I think the answer is a whole lot simpler than that. Thoma’s speaking on the normative – that, as we become a richer society, we ought to spend more on social services for the poor. I agree with that. But in fact one doesn’t need to share my progressive views to see that G/GDP must rise unless we tempt a catastrophically – yes, catastrophically – amended social contract. In other words, the answer is positive.

Conservatives are right. Government sucks at providing most goods and services. Here’s a list of some things with which Uncle Sam should never involve himself:

• Entertainment and technology.
• Apparel
• Food. (For the most part).

These are all left better to the private sector. On the other hand, due to economies of scale, monopsony advantage, and network effects, the government has a comparative advantage in the provision of:

• Medical care.
• Retirement.
• Defense.
• Infrastructure.
• Education.

Government’s comparative advantage above derives in part from standard economic analysis, but also a public, bipartisan conviction that these services should be provided by a robust public sector. This is not only a liberal consensus.

Now, the real gross domestic product is modified from its nominal counterpart with what economists call a “GDP deflator” from a given base year. It broadly tracks inflation. If price levels increase 10% in a given year, a 12% increase in nominal GDP isn’t too impressive. However, it is the broadest measure of inflation on the aggregate of (C + I + G). But take a look at individual price indices:

The above graph destroys any hope of a smaller government. Since 1988 education (purple) has increased almost seven fold. Healthcare (red) five fold. Food and beverage prices have only doubled, apparel costs have flatlined, while technology and entertainment prices have plummeted. Basically prices for everything the government is good for have a positive slope and everything it’s bad for have a negative slope. I don’t think any other graph could more clearly explode hopes for a smaller, or even flatlined, government.

The past thirty years has not been kind to the median worker. But all inflation isn’t created equal. It all comes down to relative prices. Items consumers spend on (C/GDP) benefit most from rapid globalization and technological change (Netflix… iTunes… Moore’s Law… etc.) and have experienced either incredibly slow inflation or even deflation. Things government can and should spend on (G/GDP) has experienced a disproportionately higher inflation. We can talk about “bending the cost curve” or whatever but a) nothing will bend it near consumer inflation and b) nothing will stop America from aging.

Fact of the matter is – as hard as evidence might be for the right-wing to swallow – so long as medical and educational price inflation outpaces the GDP deflator, G/GDP must rise to maintain the real value of government provision. That is to say that if we maintain a small government – forever at 20% of GDP – its real, inflation-adjusted value will continue to decline.

This is the level at which we must discuss the size of government. Once we fall into the realm of normative, conservatives will stress the inefficiencies of bureaucracy and ideals of free markets. But even at the positive, nothing can halt growth of government. This scares me because one of America’s flagship political parties is hell-bent not only on preventing growth of G/GDP, but also shrinking government despite hugely inflationary pressures to the contrary.

That scares me because I see no way to reconcile our current social contract with the demands of a smaller government. Needless to say, part of the inflation is secular and will ebb as the population ages into an equilibrium. It will ebb as technology moves from entertainment to medical care. It will ebb as online education revolutionizes the university. But it will ebb at a higher level than 1988 – and the size of government must reflect that fact of nature.

Without feeding the beast we will be starving ourselves. And it doesn’t take a progressive ideologue to see this.