We can almost get there using available data, and with a bit of determined accountancy we could probably get the whole way.

An examination of budget papers can largely explain the 14 September statement by Defence Minister Richard Marles that, by NATO’s measure, Australia is spending not 2.03 percent of GDP on defence but 2.8 percent. The difference is mostly made up in Defence pensions, which NATO would count as military spending but Australia does not. Add in the effect of different accounting rules for equipment purchases, and it’s fairly straightforward to find at least 2.6 percent.

Marles said, ‘if you look at the way in which NATO accounts for its own spending in terms of percentage of GDP, based on that metric, our spending on GDP today in terms of defence is around 2.8 percent.’

This was no offhand remark, since he repeated it on that day and the next.

The relevant political context is clear: the current US administration talks about defence spending in percentages of GDP to its allies and expects an answer in kind.

Apart from the reference to NATO accounting, Marles didn’t say exactly how he had got to 2.8 percent of GDP for Australia ‘today’, which we can take as meaning fiscal 2025–26. So let’s have a go at covering the 0.8-percentage-point gap.

NATO has a bespoke method of calculating defence spending. It used it in June when it set an alliance-wide spending target of 3.5 percent of annual GDP by 2035 (plus 1.5 percent for vague ‘defence-related’ spending). Its method differs from how member states generally report their defence expenditure, sometimes by a lot. The British Ministry of Defence, for example, reported spending £53.9 billion in the 2023–24 financial year, but NATO assessed Britain as having spent £61.2 billion on defence in the same period.

Much of the difference arises because the NATO definition includes pensions for defence personnel, both uniformed and civilian. Some countries do include pensions in what they call their defence spending. But Australia’s usual headline defence budget, the $59 billion in consolidated defence funding (on the bottom line of Table 4a of this year’s Portfolio Budget Statement [PBS]) excludes them.

The other big complication is that NATO measures spending as payments are made, not when financial liabilities are incurred and assets depreciate. That matters especially for big capital assets, such as warships, infrastructure and software. In an accrual-based system such as Britain’s or Australia’s, the cost of a $10 billion ship is spread out over its life. It gets reported as, say, a $400 million expense each year for 25 years. In a cash-based system like NATO’s, the full $10 billion would be reported as the ship was built and the government made progress payments. Then nothing would be counted after that. So it’s quite possible for a NATO measure of Australian spending to vary by billions from the budget papers without an iota of real-world difference.

How Marles’s hypothetical got to 2.8 percent of GDP (roughly $75 billion) boils down to this question: how much cash, between 1 July 2025 and 30 June 2026, will the Australian public sector exchange for military equipment, wages and contract services? That means guns, bullets, aircraft, trucks, spare parts, buildings, research and development, office chairs, the labour of forces personnel, pensions to former and current staff, and so on and so forth. It’s not about what expenses Defence expects to take on, with or without paying that year, and it’s not about the appropriations that Parliament legislates—which are the figures ASPI usually looks at in its Cost of Defence budget brief. This time, it’s just about the cash.

How the minister’s office worked out this cash estimate is not clear, but some useful figures are deep inside the PBS. Australian government agencies do estimate what they expect cash movements to be over the budget period. Let’s take a look.

$69 billion is about 2.6 percent of GDP. So we’re most of the way to 2.8 percent.

Unfortunately, so far we’ve covered only the easy parts. A figure calculated to the NATO standard would exclude some hard-to-gauge things that are somewhere in our numbers, and it would include some that are not.

For instance, Australia’s cash spent investing in defence industries, here or abroad, would have to be taken out: it would count as defence-related spending for NATO purposes, not defence spending proper. (Had Marles intended to measure Australia’s spending against NATO’s 5 percent overall target, defence-industry money could be left in.) The PBS generally does not provide information granular enough to nail down investment in industry, though we can be fairly sure it’s a significant part of what the Australian Submarine Agency spends its money on.

The PBS is also of very limited help to us in identifying the remaining $4 billion or so—though someone surely could do it if armed with the detailed numbers and a determination to put a ‘defence’ classification on whatever could be justified.

This could be done because NATO would allow such classification of money paid in other portfolios for goods or services that contributed to defence. Examples might include the health department’s development of regulations related to nuclear-powered submarines; Department of Education grants for dual-use technology research; and the Bureau of Meteorology’s provision of weather services to the ADF. But we could count those only in proportion to Defence’s share of that use, which wouldn’t be easy to estimate.

We’d also be able to count non-military forces such as the Australian Border Force or even the Federal Police, but only to the extent that they were equipped as a military force and could realistically be deployed abroad in support of the ADF.

Getting to the 2.8 percent line would require a considerably deeper and broader budget analysis than we have room for here, so this is where we leave it for now.

In government accounting, there are often several numbers that can legitimately claim to be ‘defence spending’. The NATO measure is just one. It’s not deceptive, at least not intentionally. It’s simply meant to provide a consistent basis on which the NATO allies can put a top-line number on their ‘inputs’ to defence, and it does that tolerably well. But inputs don’t tell you much about outputs. Achieving a particular GDP ratio is no guarantee that the money’s spent wisely or well, let alone of whether it’s translated into real capability.