How big is US government debt? If you google this question looking for a number, your first hit is likely to be a site like this, giving a figure (as of June 2016) around $19.5 trillion, or a bit over 100 percent of GDP. That’s the total public debt as reported by the US Treasury.
If you are reading this blog, you probably don’t take that number at face value. You probably know the preferred number is federal debt held by the public. As of June 2016, that’s $14 trillion, or a bit over 70 percent of GDP. That’s the number more likely to be used in academic papers or by official bodies. (Wikipedia seems to mix the two numbers at random.)
Debt held by the public is meant to exclude debt the federal government owes to itself. For the US, that means subtracting the $2.8 trillion in debt held by the Social Security trust fund, the $1.7 trillion held by by federal employee retirement funds, and $1 trillion various other federal trust funds. It leaves in, however, the debt held by the Federal Reserve.
I wonder how many people, the sort of people who read this blog, know that. I wonder how many people know that today, one fifth of the federal debt “held by the public” is actually held by the Fed. I certainly didn’t, until recently.
Here’s a breakdown of federal debt by who owns it. Total public debt is the whole thing. Debt held by the public is the heavy black line. Debt held by the Fed is the blue area just below that line. (Source is various series from the Financial Accounts.)
As you can see, the Fed accounts for quite a bit of federal debt holdings — $2.5 trillion, 16 percent of GDP, or 19 percent of debt “held by the public”.
There’s some other interesting stuff in there. Most obviously, the dramatic fall in the share of debt held by households and nonfinancial businesses (the orange area), and rise of the foreign share (yellow). In the 1950s Abba Lerner could talk with some plausibility about the demand-boosting effects of federal interest payments to households; but it’s silly to suggest — as some modern Lernerians do — that higher rates might boost demand through this channel today. The declining share of the financial sector (red) is also interesting. I’ve suggested that this was a factor in rising liquidity premiums and financial fragility. If, as Zoltan Pozsar argues, we’re seeing a lasting shift from “market liquidity” to “base liquidity” this may include a permanently larger share of federal debt on bank balance sheets.
But what about the Fed share? Should it be counted in debt held by the public, or not? I can’t find the reference at the moment, but I believe there is no consistent rule on this between countries. (As I recall, the UK excludes it.) In any case, the phenomenon of large central bank holdings of government debt is not unique to the US. Here, from the OECD (p. 41), are the shares of government debt held by central banks in various countries:
If you want to say that debt held by the Fed definitely shouldn’t be counted, I won’t object. After all, any interest earnings on the debt are simply returned to the Treasury at the end of the year, so this debt literally represents payments the government is making to itself. But that’s not what I want to say. To be honest, I can see valid arguments on both sides — yes, the Fed is a part of government just as much as the Social Security Administration; but on the other hand, the Fed’s holdings were acquired in market purchases from the private sector, while the holdings of the various trust funds are nonmarketable securities that exist only as bookkeeping devices for future payments to beneficiaries. And if you think the Fed will reduce its holdings in the near future, then it makes sense to count them for any target you might have for holdings by the private sector. But of course, in that case how much you count them will depend on whether, when and how much you think the Fed will unwind its 2009-2013 balance sheet expansions. And this is my point: There is no true level of the federal debt. The “debt” is not an object out in the world. It is a way of talking about some set of the payment commitments by some set of economic units, sets whose boundaries are inherently arbitrary.
Again, debt “held by the public” does not include the notional debt in the Social Security Trust Fund, or in the various retirement funds for federal employees. But what about the debt (currently about 5 percent of GDP) held by state and local governments in similar trust funds? Fundamentally, these represent commitments by the federal government to help with pension payments to retired state and local government employees. But this is the same commitment embodied in the Social Security Trust Fund. And on the other hand, the federal government has a vast number of payment commitments to state and local governments — transfers from the federal government make up more than a quarter of total state government revenue. Why count the commitments that happen to be recorded as debt holding in retirement funds as federal debt but not the rest of them?
