Brad DeLong is annoyed with people who are scared of invisible bond-market vigilantes. And he’s right to be annoyed! It’s extraordinarily silly — or dishonest — to claim that the confidence of bondholders constrains fiscal policy in the United States. As he puts it, “Any loss of confidence in the long-term fiscal stability of the United States of America” is an “economic thing that does not exist.”

So he’s right. But does he have the right to be right?

I’m going to say No. Because the error he is pointing to, is one that the economics he teaches gives no help in avoiding.

The graduate macroeconomics course at Berkeley uses David Romer’s *Advanced Macroeconomics, 3rd Edition*. (The same text I used at UMass.) Here’s what it says about government budget constraints:

What this means is that the present value of government spending across all future time must be less than or equal to the present value of taxation across all future time, minus the current value of government debt. This is pretty much the starting point for all mainstream discussions of government budgets. In Blanchard and Fischer, another widely-used graduate macro textbook, the entire discussion of government budgets is just the working-out of that same equation. (Except they make it an equality rather than an inequality.) If you’ve studied economics at a graduate level, this is what government budget constraint means to you.

But here’s the thing: That kind of constraint has **nothing to do** with the kind of constraint DeLong’s post is talking about.

The textbook constraint is based on the idea that government is setting tax and spending levels for all periods once and for all. There’s no difference between past and future — the equation is unchanged if you reverse the sign of the t terms (i.e. flip the past and future) and simultaneously reverse the sign of the interest rate. (In the special case where the interest rate is zero, you can put the periods in any order you like.) This approach isn’t specific to government budget constraints, it’s the way everything is approached in contemporary macroeconomics. The starting point of the Blanchard and Fischer book, like many macro textbooks, is the Ramsey model of a household (central planner) allocating known production and consumption possibilities across an infinite time horizon. (The Romer book starts with the Solow growth model and derives it from the Ramsey model in chapter two.) Economic growth simply means that the parameters are such that the household, or planner, chooses a path of output with higher values in later periods than in earlier ones. Financial markets and aggregate demand aren’t completely ignored, of course, but they’re treated as details to be saved for the final chapters, not part of the main structure.

You may think that’s a silly way to think about the economy (I may agree), but one important feature of these models is that the interest rate is not the cost of credit or finance; rather, it’s the fixed marginal rate of substitution of spending or taxing between different periods. By contrast, that interest is the cost of money, not the cost of substitution between the future and the present, was maybe the most important single point in Keynes’ *General Theory*. But it’s completely missing from contemporary textbooks, even though it’s only under this sense of interest that there’s even the possibility of bond market vigilantism. When we are talking about the state of confidence in the bond market, we are talking about a finance constraint — the cost of money — not a budget constraint. But the whole logic of contemporary macroeconomics (intertemporal allocation of real goods as the fundamental structure, with finance coming in only as an afterthought) excludes the possibility of government financing constraints. At no point in either Romer or Blanchard and Fischer are they ever discussed.

You can’t expect people to have a clear sense of when government financing constraints do and don’t bind, if you teach them a theory in which they don’t exist.

EDIT: Let me spell the argument out a little more. In conventional economics, time is just another dimension on which goods vary. Jam today, jam tomorrow, jam next week are treated just like strawberry jam, elderberry jam, ginger-zucchini jam, etc. Either way, you’re choosing the highest-utility basket that lies within your budget constraint. An alternative point of view – Post Keynesian if you like – is that we can’t make choices today about future periods. (Fundamental uncertainty is one way of motivating this, but not the only way.) The tradeoff facing us is not between jam today and jam tomorrow, but between jam today and money today. Money today presumably translates into jam tomorrow, but not on sufficiently definite terms that we can put it into the equations. (It’s in this sense that a monetary theory and a theory of intertemporal optimization are strict alternatives.) Once you take this point of view, it’s perfectly logical to think of the government budget constraint as a financing constraint, i.e. as the terms on which expenditure today trades off with net financial claims today. Which is to say, you’re now in the discursive universe where things like bond markets exist. Again, yes, modern macro textbooks do eventually introduce bond markets — but only after hundreds of pages of intertemporal optimization. If I wrote the textbooks, the first model wouldn’t be of goods today vs. goods tomorrow, but goods today vs. money today. DeLong presumably disagrees. But in that world, macroeconomic policy discussions might annoy him less.

Hello,

I am doing some research on the governmental budget constraint that you list above. I am an ex-mathematician that ended up in finance. I have doubts about the internal consistency of that relationship; there are embedded assumptions that appear incorrect. In most references I have, the relationship is assumed without explanation. Does the Blanchard and Fisher text cover the assumptions in detail? (My prefernce is for a discrete time version, since doing integrals properly is a pain.) Thanks.

It's derived from the household budget constraint. I think the assumptions are stated clearly.

There is the no-ponzi game condition, which require that debt goes to zero as time goes to infinity. I think that is a bad assumption in terms of describing the real world, but it is explicitly stated and I don't see that it creates any logical inconsistencies.

Can you say more about what you have in mind?

The budget constraint in integral form is based upon an assumption that the term premium is zero in the yield curve. (I looked at discrete time versions.) Any form of term premium means that the error in the relationship can be arbitrarily large. Although the mathematics are perhaps technically correct, missing that assumption is a big problem, considering that it contradicts pretty much everything we know about fixed income markets. And there is a certain amount of irony in that the people who worry about Ricardian Equivalence also worry about bond risk premiums rising. I ran through my argument on a post on my blog "What is Ricardian Equivalence and why it does not hold" (bondeconomics.com); a link to it is under the "DSGE" navigation tab at the top of the page.

As for the assumption that the NPV of debt goes to zero at infinity, I think the assumption yields fairly pathological results. I want to get my hands on a better explanation of the theory, so that I can write a more useful critique. Most of my sources only cover it in one or two lines, and so I do not know exactly what the detailed logic is supposed to be behind the assumption.

I'll order the text, and try working with it.

Thanks.

Blanchard and Fischer is a big textbook and the government budget constraint only takes up about 10 pages, so unless you need a good reference on modern macroeconomic theory in general I'm not sure it's worth your buying it.

You might want to take a look at the discussion in Michael Woodford's Interest and Prices. It's the canonical New Keynesian text and there's a copy online at libgen.net, I believe. Woodford assumes "that all government debt consists entirely of riskless one-period nominal bonds." So there won't be any term premium in his model. In general, I think the level of abstraction in these models is such that a lot of what we know about real world markets is not going to apply to them.