Should the Cost of War Include Interest Payments?
When tallying up the "total costs" of a government operation, proponents will typically include only the most obvious items, whereas opponents will throw in the kitchen sink. The obvious motivation is to make the operation look relatively cheap (for the proponents) or horribly expensive (for the opponents).
In such exercises, one of the tricky items is the government's interest payments on debt that was used to pay for a particular operation. For example, when assessing the "total cost" of the Iraq invasion to date, should we estimate how much of the federal government's interest payments this year are due to the growth in the federal deficit necessary to pay for the invasion in years past?
In this article we'll look at both sides of the issue. In the end, I'll come down in the camp of those who say government interest payments are in fact an additional burden.
Terminology: "Costs" versus Harms
Before diving in, we should make a pedantic point about terminology. In these types of analyses, it is common to talk about "the cost" of, say, the Iraq invasion, and to add up things such as the estimated monetary value of soldiers' lives (if they are killed), the estimated value of the physical resources diverted to the war effort, etc.
Strictly speaking, this is an abuse of language. As the famous LSE essays stress, in economics a cost can only be borne by the decision maker. The cost of an action is the subjective value of the next-best alternative that can no longer be achieved. Furthermore, costs can never be "realized," because the decision maker must compare the actual outcome with what might have been.
For example, if we ask a new bride, "What was the cost of your decision to marry John?" it would be wrong for her to say, "Oh, let's see, there was the $10,000 that my dad spent on the reception, then the money we spent on the limo. …" On the contrary, the bride should compare her current evaluation of a life spent with John to what she thinks she would have experienced had she chosen a different suitor or remained single. If the bride is a member of the royal family and had been pressured to marry (Sir) John for political and business reasons — even though she is in love with a commoner named David — then she might reckon the true cost of her decision to be quite high.
In fact, most of the things that we normally describe as "costs" are really harms or damages. If I lose control of my car and hit you, putting you in the hospital for three months, in terms of strict economic theory I haven't "imposed a cost on you," because that doesn't even make sense. Rather, I imposed harms or damages on you.
Notwithstanding this pedantic quibbling, let's proceed to a separate issue: whether a reckoning of the monetary expenditures on a war should include government interest payments.
The Case for Not Including Interest Payments
Preble gives good ways of grasping the huge cost of America's foreign policy. In 2007, for example, the Pentagon's budget, plus its special funding for the wars in Iraq and Afghanistan, totaled $622 billion, or $2,065 for every American resident. … Preble goes further, noting that the number for the U.S. should include the part of the Department of Energy's budget for nuclear weapons ($17.1 billion), the Department of Homeland Security, the Department of Veterans' Affairs, and the Treasury Department's expenditures on military retirement costs. …
Preble goes too far. … He estimates the part of interest on the national debt that is due to military spending and adds that in to the cost of foreign policy. But that's double counting. The cost has already been taken account of in the original expenditure. The fact that it was paid partly by debt rather than solely by taxes is irrelevant.
Henderson's point is pretty straightforward. Suppose that in the year 2000 the government spent $100 billion on a bunch of tanks, airplanes, and warships. What was the total monetary expense involved?
First, assume that the government ran a balanced budget in the year 2000, meaning that all expenditures were financed out of current taxation. In that case, most people would say "the cost" of the military buildup was $100 billion (in year-2000 dollars).
But instead, what if the government ran a $100 billion budget deficit? Does this change our answer?
Henderson (and many other economists) would argue no. The opportunity cost of the decision to build those military items is still the same; the same amount of real resources (steel, labor hours, rubber, etc.) were used up to build the new tanks, airplanes, and warships, regardless of whether the government had a balanced budget or a deficit. So from that perspective, Henderson would argue, clearly "the cost" of the government's decision is still $100 billion in forfeited goods and services that private consumers could have otherwise enjoyed in the year 2000.
But what about the higher interest payments that the government will be forced to make in the years 2001, 2002, etc.? How can Henderson justify ignoring those "costs" of the military purchases?
This is what Henderson meant about "double counting." Suppose that in the year 2000 the government had issued bonds (to cover the $100 billion deficit) with a 5 percent interest rate. Then in the year 2001, taxpayers would be on the hook for $5 billion in interest payments. Assuming the government kept rolling over the debt, the same would be true in 2002, 2003, and so on.
