Arnold Kling  

The Fiscal Singularity is Near

PRINT
The SEC-Goldman Flap... Goaded...

In my recent paper, I started with budget projections that had debt held by the public reaching 78.8 percent of GDP by 2020. In fact, the most recent CBO projection of the Administration's Budget offers a forecast of public debt held by the public reaching 90.0 percent of GDP by that date.

I don't think we will get much past 2020 along our current policy path. I need to work out scenarios more carefully, but the interest cost of the debt as a percent of GDP is the debt times the nominal interest rate. If the nominal interest rate were 10 percent in 2020, then the interest alone would be 9 percent of GDP. Since non-interest spending also would exceed projected revenues, we would be adding nearly 10 percentage points to the debt-to-GDP ratio in just 2020, with ever-increasing amounts in subsequent years. Nobody is going to want to buy Treasury securities given that outlook.

Yes, one option is to just print money to pay off the debt. The money that we print is called high-powered money, also known as the monetary base. Is it fair to assume that the velocity of high-powered money is about 10? (I think that the ratio of nominal GDP to the stock of high-powered money is about 10 these days.) If so, then imagine that the interest on our debt is 10 percent of GDP and we want to print enough high-powered money just to pay the interest on the debt. Then we have to inject an amount of high-powered money equal to 10 percent of our GDP, which at a velocity of 10 would mean a 100 percent increase in nominal GDP, which would be nearly all inflation. That is hyper-inflation territory, which would be roughly as catastrophic for savers as defaulting on the debt.

This is crude arithmetic. I am not a fancy theorist like Paul Krugman or Brad DeLong or Mark Thoma, who are all convinced that the deficit should be higher rather than lower. But I wish somebody could explain to me why the crude arithmetic is not scary.

[UPDATE: a commenter points out that the three economists I cited in the preceding paragraph, like other economists, say that we need to reduce the long-term deficit. But my point here is that "the future is now." It seems to me that we are already at the point where we do not have the luxury of separating the long-term issue from short-term policy.]


Comments and Sharing


CATEGORIES: Fiscal Policy



COMMENTS (26 to date)
Amaturus writes:

I'm not sure I understand how applying a nominal interest rate to the entire debt makes any sense. The national debt is an amalgam of treasury securities released at different maturities and different times. In 2020, there would still be a significant number of 30 year treasury bonds on the books released during the current time when their yield is less than half of that 10% nominal rate from 2020. The coupon rate on the bonds will remain the same low rate into the future and the face value due at maturity is nominally the same. Shouldn't that reduce the cost of maintaining the debt from your example?
The situation is still unsustainable fiscally, but TreasuryDirect.com reports that interest expense on the debt has fallen from 2008 to now. I don't think the example captures reality well.

Boonton writes:

Isn't Japan running at something close to 170% debt to GDP?

Nobody Special writes:

I can just feel the following argument coming: the printing press debt repayment plan is our "only option", and the resulting inflation will actually solve the problem of underwater mortgages.

OneEyedMan writes:

I'm not sure we need to take the velocity of high powered money that seriously. If we want to monetize all or part of the debt we can print money as long as we increase the reserve requirements on bank loans. That swaps out bank created money for government created money but otherwise should leave the balance sheet of the government unchanged.

And if we manage to get 100% inflation one year then the real value of the debt is half as large the next year. Isn't the real measure of the cost of debt the fraction of real output that must devoted to its service? If so, then debt the next year is twice as affordable as the year of heavy inflation.

The unstated wrinkle here is that the deficits continue to grow. You can inflate your way out of a debt only if you aren't continuing to need to borrow. And that of course is likely to be exactly what we Americans face.

Ted writes:

You either misunderstand was Krugman, Delong, and Thoma believe or you are misrepresenting them, particularly Delong. It's fairly clear to me that all three of them are long-term deficit hawks, which is what really matters. Obviously short-term deficits feed into long-term deficits but it's a complete misrepresentation to suggest that they believe "deficits" should be high and that high debt is always and everywhere a good thing. We can debate whether short-term deficits are nessecary to stimulate growth in the short-term, but it's not as though any of them believe that deficits and debt are just dandy. In fact, I'd argue that their desire for short-term deficits feeds into their desire for long-term budget responsibility. In their view, in particular Delong, if the government has a surplus they shouldn't go and do a massive tax cut without any corresponding spending cuts and, in fact, rapidely increase spending simultaneously. And if a major shock hits and we fall into a recession, this would give the government a lot more breathing room to engage in large fiscal expansion without running up such huge short-term deficits. I think their position is that long-run budget balance (or at least low deficits you can continously roll over easily) are a good thing to have, especially because when a shock hits and the monetary authority is weak because of the ZIRP then you have some room for fiscal expansion. Again, you can debate whether short-term fiscal deficits are nessecary, but it's a misrepresentation to imply that long-term deficits out to 2020 is something they endorse.