For that matter, what about the future claims of Social Security recipients? They certainly represent payment commitments by the federal government. Lawrence Kotlikoff thinks there is no difference between the commitment to make future Social Security payments and the commitment to make payments on the debt, so we should add them up and say debt held by the public is over 200 percent of GDP. Other people want to add in public pensions of all kinds. Why not throw in Medicare, too? True, retirement benefits are not marketable, but checking your expected benefits at https://www.ssa.gov/myaccount is not much harder than checking your bank balance online. And for the MMT-inclined, don’t future Social Security benefits have as good a claim to be “net wealth” for the private sector as federal debt, maybe better?
One takeaway from all this is the point eloquently made by Merijn Knibbe, that one of the big problems in the economics profession today is the complete disconnect between theory and measurement. Most public discussions and economic models — and a lot of empirical work for that matter — treat “debt” as an object that simply exists in the world. (It’s worth noting that the question of how exactly debt is defined, and who it is owed to, does get some attention in undergraduate econ textbooks, but none at all in graduate ones.) It seems to me that the large share of debt held by central banks is a case in point of how we have to make a conscious choice about which commitments we classify as “debt”, and recognize that the best place to draw the line is going to depend on the question we’re asking. We need to treat economic categories like debt not as primitives but as provisional shorthand, and we need to be constantly walking back and forth between our abstractions and the concrete phenomena they are trying to describe. You can’t, it seems to me, do useful scholarship on something like government debt, except on the basis of a deep engagement with the concrete practices and public debates that the term is part of.
More concretely: Whenever you take a functional finance line, someone is going to stand up and start demanding in a prosecutorial tone whether you really think government debt could rise to 10 times or 100 times GDP. How about 1,000 times? a million times? — until you say something noncommittal and move on to the next question (or mute them on Twitter). But of course the answer is, it depends. It depends, first, on the concrete institutional arrangements under which debt is held, which determine both economic impacts and financial constraints, if any. (For example, whether the debt held by central banks should be counted as held by the public depends on when or if those positions will be unwound.) And it depends, second, on how we are counting debt.
Consider a trust fund holding federal debt. What the federal government has actually committed to is a stream of payments in the future which in turn will allow the fund to fulfill its own payment commitments. Converting that flow of future payments to a liability stock in the present depends on the discount rate we assign to them. But we can follow that same procedure for any future spending, whether or not it is officially recognized as someone’s asset. As Dean Baker likes to say, given that we don’t prefund education, the military, etc., pretty much all government spending could be called an unfunded liability for the federal government. How big a liability depends on the discount rate. If the discount rate is less than the nominal growth rate, then the present value of future spending grows without limit as we consider longer periods.
Here’s an exercise. Let’s go full Kotlikoff and call all future government spending a liability of taxpayers today. Say that federal spending is a constant 20 percent of GDP and nominal growth is 5 percent per year. If we use the current 10-year Treasury rate of around 2 percent as our discount rate, then the present value of federal spending over the next 20 years works out to, let’s see, $10 quadrillion, or 55,000 percent of GDP. That’s $30 million per person. Whoa. Can I have a Time magazine cover story now? [No I cannot, because I am bad at math. See below.]
So yeah. 20 percent of debt “held by the public” is actually owed to the Fed. An interesting fact which perhaps you did not know.
UPDATE: As commenter Matt points out below, the math in the next-to-last paragraph is wrong. The calculation as given yields $110 trillion, a measly 600 percent of GDP. On the other hand, if we stretch it out to the next 30 years, we get nearly $200 trillion, which is 1,000 percent of GDP or more than $600,000 per person. I guess that will do.
Shouldn’t be surprising.
The SNA treats the central bank as part of the financial sector.
Second, purely as a consistency check you have to count the debt held by the Federal Reserve because reserves and currency notes are liabilities of the Federal Reserve.
In other words, if the Fed buys $1tn of debt from the market, it shouldn’t make a difference in the net debt of the government + the central bank.
But if you do not count debt held by the Fed as “debt held by the public”, you’ll see a reduction of liabilities of “central bank + government” if the Fed buys bonds from the market.
So purely for consistency, the debt held by the Fed is counted.
Just simple accounting.
The SNAs are not the final word. I am sure you will agree that there are some contexts in which it makes a difference how much of “financial sector” holdings actually belong to the central bank.
