After reading Nick Rowe, Brad DeLong, and Paul Krugman, I now understand who bears the burden of the government debt. Any member of the younger generation who reads the stuff that these old guys have written on government debt will be hopelessly confused. And that will be a burden on us all.
The idea that a larger government debt is a burden for future generations is so strongly intuitive as to be part of the standard narrative for anyone who wants to tell you that more government debt is a bad idea. But is that correct? When I teach undergraduates about government finance, I find it instructive to start with the Ricardian equivalence theorem. In a frictionless world, government debt is irrelevant. A tax cut that increases the government debt today has no effect because everyone understands that government debt is just deferred taxation. Lifetime wealth does not change, and everyone saves their tax cut today so as to pay the higher future taxes that are required to pay off the government debt in the future. Ricardian equivalence is a useful starting point, as it makes clear what frictions might cause Ricardian equivalence to break down - and that's a route for thinking about how policy might work to improve matters. Distorting taxes, intragenerational distribution effects from tax policy, and credit market frictions all potentially make a difference. But that doesn't make Ricardian equivalence "wrong" or useless. Indeed, it is an important organizing principle, and needs to be taken seriously.
One key departure from Ricardian equivalence arises because of intergenerational redistribution effects from government tax policy. Ricardian equivalence works because changes in the timing of taxes do not matter for anyone's wealth - the present value of tax liabilities for each individual is unchanged. But what if the government is cutting one person's taxes today, and paying off the government debt in the future by taxing someone else? Surely that comes into play in reality, as the government can cut taxes today, increase the government debt, and potentially not pay off the debt for 100 years, at which time the people who were on the receiving end of the current tax cut are long dead?
Robert Barro had an answer for this. If generations are tied together through altruism (we care about our children) and bequests, then we behave as if we are chained together with our descendants, and might as well be infinite-lived households. If I receive a tax cut today, then I save more, and give my descendants a larger bequest that will allow them to pay their higher future taxes. But surely this is going to far. Following the logic of Barro's argument, chains of altruism tie everyone together in ways that make everything neutral. There is certainly altruism in the world, but I think we all recognize that some collective action is required to make outcomes more efficient.
A standard vehicle for thinking about intergenerational distribution is the overlapping generations (OG) model (DeLong knows that the OG model exists, which is a start, at least). Peter Diamond's version of the OG model, with capital accumulation and production, is useful, but I'll simplify here to get the ideas across. Suppose two-period-lived people, with two generations alive at each date - young and old. The population grows at rate n, and each person is endowed with y units of consumption good when young, and zero when old. Storage is possible, with r the rate of return on storage from one period to the next. This is useful, as it ties down the real interest rate at r (so long as there is some storage in equilibrium). Suppose also that the government has access to lump sum taxes on the young and the old. The government can also issue one-period real bonds, with the real interest rate on government bonds = r in equilibrium.
Suppose that the government increases the quantity of government debt in the current period, T, by an amount b per young person currently alive. Since the real interest rate is constant at r forever, the effect of this change in government debt on economic welfare for each generation depends only on how taxes change for the rest of time. Suppose that the increase in debt is reflected in a lump sum transfer of x(T) to each person currently alive, so
x(T) = [b(1+n)]/(2+n)
Now, clearly the current old are better off as a result. They receive a transfer and are better off. What about everyone else - the current young, and future generations? For those people, it depends.
Scenario 1: Pay off the debt at T+1: In this scenario, to pay off the principal and interest on the government debt in period T+1 requires total taxes per young person alive equal to [b(1+r)]/(1+n). Supposing the government spreads the tax burden equally across people alive in period T+1, the tax per person at T+1 is
-x(T+1) = [b(1+r)]/(2+n)
Therefore, as long as n > 0, so the population is growing, the young people in period T (who are old at T+1) are better off, as their lifetime wealth increases. The old in period T+1 are worse off, as they pay a higher tax.
In this case, it is clear that the government debt is a burden on future generations - specifically the next one, which pays the taxes to retire the government debt.
Scenario 2: Hold the debt constant forever at its higher level: Under this scenario, the government debt per young person in period T+i is
(where x^i is x to the power i). If taxes are the same for young and old during any period, this implies that the tax per person in period T+i is
-x(T+i) = rb/[(2+n)(1+n)^(i-2)]
In this case, the young in period T are better off in present value terms, but anyone born in periods T+1 and beyond is worse off.
Compared to scenario 1, we are now spreading the burden of the government debt across all future generations. Note however, that the government debt per capita vanishes in the limit. The experiment increases the government debt by a fixed real amount and, as the population grows, the burden of the debt in per capita terms falls.
Scenario 3: Hold debt per capita constant at the higher level forever: In this case, if everyone at a point in time bears the same tax, the tax in period T+i is
-x(T+i) = [b(r-n)]/(2+n)
As in scenarios 1 and 2, the young in period T are better off, but now things are more interesting. If r > n, so that the real interest rate is higher than the population growth rate, we get a similar scenario to what we had previously, except we are now spreading the burden of the debt equally across future generations. However, if r < n, we actually make everyone better off. When r < n, it is socially inefficient to use the storage technology, and an optimal arrangement would have government debt driving out use of the storage technology. Diamond's classic paper shows how this works in a more standard neoclassical growth framework. In general, an economy with r < n is dynamically inefficient, and government debt is one means (as is social security) for effecting the appropriate intergenerational transfers. The interesting thing about the scheme in scenario 3 when r < n is that it works as a Ponzi scheme - effectively, each generation borrows from the next, and everyone is better off. Magic!
What's the bottom line? Increasing the quantity of government debt does indeed represent a transfer in wealth from future generations to those currently alive (or, in reality, those currently working). The only important qualification arises if society is not efficiently distributing wealth across generations. In that case, the inefficiency can be corrected with an increase in government debt. But unless we want to argue that the intergenerational redistribution being done by the U.S. social security system is insufficient, that seems a difficult argument to make.