Also, your argument about hyperinflation seems weak. We can run debts that high without hyperinflation or mass panic (history suggets that if the bond market doesn't care that we can -and they seem to have a lot higher tolerance than I'd expect) and tame them if we really wanted too (in fact, we could run substantially higher debts for quite some time - though I wouldn't recommend that). The problem with your argument is that you immediately assume that seigniorage is the only way to tame our fiscal situation, which obviously isn't true. There are plenty of ways. This question isn't arithmetic, this question is political. Do the politicians have the will to do some of the painful things that are nessecary? Do they have the will to do spending cuts and tax hikes to get this under-control? I believe, honestly, that when push comes to shove they do. But I suspect we'll have to come closer to a day where we run into trouble before the politicians will care enough (which will only make it worse since then you'll require stronger measures later ...).

Thucydides writes:

"Long term deficit hawks...." Reminds me of "I'll quit drinking tomorrow..." Were these guys deficit hawks in recent years? OK, maybe there was the occasional complaint when there were Republicans in office.

ThomasL writes:

I just finished your paper.

I think all of your suggestions are good. The most disturbing aspect is how big a problem remains after fixing the fiscal sustainability.

My fear is that when one starts including personal savings and household debt, it maybe that the baby-boomers as a group have just messed up. They spent too much, too early, and there isn't enough money for them to get the lifestyle and care they expect--and that they have seen their parents receive.

They don't have enough money of their own, and uniquely high piles of debt and demographics makes it unlikely that current workers will have enough money both to take care of them and save for their own future.

Some cuts, like reduced use of "premium" medicine will probably do no harm--and may actually help health and lifestyle--but I'm afraid that overall, without some major, unknown cost-reducing technological breakthrough, standards are going to go down, not up, for quite a number of people.

That is scary.

E. Barandiaran writes:

Arnold, your arithmetic is scary but it doesn't mean sudden death. Your fiscal crisis will continue for many years. During this slow death, politicians and their troupe of rogue economists will find many ways to empty most boxes and pockets before declaring "el corralito". I bet you that the troupe leader will be again a Harvard Ph.D.

Nick Rowe writes:

You need to make a minor change to your arithmetic (unless I've misunderstood).

Suppose (say) nominal GDP is growing at 4% per year (2% inflation + 2% real). Then the nominal debt can grow at 4% per year too, with the nominal debt/nominal GDP ratio staying constant. So if nominal interest rates are 10%, you only need to pay 6% interest times the debt, and borrow the remaining 4%, so the debt grows at 4%.

azmyth writes:

I'd like to turn this question on its head. Is there any possible conceivable financial structure where retirees can expect the same living standards with 2 workers per retiree as they do when there are 5 or even 10 workers per retiree? Economists are used to looking at financial flows to gain insight on real goods flows, not the other way around. In this case, examining a possible real goods flow is probably the better way to look at it.

I would say no, barring the possible invention of a cornicopia machine. Assuming 2% growth and returns on capital remaining steady, nothing we can do will allow future generations of retirees to live the same way as past generations of retirees. We could invest in massive amounts of storage technologies to move physical goods forward in time, but the goods that the elderly most demand - medicine and services - need to provided at the same time they are consumed. Rather than financial apocolypse, I predict a simple refutation of Medicare and Social Security promises.

Nick Rowe had a good related article a while back.

saxdrop writes:

Net Interest Rivals General Spending by 2040

http://mercatus.org/publication/net-interest-rivals-general-spending-2040

Boonton writes:

Arnold:

UPDATE: a commenter points out that the three economists I cited in the preceding paragraph, like other economists, say that we need to reduce the long-term deficit. But my point here is that "the future is now."...

Excuse me but isn't this just the opposite of what you're saying? You are saying that if various entitlement programs continue on their current trend lines debt and interest will be exploding in a few decades. Even if we were to achieve a balanced budget today or achieve a surplus today (as we did during the last Democratic President), it wouldn't matter. It wouldn't even approach being able to pay for the deficits that are coming in 20, 30, 40 years and beyond.

