Saturday, September 29, 2012

Phantoms in the Dark Part 6: The Barro-Wight

I need to get to the crux of this, which is not really the issue of whether current deficits hurt or help growth, or whether the burden of current deficits are borne by the present population or by some distant future population.  No, the crux is really the question of how much total debt---not deficit---is too much, and whether there’s a point at which the total debt starts to slow growth down, or even reverse it, or otherwise cause damage to the economy.  That’s the real phantom here.  That’s the big one, the one that’s scaring everyone.  That’s the issue that makes people all over the political blogs claim that the country is, or will be, “bankrupt”, and that caused even Barack Obama to say at one point that we’ve run out of money.  So I need to discuss that, and I’ll do that soon.

But I’m aware, because I was reminded, that I never talked about the Barro/Ricardian equivalence idea about deficits, item number 3 in Nick Rowe’s list, which still talks only about deficits rather than total debt.  Barro’s idea is that whenever there is a federal budget deficit ordinary households will save more and spend less in order to prepare for what they think are the inevitable future taxes necessary to pay for the deficit.  As a result the total savings and total demand for goods and services is, according to this theory, the same if the government borrows to cover its costs as it would be if the government simply taxed the population and balanced the budget in the first place.  Deficits don’t burden future populations, this theory says, but only because the current populations rationally expect future taxation to pay for the deficits and all accumulated interest until the tax is applied.  Households rationally, and carefully, save just enough to pay the equivalent taxes, and pass their savings down the years until the taxes are imposed.

I am a little hesitant to approach this topic because it’s hard to know where to start: Robert Barro is a staggeringly smart economist, but I’m just puzzled by the whole set of assumptions that have to be made to make that concept work.  They seem to me to be the kind of assumptions you could make only if you have been absorbed in abstract mathematical models so deeply that you can no longer see the real world, absorbed so deeply that you think those models are the real world.  I can understand that.  I’ve been there, in my long-ago days in graduate school.  The mathematics were so beautiful that I wanted to live in the world they described instead of the world that exists.  But when we’re talking about actual policy prescriptions, we can’t pretend that the math is everything: we have to implement policy in the real world, and the results we get will be determined by how the real world works, not by how some seductive abstract model with perfect mathematical curves works.

But in any case, this post will be longer and wonkier than most of my posts, and longer and wonkier than I would like it to be.  I don’t know how else to express my problems with this item from the list. 

My examples will come from this version of Barro’s explanation of Ricardian equivalence.  To show what I mean about the assumptions that support the model, let’s start with this quote from that paper:

“I will sketch the standard model. The starting point is the assumption that the substitution of a budget deficit for current taxation leads to an expansion of aggregate consumer demand. In other words, desired private saving rises by less than the tax cut, so that desired national saving declines. It follows for a closed economy that the expected real interest rate would have to rise to restore equality between desired national saving and investment demand. The higher real interest rate crowds out investment, which shows up in the long run as a smaller stock of productive capital. Therefore…the public debt is an intergenerational burden in that it leads to a smaller stock of capital for future generations.”

It’s true that this model is fairly widely accepted in some parts of the economic world.  In fact, some parts of that world see this model, that government deficits reduce the total savings in the economy and that they therefore “crowd out” private investment, to be so self-evident that it requires no proof; there are those who are astonished and exasperated when anyone questions it. 

And I, on the other hand, am astonished and exasperated that anyone fails to question it, since to me it seems to be self evidently wrong. 

So let’s look at it.  Here are the assumptions I see in the quote above:

He assumes that deficit spending absorbs savings in the economy---that it competes with business for a limited supply of loanable funds.  It’s not a fixed supply, since he claims that a tax cut (or an increase in income caused by increased government spending) will increased desired savings, but, he asserts, not by as much as the deficit the government must cover. 

But as I pointed out a few posts ago, when the government spends, that act creates savings in the economy; it increases total savings by at least the amount of the government spending.  Someone in the economy, a business or a household, receives what the government spends.  I’m not proposing anything radical by saying this, these are just two sides of the same transaction.  Until that person or business spends the money, it is an increase in savings.  But what happens when a household does spend the money on consumption goods, for example?  Doesn’t that decrease savings?  No, it doesn’t.  The money is subtracted from the household’s bank account by the transaction---and is added to the bank account of the business that sold the consumption goods.  The savings is still there, but now it is corporate savings instead of household savings.  You can go through all the other possible kinds of transaction, such as investment or other spending by businesses, and you will find that because every transaction has two sides, one person buying and another selling, the money, the savings created by government spending, never leaves the system.  It always shows up as savings in someone’s account, unless the government taxes it away.   So when the government spends money, and then borrows to cover the cost, it is simply borrowing the savings that were created by its spending, not absorbing any share of a pool of savings that existed prior to those actions.  (That is not, of course, the end of the story: the actual savings that have shown up in people’s and business’s bank accounts may be more savings than they desire at their new level of income, and they will try to reduce it by spending or investing.  They will continue to do this until the total nominal national income rises enough to make the actual savings acceptable to the people and businesses who end up holding it---but that’s a much longer story than I can tell in this post.)

