The Marginal-Seat Premium

Australia elects one member for each of 151 electorates, and on paper every one of them is up for grabs at each election. The electoral map tells a narrower story. Set the two-party margins from the 2022 result side by side, and most of the country turns out to be settled before a single advertisement runs. The real contest — the seats that can change hands, and with them the government — is fought in a few dozen places, many of them the same places election after election.

That geography is not a curiosity. It is the incentive structure of Australian politics. If only a small number of seats are ever in play, a government that wants to stay one has an obvious reason to concentrate its attention, and its discretionary spending, exactly there. The auditors keep finding that this is what happens. This is what the electoral map that makes it rational actually looks like.

Where the contest actually is#

Start with the margins. The Australian Electoral Commission publishes, for every division, a two-party-preferred figure — the Labor-versus-Coalition split once minor-party preferences are distributed. The distance from 50% is the margin, and by the AEC’s own convention a seat under 6% is marginal, 6 to 10% is safe, and anything above that is very safe.

Of the 151 divisions, 47 were marginal. 104 — more than two in three — were safe or very safe. Narrow the field to the 135 seats the major parties actually hold, and only 36 of them, about one in four, are marginal. At the sharp end, just 14 seats sat on a margin under 2%: the entire theatre in which the outcome was genuinely uncertain, out of a parliament of 151.

Fourteen of the 151 divisions were decided by less than two per cent. The other 137 were, in two-party terms, already spoken for.

AEC two-party-preferred results, 2022

How few seats decide a government#

The consequence is a government balanced on a very thin edge. Labor holds office with 77 seats. The majority line is 76. That leaves a buffer of exactly one seat — and the seats that make up the buffer are among the closest in the country.

Labor’s four most marginal seats — Gilmore (0.17%), Lyons (0.92%), Lingiari (0.95%) and Bennelong (0.98%) — all sit under one per cent. A uniform two-party swing of just 0.92%, enough to take the second of them, would have pushed the government below 76 and into minority. Put the other way: the difference between majority government and a hung parliament was under one vote in a hundred, spread across a couple of electorates. Twenty seats changed hands in 2022; the winner of the whole contest was settled in far fewer.

This is the structural fact beneath the campaign. A party does not need to persuade the nation. It needs to hold its own marginal seats and take a handful of its opponent’s — perhaps two dozen electorates in all. Everything else is, electorally, decoration.

The competition is not spread evenly#

Nor are those contestable seats scattered at random across the map. The share of a state’s seats that are marginal ranges widely.

Among the large states, Victoria was the least contested — a quarter of its 39 seats marginal — while Western Australia (40% of 15) and Queensland (37% of 30) were the most. The two territories sit at the extremes only because they have two and three seats between them. The point is not any single figure but the shape: electoral competition pools in particular states and particular suburbs, and a government reading this map knows precisely where its parliament is won and lost.

The premium#

Which is where the money enters. If a narrow set of seats decides power, directing public spending to those seats is a rational, if improper, use of incumbency — and Australia’s audit office has documented it doing so more than once.

In 2020 the National Audit Office examined the Community Sport Infrastructure Program — the affair that became known as “sports rorts.” It found that the awarding of grants was not the merit process it appeared to be: a parallel assessment run out of the minister’s office ranked candidate projects against the electorates they sat in, and the funded list diverged from the one the sporting agency had recommended. The following year the same office turned to the Urban Congestion Fund’s commuter car parks and found the selection of projects was not demonstrably merit-based, with sites concentrated in Coalition-held and targeted electorates rather than where the transport case was strongest.

The pattern is bipartisan in principle even where the documented cases are not: the map rewards whoever holds power. Discretionary grant programs, with ministerial sign-off and loose published criteria, are the instruments best suited to it. The electoral geography is the standing invitation; the audits are the record of it being accepted.

What the data did not show#

One prediction registered in advance did not survive contact with the numbers, and it is worth saying so plainly. The assumption going in was that safe seats breed disengagement — that where the result is a foregone conclusion, voters tune out, and informal voting climbs. If true, informal rates would rise with the margin.

There is no such gradient. Across all seats the correlation between margin and informal voting is essentially zero (Spearman ρ = +0.04); among Labor-held seats it runs slightly the other way. Mean informal voting is near-identical by party — 5.4% in Labor seats, 5.3% in Coalition seats.

The pattern in the data is a cluster instead. The highest informal rates in the country sit in a tight group of western-Sydney seats — Blaxland (10.8%), McMahon (10.6%), Fowler (10.5%), Werriwa (9.8%), Watson (9.7%) — some very safe, some marginal. It is a pattern about place and population, not about electoral safety; the AEC’s own research links high informal voting to ballot complexity and to areas with more residents from non-English-speaking backgrounds. The tidy theory was wrong, and the map that replaced it is more specific and more interesting.

What could explain this instead#

A narrow band of marginal seats, and a record of money flowing toward them, is not by itself proof that the electoral map causes the spending. Several other readings fit the same facts. Need and marginality overlap. Marginal seats are disproportionately outer-urban and regional — the places that plausibly have the weakest sporting facilities or the worst commuter congestion in the first place. Some share of the grants that landed in marginal seats may have landed there on merit. The audits’ point is narrower and stronger: that the documented processes were not merit-based, whatever the underlying need. Marginality is endogenous. Seats are marginal partly because the parties contest them hardest, and they contest them hardest partly because they are marginal. The map and the behaviour shape each other; a single election cannot separate cause from effect. Two elections are a snapshot. Boundaries move at each redistribution, and a seat safe in 2022 can be marginal by the next poll. The claim here is about the standing shape of competition, not about any individual seat’s fate. The two-party lens understates the crossbench. Sixteen seats are now held by Greens and independents whose contests the two-party margin does not capture. The competition map is, if anything, broader than the pendulum shows — but broader in ways that cut against, not toward, the major parties’ traditional targeting.

None of these dissolve the core facts. Competition is scarce and concentrated; government turns on a handful of seats; and the audit record shows discretionary money bending toward the places where power is decided. The descriptive claim — the contest is narrow, and the spending follows it — rests on firmer ground than any single causal story about why.