On your second point, you are right that if we are going to net out central bank holdings of government debt, there is an argument for adding central bank liabilities. (I meant to say this in the post — thanks for mentioning it.) But you are wrong to say that which way we do it makes no difference. Central banks hold private assets as well, so reserves + currency are greater than public debt holdings.
if you do not count debt held by the Fed as “debt held by the public”, you’ll see a reduction of liabilities of “central bank + government” if the Fed buys bonds from the market.
That’s right. And for some purposes that may be the right way to look at it.
purely for consistency, the debt held by the Fed is counted.
There are lots of ways to count things consistently.
True the SNA is not the final word. However it makes some genuine effort in defining things correctly and that’s the reason statisticians try to adopt it more and more.
It’s fine if the Fed holds private assets. It really doesn’t affect the argument that the government debt held by the Fed should be counted in the measurement.
Because when netting, the assets nets with liabilities created. So if the Fed buys $1 bn of corporate bonds, its assets and liabilities rise by $1bn. So that doesn’t affect the correct net number.
It will affect the future assets and liabilities because interest payments and market fluctuations change numbers. But that’s in future.
So if the Fed buys $1bn of corporate bonds, while its liabilities are rising but it nets with the assets.
I mean imagine the central bank buys government bonds worth 20% of GDP and a politician comes and says “we’ve reduced the debt”. This example highlights that that’s not the way to count.
Of course the question whether the absolute number is important is another question but nobody should argue that “our debt has reduced” if the central bank buys government bonds.
Your point about social security is interesting but it isn’t exactly identical because social security works in a different way. Again the SNA addresses this issue correctly.
About consistency: self-consistency leads to uniqueness in various ways. For suppose the central bank buys all government debt. Would one say the government’s debt is zero?
I do not see any reason to change measurements so that the Fed’s holding is not counted.
I don’t share your belief that there is one correct way.
You simply assert that it doesn’t make sense to say that when a central bank purchases government debt, the debt has been reduced. My position is that whether or not it makes sense depends on why you are interested in the level of the debt. For instance, if you are concerned about the burden of future interest payments, they have been reduced just as much as if the debt were retired in some other way.
That isn’t true. Central banks pay interest on reserves. Plus it’s not the case that the central bank purchase is permanent. So the Fed is looking to reduce its balance sheet in some 10-15 years. And even now it pays interest on reserves.
But if you treat it as if the debt is reduced, it doesn’t make sense because that ignores interest costs.
So the Fed holding of Treasury is a closer approximation to the truth (that the government is paying interest). It’s effectively (indirectly) paying interest even on the debt held by the Fed to the private sector. It’s of course lesser than the case if the bonds were held directly by the private sector (because the Fed is making profits on its holdings and remitted to the government) but that is closer to the truth than saying that it’s as if the debt held by the Fed is not effectively in existence.
In effect, if public debt is 80%, central bank holding is 20%, 80% is still closer to truth than 60% because the latter would ignore interest on central bank liabilities.
Central banks pay interest on reserves. Plus it’s not the case that the central bank purchase is permanent.
Maybe, maybe not. Depends on the bank, and the time period. My point is NOT that we definitely should exclude central bank holdings from the headline number, it’s that the right choice depends on context. The factors you identify are good reasons to include central ban holdings in the “public” number, but those factors do not always apply.
In effect, if public debt is 80%, central bank holding is 20%, 80% is still closer to truth than 60% because the latter would ignore interest on central bank liabilities.
But what if central bank liabilities pay much lower interest than government debt (as they do)? Then maybe the right number in this case would be 65%, or 70%? Again, I’m not advocating for a particular number, I’m advocating for a critical attitude toward these statistics — for thinking about their relationship to the concrete reality they describe. You have to remember that the map is not the territory.
Well how about inverted yield curves?
I am fine with reporting 60% debt held by the private sector and 20% reserves as opposed to just saying the debt held by the private sector is 60%.
At another level, I am not for being apologetic for public debt. In the example above, 60% sounds a bit apologetic. One needs to say that the absolute level by itself is not a bad thing as understood.
You should never go full Kotlikoff.
There’s actually no convincing model that says anything about the debt-to-GDP ratio in the abstract; we need to look at how the holdings affect the behaviour of the owners of the debt.