But even if it were not true as an accounting artifact that government spending creates new nominal savings, the quote from Barro above would be questionable, since it seems to assume that all savings that existed prior to the government action were already demanded by businesses or households for investment purposes. That is not necessarily true; in fact it is clearly not true right now.   There were $1.452 trillion dollars in excess reserves at the Federal Reserve bank as of September 19, 2012.

Now let’s get to the Ricardian model itself.  Barro says:

"The Ricardian modification to the standard analysis begins with the observation that, for a given path of government spending, a deficit-financed cut in current taxes leads to higher future taxes that have the same present value as the initial cut."

Alternatively, he would say that an increase in government spending without increasing taxes in the present must imply an increase in taxes in the future that has the same present value as the spending increase.  And he goes on to explicitly state:

"This result follows from the government's budget constraint, which equates the total expenditures for each period (including interest payments) to revenues from taxation or other sources and the net issue of interest-bearing public debt."

But if you have read the prior Phantom posts, you know that there are economists out there, from at least the time of Abba Lerner’s paper on Functional Finance, who do not accept the idea that the government even has a budget constraint.  Nor do I.   The usual expressions of an intertemporal government budget constraint that include only taxing, borrowing and spending as variables across time.  But if we include the government’s ability to create money into an expression of an intertemporal budget condition, or intertemporal budget equation, we see that it is no longer an intertemporal constraint: a  government that controls its own money supply can always purchase anything that is offered for sale in its own currency, either now or at any time in the future.  For other economic reasons, for example to avoid excessive demand that would drive inflation, the government may choose not to finance its spending by money creation, but its equally true that to stimulate demand and achieve full employment the government may prefer that means of finance in some periods in the future.  All of that simply means that the full intertemporal budget equation, including all methods of financing government activity, might be useful along with other equations in planning.  But it is not a constraint on government financing, and it can't be used in any simplistic way to forecast tax rates in the future.

So this whole Ricardian alternative depends on assumptions of eternal full employment of resources, eternal full employment of any existing savings, and the existence of an intertemporal government budget constraint, none of which are true in the real world.

But none of these is the biggest failing, in my view, of the Ricardian equivalence theory. The theory itself after all of this build up states that:

"household's demands for goods depend on the expected present value of taxes---that is, each household subtracts its share from the expected present value of income to determine a net wealth position."

This is the one that floors me, frankly.  I don't understand how anyone can suggest this with a straight face. 

Let me make this clear.  Future tax rates do not depend on current deficits even in theory, and if there is any such dependence as a practical matter it is a pretty feeble one.  As evidence, we could simply point to the post-WWII history of the United States, where deficits have been common but tax rates have consistently declined.  In fact when deficits exploded under Reagan/Bush in the eighties and under George W. Bush  in the oughts, tax rates declined a great deal in both decades.  The top tax rate at the end of WWII was above 90%.  In spite of all the decades of deficits since then the top tax rate is now 35%.

But ignoring all of that, ignoring all the questionable or simply false assumptions discussed above, and the feeble---in fact completely invisible---connection between observed tax rates and actual deficits, does anyone seriously believe that households---all of them, rich and poor, educated and not---take the time to calculate a discounted present value of all expected future taxes before they buy milk at Safeway?   People simply do not behave like this.  I don't mean that a few irresponsible people who fail to calculate their net wealth position including the discounted present value of all future taxes while they’re standing in the produce aisle weaken the theory.  I mean that because of this problem alone the theory is bunk because absolutely no one behaves like this.  If there are exceptions, I don’t know them.  I strongly doubt that Robert Barro behaves like this.  

Here's why I doubt it: the failure to do this calculation isn’t laziness.  And it isn't ignorance.  People do not try to make that calculation because neither an average person nor the world’s greatest economic expert can make it, so no rational person would try to make it as a basis for their daily decisions. While the CBO and other forecasters do try to peer into distant futures with huge macroeconomic models, they do so with a humility born of frequent error.  The CBO should not be blamed for these errors.  The distant future is impenetrable and unpredictable, even to the brightest and best-educated people on this planet.  As I pointed out in an earlier post, no one a quarter of a century ago could possibly have predicted the federal budget surpluses of the late nineties; they could not have predicted the dot-com boom or the subsequent dot-com collapse, because a quarter of a century ago even the concept of a dot-com was unknown and unimagined, except perhaps in the secret inner mind of Timothy Berners-Lee.  And if the best educated people on earth were trying to predict the stream of government revenue in the future they would include tax rates as only one of many variables: economic growth, population growth, income distribution, employment, the probability of new resources, new technologies, wars, booms, busts, and many other things would also enter their calculations.   And they still would not get it completely right.

