One of the most significant, but least remarked upon, features of this year's budget is Malcolm Turnbull's decision to greatly expand the federal government's involvement in the construction of public infrastructure.
He did so under unprecedented and sustained public pressure from the Reserve Bank, seconded by the International Monetary Fund and the Organisation for Economic Co-operation and Development.
But how could the government be stimulating demand at a time when it still had a big budget deficit it needed to get back to surplus ASAP?
By distinguishing between the deficit arising from recurrent spending on its day-to-day operations, and the deficit arising from its investment in capital works, whose benefits to the community would flow for decades.
With the economy's downturn long past, the government should certainly be striving to get its recurrent finances – summarised by the budget's "net operating balance" – back to a healthy big surplus.
But no such stricture should apply to borrowing to improve the nation's infrastructure – always provided the money is well spent.
There's nothing new about this. The state governments have divided their budgets between operating expenses and investment in capital works for years. The national governments of New Zealand and Canada do the same.
So why haven't the feds been doing it? Because Treasury's never liked the idea. That's why, if you read the budget papers carefully, you find Treasury's found a way to do it and not do it at the same time.
The papers say they've always told us what the recurrent budget balance is, it's just that it's been buried somewhere up the back and called the net operating balance, or NOB.
But Treasury has had to admit that, for reasons that make sense only to accountants like me, the NOB regularly overstates recurrent spending by treating as an expense the cost of the feds' annual capital grants to the states to help with their infrastructure spending.
In the coming financial year, this overstatement is worth more than $12 billion, meaning the true recurrent deficit is actually quite small – $7 billion – and expected to be back in balance in the following year, 2018-19.
So, no great worries there.
For the first time, Treasury has been obliged to reveal clearly exactly how much the feds have been, are, and expect to be, spending on capital works for the 14 years from 2007-08 to 2020-21 (see budget page 4.10).
In 2007-08, the last of Peter Costello's budgets, total federal capital spending was allowed to fall below $10 billion, but generally it's been between $30 billion and $40 billion a year. That's roughly 10 per cent of all the feds' spending.
But here's the big news: in the coming financial year, it's expected to rise to a (nominal) record of more than $50 billion, up from about $43 billion in the year just ending.
This will represent 12 per cent of total federal spending, and be equivalent to 2.8 per cent of gross domestic product.
Again for the first time, the budget papers give us the breakdown of the feds' total capital spending. First there's "direct capital investment" of $13.5 billion, which is mainly spending on defence equipment.
Next is "capital grants" of $14.2 billion. This is money given to other entities – predominantly, the state governments – to help them pay for their own capital works spending, mainly roads.
Last is an odd one, that Treasury usually prefers us not to notice: "financial asset investment (policy purposes)" worth $22.9 billion, up almost $6 billion on the year just ending, and the main cause of the coming big increase.
What's that financial asset investment thingy? It goes back to 1996 and a loophole Treasury carefully built into the budget figuring at the time of the Charter of Budget Honesty (!) and the introduction of the "underlying cash balance" as the preferred measure of the budget's deficit or surplus.
Get this: if the government simply pays some private construction company to build some infrastructure for it, the cost is counted as part of the underlying cash deficit.
But if the government sets up its own company and gives it the same money, in the form of share capital or a loan, so the company builds the infrastructure (or pays another company to do it), the cost isn't counted in the underlying deficit.
Rather, it's tucked away in the "headline cash balance" that few people notice (see budget page 3.36). (The other big item stashed in the headline deficit is the net increase in the stock of HECS HELP student debt owed to the government, expected to be an extra $8 billion in the coming year.)
It's by this means that the Labor government was able to spend many billions constructing the national broadband network without a cent of it showing up in the underlying deficit.
In the coming year, the Turnbull government expects to buy $1.5 billion more in NBN shares and lend it $9.3 billion – all to finance further construction spending.
As well, it's setting up a company to own and build the second Sydney airport, and another to own and build the Melbourne to Brisbane inland freight railway.
Combined, these two new projects are expected to cost $1.8 billion in the coming year, rising to an annual $3.2 billion in 2020-21.
But if spending on infrastructure is now regarded as "good debt", why is Treasury still using this legal hair-splitting to conceal the cost of the new infrastructure spending push?
Because it's fighting a rear-guard action. Although it's agreed to give the NOB "increased prominence" in the budget papers, the underlying cash balance "will continue to be the primary fiscal aggregate".
And just to prove Treasury's lack of repentance, no modification has been made to the wording of the government's "medium-term fiscal strategy" to "achieve budget surpluses, on average, over the course of the economic cycle".