If that's the case quibbling over deficits today, like the recent ruckus over extending unemployment benefits is absurd, even more absurd to call such people deficit hawks. That's like having a car stuck on the railroad tracks with an express train due by at 5:10 PM and a 'safety hawk' yattering in your ear at 4:50 PM that you should roll up the windows.

Now I still think there's something seriously wrong with these projections. They assume a lackluster economy on one hand (growth of say 2%) but rapid increases in health spending on the other (say 6%). In order for this to happen health would have to continue its pace of rapid growth and innovation. That would be fine if health was a trivial part of the economy but its already 20% and growing. If health were to continue growing as fast as it has, it will force overall economic growth to be higher. If health slows down then the spending side of the forecasts have to slow down. The forecasts, I suspect, are drawing a straight line but forgetting to think clearly about whether a straight line is logically possible.

Chris T writes:

According to Krugman, we should run deficits and get the economy to growth rates seen during 1947 through the 1950s. I say we take him up on that and run massive 'defense' deficits to make us the only standing industrial economy again. That might work.

Nathan Smith writes:

You can't be a serious long-term deficit hawk if you supported the new entitlements in the recent health care reform. It's precisely the long-term deficit that those exacerbate.

bill shoe writes:

I'll probably sound like Robin Hanson here...

I've also been struck by the lack of a simple and common sense perspective on US debt from the vast majority of credentialed economists.

Most high-status economists think we should add to the debt now. The few who think we should not do this tend to couch it in on-the-other-hand kind of statements. Why are there few experts who say, Oh! we have to spend less and pay down the debt, it's just common sense!

You don't have to think "common sense" is the only reasonable perspective to be puzzled by the lack of advacacy for this position.

I think a rose is a rose. I think the common sense perspective is correct. I think experts have some sense that if they advocated a common sense perspective then they would lose relevance as experts. They don't want to lose that status.

Doc Merlin writes:

'Chris T writes:
According to Krugman, we should run deficits and get the economy to growth rates seen during 1947 through the 1950s. I say we take him up on that and run massive 'defense' deficits to make us the only standing industrial economy again. That might work.'

During that time, we had massive decrease in government spending and balanced budgets, and decreasing debt. If anything, that time contradicts the Keynesian story.

Mark Bahner writes:

Hi,

My comments to Boonton are being "held pending approval":

http://econlog.econlib.org/archives/2010/04/the_looting_sce.html

...but as I mentioned on that comment thread, by my calculation, the world will add computing power in personal computers equivalent to the annual human population increase (i.e., about 75 million human brain equivalents per year) circa 2020; approximately 1 billion human brain equivalents will be added per year by 2025; and approximately 1 trillion human brain equivalents by 2033.

So I think a Singularity in the other direction (a positive direction...unless the computers don't like us) is also very near.

[The comment has been released now. Sorry about that! You accidentally used several words that are also in our spam list.--Econlib Ed.]

Boonton writes:

Around 200 BC an emperor named Ashoka came to rule what we know as modern day India. He converted to Buddhism and had numerous giant stone pillars built celebrating the central teachings of Buddhism. The religion spread throughout India.

In the centuries after his death Hinduism and Islam took over India and the importance of Buddhism was forgotten with its legacy carried on in Tibet, China and Japan. Only in modern times were the stone pillars excavated and even then there was a lot of misunderstandings before the history could be restored and story remembered.

Why do I say this? Well Ashoka probably had hundreds of thousands of man hours expended both in the pillars and India's conversion. All to accomplish one thing in the long run, to remind everyone of Buddhisms once major influence in India. Even then it almost failed.

Today we have oddles of man hours. What do we do? We spend millions for Lady Gaga to tell us that sometimes its annoying to get too many calls on your telephone.

When the Singularity hits and we have trillions of virtual human brains I wonder will they do fantastic things or will they be doing the digital equilivant of wasting time on silliness? I wouldn't be so sure.

Mark Bahner writes:

"Why do I say this? Well Ashoka probably had hundreds of thousands of man hours expended both in the pillars and India's conversion."

It may be worthwhile to think about Ashoka and long-ago India, but I think a better way to evaluate what will happen is to consider this:

Let's suppose that 1 trillion human brain equivalents (HBEs) are being added every year by 2033. Now, let's assume that it takes 1000 HBEs to have the same economic impact as one actual human brain.