Someone like Henderson would say that if we want to include these annual $5 billion payments as part of the total expense, then we are no longer talking about resource usage. Instead, we are talking about a financial transfer from taxpayers to bondholders. In other words, when Uncle Sam gives $5 billion to bondholders, that action per se doesn't reduce the amount of goods and services that the private sector can produce. (In contrast, when the government spends $100 billion to private contractors to produce tanks, that does reduce the amount of goods and services that the private sector can produce.)
Henderson just wants us to be consistent with our accounting. If we want to count the annual $5 billion transfers — from taxpayers to bondholders — as a "cost" of the military buildup, then we can't also count the original $100 billion expenditure in the year 2000. That's because the taxpayers didn't have to shell out $100 billion at that time; the private bondholders lent the money.
We can now understand Henderson's position, which (to repeat) many economists would endorse. The standard way to measure the size of a government expenditure is to look at the actual price tag. Whether the expenditure is financed through taxation or borrowed funds, the impact is the same (according to this line of thinking). If we focus on the opportunity costs in terms of real resources, the method of financing is obviously irrelevant. And if we insist on including interest payments as true "costs," that means we are focused not on society as a whole, but on taxpayers; and then by consistency we can't include the original expenditure as a "cost" of the program, since the taxpayers didn't foot the bill at the moment of purchase.
The Time Element
In this standard approach, it is still important to keep our eye on the element of time. For example, if the government keeps rolling over the debt from our hypothetical $100 billion purchase made in the year 2000, some people might think that the "total cost" is therefore larger. In other words, if the government makes annual payments of $5 billion on the new debt, from the year 2001 through 2101, then surely that will mean a higher total expense compared to the situation where the government pays off the debt, say, in the year 2005?
No, not in the standard finance framework. When looking at cash flows over time, we have to reduce them to a common denominator. A dollar spent in the year 2005 is worth more than a dollar spent in the year 2085. Once we use present values to convert different cash flow streams into the equivalent dollar sum expressed in today's dollars, we get Henderson's answer: the total expense of a government program is the same, whether it is financed through taxation or borrowing.
One last warning on this subject of time: Some readers might think that it's crazy for the standard economist to argue that a long-term debacle shouldn't somehow raise the "total cost" of a program. In our example, how can it possibly be correct to say that the taxpayers in the year 2040, who must pay $5 billion in interest because of the military expenditure made 40 years earlier, aren't in reality "paying for" that reckless decision?
To be clear: They are paying for it. Specifically, they are paying $5 billion in "2040 dollars," and they will have to pay another $5 billion in "2041 dollars," etc. Yet if the original program had been paid for through taxes, then instead of a constant stream of annual $5 billion payments, the taxpayers would have paid a lump sum $100 billion in "2000 dollars." That is a crucial point. Since "2000 dollars" are worth more than later dollars, the two cash flow streams have the same market value. That's the sense in which Henderson would argue that the financing decision is irrelevant; either way, the program to buy tanks, airplanes, and warships carried a price tag of $100 billion in year-2000 dollars.
Problems With the Standard Analysis
The biggest flaw with the standard analysis is that government interest payments to bondholders do "hurt the economy," because they are financed through coercive taxation. (The government can also pay interest on its debt through further borrowing, which just postpones the problem, or through inflation, which is a form of indirect taxation.) Once we take this aspect into account, it's no longer true that interest expenses are a form of "double counting."
Let's return to our scenario. In the year 2000, the government borrows $100 billion from investors in order to bid away steel, rubber, labor hours, and other real resources from citizens. By channeling these resources into building tanks, planes, and warships, the government's action has necessarily reduced the potential output of automobiles, civilian airliners, and other goods and services in the private sector.
Yet this isn't the end of the disruption. In the year 2001, the government must raise $5 billion in taxes, in order to pay its annual interest expense on the outstanding $100 billion debt. If the government were a private corporation, engaged in the peaceful activity of buying inputs and selling outputs to willing customers, then bond payments would be a wash, as far as society were concerned. In order to raise the $5 billion needed for interest payments, the private corporation would have to produce "$5 billion worth" of goods and services to customers who voluntarily spent the money. The corporation's actions wouldn't diminish the potential output of goods and services in the community.
Yet when the government needs to come up with $5 billion, it simply takes it from the taxpayers, and this does shrink the total potential output of the private sector. For example, if the government levies a percentage tax on income, then people will not work as much as they otherwise would have.
