Suppose a country has a progressive tax system.  If everyone equally consumes government benefits, isn’t everyone with below-median income automatically a net fiscal burden – i.e., a person who withdraws taxes more from the Treasury than he contributes?

Naive analysts usually assume that the answer has to be “yes.”  More sophisticated analysts recognize, however, that some government services are “non-rival.”  Once these goods exist, the marginal cost of allowing one more person to consume them is zero.  National defense is a standard example: If we had a baby boom, no one would say, “We’ll need more nuclear weapons to protect these kids.”  Regardless of your views on foreign policy, the cost of defending (N+1) people is no greater than the cost of defending N people. 

The higher the share of non-rivalrous goods in the budget, the easier it is for low-income taxpayers to carry (or more than carry) their own weight.  Before you compare individuals’ tax payments to their consumption of government services, you should subtract out all spending on non-rivalrous goods.  For example, if 20% of government spending goes to non-rivalrous goods, you should compare individuals’ tax payments to .8*average government spending.

On reflection, though, this analysis is still woefully naive.  In the real world, goods are not either 0% rival or 100% rival.  There’s a continuum.  Some goods are close to the 0% pole, others to the 100% pole.  But most goods are semi-rival – somewhere in the middle.

Consider roads.  Most roads are virtually empty in the middle of the night.  The marginal cost of driving at 2 AM is only a slight amount of wear-and-tear on the asphalt.  During rush hour, however, the marginal social cost of driving skyrockets.  When you drive a 5 PM, you aren’t just slightly damaging the pavement; you’re slightly inconveniencing thousands of other motorists.  During mildly congested hours, the marginal cost is somewhere in between.  Instead of classifying roads as either “non-rival” or “rival,” then, you should admit that they’re semi-rival, then try to accurately place them on the continuum.

In principle, switching from binary classifications (“rival” or “non-rival”) to continuous classifications (degree of “semi-rivalry”) could make our estimates of fiscal externalities more or less optimistic.  In practice, however, the effect is strongly optimistic. 

Why?  Because when empirical researchers use a binary standard, they usually treat “rival” as the rule, and “non-rival” as the exception.  Only goods close to the 0% rival pole – especially national defense, interest on the national debt, and R&D –  go into the “non-rival” category.  Everything else gets tossed into the rival category.  If 20% of spending goes in the non-rival category, this approach implies that spending is 80% rival overall.

With a continuous standard, reclassification has almost no effect on the classic non-rival goods.  Fine, maybe they’re 1% rival rather than 0%.  But reclassification has a huge effect on the so-called “rival” goods.  When you break them down, you could easily find that one-fourth are 25% rival, one-fourth are 50% rival, one-fourth are 75% rival, and one-fourth are 99% rival.  Switching to continuous classification therefore brings overall rivalrousness from (.2*0+.8*1)=80% down to (.2*.01+.2*.25+.2*.5+.2*.75+.2*.99)=50%.  Quite a shift!

But doesn’t this prove far too much?  If most government services are semi-rival, doesn’t the same hold for private goods?  Of course.  Look around.  In many (most?) businesses, the marginal cost of serving one more customer is trivial.  But there’s no reason to run from this symmetry.  In the private sector, people who buy bargain tickets do not “burden” customers who pay full fare.  In fact, due to semi-rivalry, people who buy bargain tickets benefit full-fare customers by making a wider variety of flights profitable.  The same can easily hold in the public sector as well.