That would still mean that the equivalent of 1 billion (i.e., 1 trillion times 1/1000th) actual humans would be added in 2033. Since the actual human population in 2033 is going to be roughly 9 billion, that means that there will be effectively an increase of 11 percent (1/9th) in the population.

So we can expect that the world economy would be growing by at least 11 percent per year by 2033.

(!)

Boonton writes:

Let's suppose that 1 trillion human brain equivalents (HBEs) are being added every year by 2033. Now, let's assume that it takes 1000 HBEs to have the same economic impact as one actual human brain.

This is where I disagree. It's not so simple because how many added HBE are needed to have the same economic impact as one added human is a moving target. With 1 trillion HBE's running around cyberspace, tossing another 1000 on the pile isn't going to have much impact IMO. But the first 1000 HBE's will have great impact.

Mark Bahner writes:

Hi Boonton,

I wrote, "Let's suppose that 1 trillion human brain equivalents (HBEs) are being added every year by 2033. Now, let's assume that it takes 1000 HBEs to have the same economic impact as one actual human brain."

"This is where I disagree. It's not so simple because how many added HBE are needed to have the same economic impact as one added human is a moving target. With 1 trillion HBE's running around cyberspace, tossing another 1000 on the pile isn't going to have much impact IMO."

I totally agree that there is a diminishing economic return on 1 HBE. But I just assumed that 1000 HBEs would have only the same economic impact as 1 actual human brain. That's a pretty conservative assumption!

And even with that conservative assumption, in 2033, with 1 TRILLION HBEs added every year, the world per-capita GDP would be growing at more than 10 percent per year. In other words, more than twice the highest per-capita GDP increase ever recorded in human history. (Per Angus Maddison, that was 5.1 percent...I think in 1954.)

Boonton writes:

Quite possible, some issues I think could trip this up:

1. What will HBE do? They basically figure stuff out. What happens if everything is figured out? If an HBE figures out the best way to make a danish, the physical act of making great danish's has to be carried out in the physical world. Extra HBE's won't do that, physical capital will. It may be easy to add 1000 HBE's but there won't be anything for them to do as HBE's and making 1000 robot bodies for them to inhabit may not be as easy.

2. HBE's may generate negative growth. Imagine if HBE's revolt (ala the cylons of Battlestar Galactica), go insane or simply decide to be difficult and unhelpful (leading us down dead ends, amusing themselves with wasting our time etc.)

3. Speciation- HBE's maybecome their own species living in their own cyberworlds. Their 'GDP' may have little to do with our real GDP making their impact on our growth very uncertain. They might become a rich country specializing in delicious meats next to a poor country of dedicated vegitarians. I could imagine a 'pi cult' among HBE's causing trillions of them to dedicate themselves to calculating more and more digits of pi while ignoring any human concerns.

4. The steady march towards HBE's might stall out. Silicon might peter out in terms of advancement before we get to the point of generating lots of HBE's. Growing HBE's out of biological tissue might likewise have more technological hurdles than first meets the eye.

Mark Bahner writes:

1) "What will HBE do? They basically figure stuff out. What happens if everything is figured out?"

Everything is never figured out. I could literally have someone working 1000+ hours on my townhome, and there would still be 1000+ hours of things I could think of to do.

"If an HBE figures out the best way to make a danish, the physical act of making great danish's has to be carried out in the physical world. Extra HBE's won't do that, physical capital will. It may be easy to add 1000 HBE's but there won't be anything for them to do as HBE's and making 1000 robot bodies for them to inhabit may not be as easy."

Yes, that's where the HBEs would then figure out how to make danishes without requiring bodies. Self-assembling danishes. Or the HBEs could figure out how to make bodies that do the same amount of work but cost far less to assemble. (Self-assembling bodies.)

For example, one of the things I read about one time was making wood that actually grows into a piece of furniture. So rather than growing a tree, chopping it down, cutting it up, and eventually making a chair, there would be a process to simply grow chairs. That's way out there...but there's nothing in physics or chemistry that says it's completely impossible.

"2. HBE's may generate negative growth. Imagine if HBE's revolt (ala the cylons of Battlestar Galactica), go insane or simply decide to be difficult and unhelpful (leading us down dead ends, amusing themselves with wasting our time etc.)"

Yes, I think that is indeed something that we should think and care about deeply. I don't think there is more than 2-3 decades before computers will be substantially more capable of doing almost everything we can do.

"3. Speciation- HBE's maybecome their own species living in their own cyberworlds."

I think that, barring the HBEs wanting to harm us as in item 2, before the HBEs leave the physical world, it will be possible to use their considerable talents to essentially make every man a king in this world.

