of the UK and the US, and New Zealand local elections). In the recent spate of electoral reform proposals in Canada, the recommendation of British Columbia Citizens Assembly for STV is the one that came closest to being adopted (defeated only because a 60 referendum vote was required). Although the results of modern STV elections are reasonably proportional, especially given that they usually take place in low magnitude districts, in a technical sense it is a misnomer to identify STV as a proportional system at all. That is, if proportional representation means that the formula for allocating seats among parties is based on the proportion of votes they receive, then STV, in which such a proportion is never computed, does not qualify - and that is probably why it was “invented” before list PR systems: appropriately for a pre-party era, it does not require that there be parties. This paper intends to pursue two lines on inquiry, both deriving from the fact that the proportionality of the results of STV depends entirely on the way in which voters rank the individual candidates. (If every voter plumped for a single candidate, STV would degenerate into SNTV; if every voter ranked all the candidates of a single party and then stopped, STV would produce exactly the same results as Droop quota largest remainder PR.) First, what voters do can be influenced by institutional decisions: they may be required (or not) to rank some or all the candidates; they may be given (or not) party cues (e.g., party labels attached to candidates otherwise listed alphabetically; candidates listed alphabetically grouped by party) or even given a short cut that allows the easy casting of a straight party ticket (voting “above the line” in Australia). Particularly if we move beyond exclusive concern with national parliamentary elections, there is a wide variety along these dimensions, in a wide variety of settings, allowing us to investigate the partisan consequences of these variations. Second, what voters do can also be influenced by party and candidate strategy. There is a range of possibilities here. In the first instance, parties can (or not; in Australia they must) make explicit agreements to encourage the mutual transfer of votes once the candidates of the voter’s more preferred party are exhausted. In the second instance, candidates can campaign primarily as party (wo)men, presumably encouraging transfers to remain within party, or primarily as individuals, thus encouraging transfers across party lines. And, of course, these two can be expected to interact, with for example personal campaigning most likely in systems with the weakest partisan ballot cues. Finally, there are important dynamics in terms of dealing with intra-party transfer strategies: in Ireland, for instance, the strategy is one of spreading preferences as equally as possible across candidates, whereas in Australia instead the parties tend to ‘plump’ all first preferences on the top-placed candidates, with preferences cascading downwards to the remaining candidates in each successive count. This summary confirms what has been noted many times over in the social choice literature, namely that STV systems vary quite distinctly (in design and operation) from one context to another. This paper will start by setting out the large range of variations in how voters might interact with STV resulting from the different institutional and strategic circumstances. The paper’s key objectives are twofold. First, we will use aggregate data (election returns) to assess variations in party-based proportionality profiles (using both standard ‘first-count’ measures as well as final-count ‘party preferred’ measures) and inter-party transfer patterns. Second, we will make use of existing survey data (such as the Australian and Irish election studies) to tease out how voters interact in detail with STV systems under particular institutional and strategic circumstances. The overarching question driving the analysis is just how ‘party coated’ STV can be in these different contexts.