Public spending should rise because it needs to, not just because it can
NationalPost.com – Full Comment
March 17, 2017. ANDREW COYNE
The federal budget is fast approaching, and with it another round of debate about federal spending. Is there too much of it, too little, or just the right amount? To help you out, I’ve brought a couple of handy charts.
Here is what federal program spending looks like over the last six decades: http://wpmedia.news.nationalpost.com/2017/03/na0318_program_spending_1_mf.png
On the other hand this, too, is what federal program spending looks like over the last six decades: http://wpmedia.news.nationalpost.com/2017/03/na0318_program_spending_2_mf.png
Huh? Two charts, both professing to show the same thing, yet with vastly different results. Both draw upon the same figures from the public accounts. Yet one shows spending declining or constant over many years, while the other shows it more or less steadily rising. One suggests we’re spending a third less than we did in the 1970s; the other has us spending nearly twice as much. How can this be?
The difference lies in what spending is compared to in each case: the benchmark. The first measures spending in proportion to GDP. The second expresses it in constant (2016) dollars per capita, that is adjusted for increases in prices and population. That may seem an arcane distinction, but as you can see it makes all the difference. In the first chart, spending seems well under control, even overly so. In the second it looks, if not quite out of control, then certainly running at full throttle. Cue the political arguments.
Which is the better measure: share of GDP or real per capita? Which gives us a more meaningful sense of the size and growth of government? There are good arguments for either, but let me first make the case for real per capita. In essence, the argument is that spending is best measured relative to the things that spending is needed for, that is to provide public services to people. So long as we’re spending the same number of real dollars per citizen from year to year, then arguably we’re providing roughly the same level of services. Any increase beyond that “steady state” would then be discretionary — a policy choice.
So it was that in his 2000 budget, fresh from slaying the deficit, then Finance Minister Paul Martin promised to hold future increases in spending to no more than the rate of inflation plus population growth — “the standard used by most economic commentators.” He didn’t of course; neither did his successors. But had he, and they, kept to that pledge, program spending today would be about $50 billion lower than it is, or about a sixth less than its current total of roughly $300 billion.
By contrast, the share-of-GDP benchmark implies that, so long as spending grows no faster than GDP, it has not grown at all. Spending is here measured not against the volume of services to be provided, but rather the resources available to fund it; it grows, almost automatically, in line with the economy, not because it needs to but because it can. This seems an excessively forgiving standard. Spending has nearly tripled in dollar terms since 2000, but on that top chart it looks as if it’s hardly increased at all.
There’s a counter-argument to this, which holds that to supply a constant level of services, public spending really does need to rise in line with economic growth. Public services, so the argument runs, are peculiarly labour-intensive. To be competitive, public service wages have to keep pace with wages in the private sector. And private-sector wages typically don’t just rise in line with inflation, but also with productivity.
Or at least they do in manufacturing. And wages for services, according to a well-known theorem (“Baumol’s cost disease,” named for the economist William Baumol), tend to keep pace with them, notwithstanding the negligible productivity gains often observed in that sector (it takes a hair stylist, for example, about as long to cut your hair today as it did 100 years ago). So even if the productivity of public services remained flat, wages would still increase in line with productivity gains in the wider economy.
Add it up, and the share-of-GDP benchmark holds: to deliver a constant level of services would require spending to keep pace not just with population growth, and not just with inflation, but also with the growth in productivity — or in other words, with the growth rate of the economy.
It may not be possible for a string quartet to perform any more efficiently than in the past, but businesses are still finding ways to automate common tasks
But government costs are not all or even mostly for labour: wages are typically less than one-fifth of federal program spending. Lots of government spending goes to acquiring goods, for example defence equipment, and much of this has very high economies of scale, meaning it has high fixed or up-front costs and low or even zero marginal costs. So even as population grows, costs do not, or not as fast.
Much more of what government spends, moreover — fully two-thirds of federal program spending — is on transfers: to individuals, to businesses and other groups, and to other levels of government. Much of these, too, are spent on purchasing goods and services from the private sector. Some transfers, such as employment insurance benefits, would tend to rise in line with average wages. But for the most part there’s no reason why this spending needs to expand faster than inflation plus population growth.
It’s not always the case, third, that service-sector productivity is fixed. It may not be possible for a string quartet to perform any more efficiently than in the past, as in Baumol’s most famous example, but businesses are still finding ways to automate common tasks, to replace or augment office clerks with computers, or even higher-end service providers with artificial intelligence. And given the public sector’s notorious inefficiency, it would be reasonable to expect there would be some additional gains on top of what is possible in the private sector, where the low-hanging fruit would presumably all have been picked. Indeed the public sector can piggy-back on private-sector productivity gains by contracting out, purchasing from others what had previously been done in-house.
Neither is it self-evident that public-sector wages must keep pace with those in the private sector. If competitiveness with private wages were the sole factor in public-sector wage-setting, we should expect to see some rough parity between the two. Instead, public-sector compensation tends to exceed that for comparable jobs in the private sector by a wide margin — between a fifth and a third, taking into account wages, benefits and working hours, according to a study by the Canadian Federation of Independent Business.
Governments aren’t obliged to accept that public-sector wage premium. They can, if they choose, find ways to narrow it. If they do then public-sector wages would be expected to grow a little slower than in the private sector, at least for a time. And if they do not then that, too, is a choice.
For all these reasons — fixed-cost purchases that grow only with inflation, other purchases that grow in line with inflation plus population, labour costs that are ripe for savings — it strikes me that real per capita remains a more appropriate benchmark for spending than share-of-GDP, at least for now.
Perhaps there will come a time when the public sector is so miraculously efficient that share-of-GDP will become the better measure. But holding spending to the more stringent real per capita standard seems the best way to hasten that far-off day.