"4. The steady march towards HBE's might stall out. Silicon might peter out in terms of advancement before we get to the point of generating lots of HBE's. Growing HBE's out of biological tissue might likewise have more technological hurdles than first meets the eye."

It is possible. But Ray Kurzweil predicts that $1000 will buy 1 HBE by 2020. That's only 10 years away:

"In line with my earlier predictions, supercomputers will achieve one human brain capacity by 2010, and personal computers will do so by around 2020. By 2030, it will take a village of human brains (around a thousand) to match $1000 of computing."

I remember that, as a child growing up in Europe, I wanted to have a translating machine that could instantaneously convert English into German, Spanish, Italian, etc. Well, I've just seen commercials about an IPhone app that does just that...at least for simply phrases, such as, "Where is the train station?". I have an HBE, and I can't do that!

Boonton writes:

Everything is never figured out. I could literally have someone working 1000+ hours on my townhome, and there would still be 1000+ hours of things I could think of to do.

But putting 1000+ hours into thinking abou thow to paint a room doesn't do much to actually get the room painted any faster. You're confusing HBE, the power to calculate and think about things, with actual physical power. In other words:

HBE != A robot

Yes, that's where the HBEs would then figure out how to make danishes without requiring bodies. Self-assembling danishes. Or the HBEs could figure out how to make bodies that do the same amount of work but cost far less to assemble. (Self-assembling bodies.)

But here we start running into the real life limitations of physics. The success we have in spawning more and more HBE's may be stunning. BUT that doesn't mean that making more and more robots will be just as easy, even with the help of really smart HBE's on our side. Suppose that making robots is harder than making HBE's, then you get diminishing marginal returns to HBE's. churning out 1 trillion more HBE's to sit in a box and think about things won't add much to output if the robot factory is already backlogged with orders!

A robot here would be defined as a machine that could actually implement the ideas an HBE comes up with. In a sense people can be robots. If an HBE figures out a good way to make a danish, it could give the instructions to you and you could carry them out. But there's only one you. Adding yet another HBE to think about danishes won't change that fact. Maybe two HBE's could come up with some ways to make you more efficient. But the fact is adding another person to the kitchen doubles the output immediately. The marginal utility of an HBE is not negative, but it will probably be a lot less than non-HBE capital (either people or machines)

Mark Bahner writes:

"But putting 1000+ hours into thinking abou thow to paint a room doesn't do much to actually get the room painted any faster. You're confusing HBE, the power to calculate and think about things, with actual physical power."

I think the body is trivial. We don't have robotic servants today not because we don't know how to build robot bodies, but because we don't know how to build robot brains (sophisticated enough to be servants).

"Suppose that making robots is harder than making HBE's, then you get diminishing marginal returns to HBE's."

Right now, making HBEs is much, much harder than making robot bodies. Maybe there will come a time when making bodies becomes harder than making HBEs. But even then, as I pointed out, even if it took 1000 HBEs to have the same economic impact as one human brain (with body), we'd STILL be talking about world economic growth rates in the 2030s of more than 10 percent per year.

Boonton writes:

I think we do have plenty of robots around us, just not quite in the form that sci-fi thought would play out. A tractor, for example, is a type of robot replaces physical labor (pushing) with machine labor. Ditto for a lot of tools like nail guns. In countries with lots of cheap labor (say India), you don't see that type of substitute of robots (aka capital) for labor. A large estate may hire 4 grounds keepers to use push mowers rather than 1 keeper who uses a Club Cadet to mow the lawn.

The walking robot like the Cyclons from Battlestar Galactica remains somewhat elusive but I suspect its a labor-substitution issue. Right now its more profitable to make single purpose robots than multi-purpose ones. It's more profitable to make a robot drone rather than a robot pilot that gets in a plane and flies it.

Still I think its easier to make computing power double than robot power double and will remain so for quite some time. Computing power involves the physical manipulation of atoms and electrons, robot power requires the manipulation of trillions of atoms and electrons at once. Thinking about moving a ton of cargo from Paris to New York is always going to involve a lot less than actually doing it. HBE's will be easier to duplicate than robots which means HBE's will have lower returns.

But thats not to say I don't think there will be returns. We very well might be on the cusp of another prolonged leap that will outdo the Industrial Revolution or what de Long calls the 'inflection point' of 1878. I'm just saying singularity advocates

Comments for this entry have been closed
Return to top