Another way of expressing this fact is to say that in order to raise $5 billion in additional income-tax receipts, the government reduces the after-tax income of taxpayers by more than $5 billion. Thus it is a negative-sum game when the government takes $5 billion from taxpayers and gives it to people holding Treasury bonds. (Note that we're not even considering the opportunity cost of government workers who have to oversee the debt financing.)
Because of what mainstream economists call the "deadweight loss" of taxation, our hypothetical annual streams of $5 billion interest payments really do make society as a whole poorer, in terms of forfeited goods and services that could have been produced in the private sector.
The Difficulty of Numerical Price Tags
Before closing, we should point out that it's still not necessarily right to count the full $100 billion in the year 2000, and then to count the full $5 billion payments every year thereafter, as the "total cost" of our hypothetical government program. Since we're focusing on the actual, "real" opportunity cost of the government program, the $100 billion in the first year is actually too high.
Consider this: We know the plan doesn't cost more than $100 billion, because if it did, entrepreneurs would bid away the resources in question, in order to produce goods and services for which consumers would be willing to spend more than $100 billion. (In other words, why sell tanks, airplanes, and warships to the government for $100 billion when you could use the same amount of steel, rubber, labor hours, etc., to sell cars, commercial airliners, and so on to consumers for, say, $110 billion?)
On the other hand, the true opportunity cost of the plan might be less than $100 billion, if our yardstick is the market value of the forfeited private-sector goods. This is because military contractors notoriously overprice their deliverables to the government. In other words, a military contractor would be able to earn a far higher "markup" on his purchase of inputs, when selling to the government, compared to a private-sector operation buying inputs and selling goods to consumers in the market.
Because of this, the steel, rubber, and labor-hours used up to produce "$100 billion worth" of stuff for the Pentagon, probably could not be used in the private sector to produce "$100 billion worth" of stuff for consumers. What's really happening is that part of the $100 billion military expenditure is a hidden transfer payment to the politically connected contractors.
So how much should we say the $100 billion expenditure in the year 2000 "really" costs? It's impossible to say. It's true that a chunk of it is a disguised subsidy to military contractors, and in that respect doesn't diminish society's capacity to produce (private) goods and services. On the other hand, in order to get the handout, the military contractors will spend lavish amounts lobbying Congress, wining and dining procurement officials, and so forth. In the end, resources are used far less efficiently than they otherwise would have been.
All we can say for sure is that $100 billion is the upper limit on the "social opportunity cost" in forfeited goods and services, because of the initial expenditure. The true number is presumably lower than that, but our estimate would vary, depending on how many different factors we want to include in the analysis.
And what of the annual $5 billion interest payments? How do we gauge their impact on the ability of the private sector to produce goods and services each year?
Again, it's impossible to say. We have already seen that it's not the sending of a check per se that is the problem, but rather the disincentive to produce caused by the government's taxation. Some taxes are less destructive than others in this respect. For example, a highly progressive income tax will discourage work effort far more than a simple "poll tax" levied on each taxpayer, regardless of income.
In practice, if I had to guess, I would say that raising $5 billion from taxpayers (in order to pay interest to the bondholders) reduced total private-sector output by more than $0 but less than $5 billion. So in our exercise we wouldn't add in the full $5 billion to "the total cost" of the original military purchases, but we wouldn't completely disregard it either.
As we have seen, it becomes incredibly complicated when we try to estimate the total "social cost" of a government policy. Ultimately, this difficulty stems from the fact that costs really only make sense in terms of an individual's subjective preferences. In that respect, costs cannot be aggregated.
Furthermore, even if we switch our framework so that "total cost" means the reduction in potential goods and services that might otherwise have been produced, we still run into thorny problems of knowing which factors to include in the analysis and which to leave out. In practice, it is impossible to actually measure all of the different factors that contribute to this question.
Ultimately, I would argue that Preble and Henderson both oversimplify: it's not really accurate to simply add up the original government expenditure plus subsequent interest expenses (as Preble did), but it's also not quite right to say that financing decisions are irrelevant (as Henderson did). If we throw caution to the wind and insist on calculating a single number, the best answer would probably fall in between the range that Henderson and Preble would give.
 I want to thank David for reviewing this portion of the article and suggesting a clarification of his position.
 In fairness, I should point out that Henderson could respond that the deadweight loss of coercive taxation also applies in the case where the government taxes citizens upfront for the entire expenditure. So it's not clear whether my observation in the text above tips the scales in favor of nondeficit financing. Even so, my main point is that the standard analysis of government finance often neglects this important factor.