So of the four items in Nick Rowe’s list on the burden of deficits, this is the one I find least plausible.  The real world is not nearly as perfect as the original “standard” model that Barro outlined in the quote at the top of this post.  The government does not actually face a budget constraint, so using one to predict tax rates in the future is misguided at best.  Unemployment does occur, not only of people, plant capacity, transportation capacity and other real things, but also of saved funds.  Future tax rates don’t really depend on current deficits, partly because future financing needs and strategies depend on many, many things, and accumulated debt is only one of them.  And finally, no one at all would ever, or could ever, make the calculations that Ricardian equivalence requires of them as they make their savings and consumption choices.


  1. "So this whole Ricardian alternative depends on assumptions of eternal full employment of resources, eternal full employment of any existing savings, and the existence of an intertemporal government budget constraint, none of which are true in the real world."

    It does depend on the intertemporal government budget constraint. But not full employment.


    1. Welcome back, and thanks again for finding my blog among the galaxy of stars that populate the interweb. It’s good to know that someone is out there reading these words I type into my keyboard late in the evening. This time I’m not just back from a dinner out: I’m just back from mowing my lawn. Not quite as much fun, but the upside is that the yard looks great.

      You spend your days teaching this stuff, which is by far the best way to learn anything, while I spend my days managing engineering projects. So I’m willing to accept my lesson and try to absorb it if I’ve made a mistake in this. But I need some convincing that full employment of both resources and finances aren’t presumed somewhere in the heart of the argument for the “Ricardian alternative”.

      Here’s the basic bones of the vision that makes me say that: I see the “crowding-out” concept that Barro used as the standard model in the paper I cited above as one side and the Ricardian alternative as the opposite side in a discussion of where the resources come from to fulfill government demand. Barro and others explain their view by starting with a presumption that the government deficit results from reduced taxation while spending is held constant, but it’s easier for me to explain this if I start with a deficit that is created by keeping taxes constant and increasing government spending, and to start by looking at the impact on the physical side of the economy rather than the financial side.

      The basic idea is that when the government consumes something, whether it’s some resource (steel to build a new ship) or a final product (whiteboards for the meeting rooms) or labor, that thing is then no longer available for use by the private economy. The crowding-out end of that thought concludes that it comes out of investment that would otherwise occur, by the process of absorbing savings that could otherwise be used by private firms for that purpose. The Ricardian alternative concludes that it comes from household consumption, because households will feel less wealthy by the amount of some presumed future taxation, and will cut back on consumption and increase savings. It’s true that anything the government uses must come from somewhere, and it’s tempting, given a set of alternative ends of a spectrum like that, to say “well, there must be some true point in the middle between the two extremes”. But neither of these visions of real displacement in the real economy accepts the possibility that the new government consumption comes from an existing pool of unused resources---piles of steel that would otherwise not be sold, or labor that would otherwise be idle---and that no one has to give up anything in the real economy to enable the new government use of these resources.

      (continued---too dang many characters for a response,apparently)


    2. It’s certainly possible to write down a simple two-period IGBC equation, enter the Riciardian-alternative assumption about how household savings reacts to a government deficit, and show that the equivalence notion holds in that equation, and to do all of that without ever mentioning employment, full or not. But if this bit of “hat-algebra” (borrowing your phrase from your recent NGDP post) shows the same result as the initial crowding-to-Ricardian spectrum in the real economy that I described above, which means that it fails to show the result from a real economy with idle resources, then the presumption of full resource employment has to be hidden in the equations somewhere. Exactly where would depend on how you write down the equations, I guess, but here’s a possibility: the Ricardian view seems to hinge in part on the idea that the only way government revenue can grow in the second period is by raising tax rates, so that an average household would pay higher taxes in the second period even if it had the same income. But of course, if there is unemployment in the first period that is reduced by the second, then government spending will decline (because dealing with unemployment is costly) and government revenues will rise (because of new economic activity from increased employment in addition to normal economic growth), and these two things will happen without any increase in tax rates.

      Barro seemed to recognize the implicit assumption of full employment in both his standard model and the Ricardian alternative, because he cites it as one of the potential theoretical objections that needed to be address. And he did spend much of the paper I quoted in my post responding to several such theoretical objections, including finite horizons for consumers, imperfect loan markets, uncertainty about the timing or amount of future taxes due to allocation of taxes across income classes, and finally the issue of “involuntary unemployment”. The quote marks are his, and might be intended to imply that he does not think such a thing can really happen, or not for long.

      I thought he did a very credible job responding to the other possible theoretical objections; he spent several pages on each, and presented some clear, logical arguments. But on the issue of “full employment” (again, the quote marks are his), he provided two short paragraphs. In those two paragraphs he presents no real response, except to express his doubts about the “formulation of Keynesian models”. His dismissal of Keynesian ideas is pretty brutal, in fact. His second sentence in this section starts like this:

      “In standard Keynesian analysis (which still appears in many textbooks)…”

      to which Keynesians of any variety can only respond: ouch.