To do that, we would need a solid model of macro behaviour. Good luck with that.
Agree 100%. My point was just that if you think of “debt” in the most generic sense — as a stream of future payment commitments discounted to the present — then we can come up with lots of possible numbers for the debt-GDP ratio, including some very high ones. To decide which number is actually relevant, we have to ask — exactly as you say — about how these future commitments affect behavior.
The Time magazine economics writer Rana Foroohar isn’t that bad, but that cover was groan-inducing.
I would be interested in some intelligent writing about the late-80s, early ’90s period, where Clinton’s New Democrat turned to deficit reduction to bring down interest rates. What is the accurate economic history?
I subscribe to Dean Baker’s and others’ narrative about how a deregulated/top-heavy economy became more and more bubble-blowing (even as monetary/fiscal policy has become too tight during, through and after the Great Recession).
Yes, I like Rana. Jacobin asked me to review her new book, which I will try to do soon.
Have you looked at Stiglitz’s book on the 90s?
During late 2008 financial crisis Fed first effort was to lend reserves to banks and sell Treasuries to drain reserves. But this meant Fed would run out of Treasuries to control the fed funds rate. Also Treasury sold extra debt to help Fed drain reserves. By early 2009 Fed switched to Quantitative Easing where it could purchase financial assets from non-banks. QE with non-banks injects excess reserves into the aggregate bank and injects transaction accounts into the aggregate non-bank. The FDIC increased insurance limits on transaction accounts so we see that Fed/FDIC were providing insured assets to banks (reserves) and non-banks (FDIC deposits). Some assets held by Fed under QE are debt owed by market sector and direct student loans held by Department of Education are consumer loans held by the federal government. The lesson from functional finance is that the combined Fed and Treasury stand ready to perform certain functions that are necessary to serve the public purpose and have done so with a fair amount of success despite the complexity of the political and financial system. The MMT position of combined Fed and Treasury is useful for analysis however there is not sufficient discussion of how the federal government can be captured by special interests when it acts like a financial intermediary. Sallie Mae held assets with cash flow guarantees by the Department of Education financed with the right to borrow from Treasury via the Federal Finance Bank (FFB) so it more resembled an off-balance sheet branch of government than a private corporation. Who had a cash flow incentive to strip student borrowers of bankruptcy protection from 1976-1998? Sallie Mae and ED because ED had to obtain fed funds to pay student loan defaults. Congress probably responded to political insider games via Sallie Mae and ED when it made honest but unfortunate student borrowers into scapegoats in bankruptcy. The main problem with functional finance is that the government allocates cash flow guarantees and bankruptcy rights via a political process subject to capture by special interest groups rather than for the greater public purpose.
Right.
I agree that a (maybe the?) big problem with MMT/functional finance is that it ignores the political economy side — it doesn’t ask what purpose the myth of financial constraints on the state is supposed to serve. Warren Mosler likes to say that the government can’t run out of money anymore than a referee can run out of points. This is true — but no game would work if the referees handed out points without limit, they only work if points are treated as scarce.
The GSE piece is important too. Not relevant for this particular post since they have only trivial holdings of Treasury debt but in other contexts how they get classified could make a difference.
I agree the level of govt debt depends on a valuation process which is somewhat arbitrary. I also agree the points on the scoreboard must be scarce per the rules of the game. In the context of United States political and economic system this means saving and cash flow insurance is allocated by rules set in the legislature that include automatic stabilizers, fiscal policy, and bankruptcy laws. A problem with govt acting as intermediary is that a loan default is recorded as a credit to accounts receivable (write-off) and a debit to net worth which means defaults to the govt tend to decrease the net worth of the federal govt. In the private sector these write-offs go away in bankruptcy. In the public sector govts either tax to get net worth closer to zero or float more debt that must be rolled over in perpetuity. So functional finance means allocating cash flow insurance and bankruptcy rights in addition to subsidizing some activities that have a public purpose. Mosler does propose significant limits on banks although he recommends unlimited govt insurance of bank liabilities for solvent banks that play by the rules if I understand his position correctly.
“I agree that a (maybe the?) big problem with MMT/functional finance is that it ignores the political economy side — it doesn’t ask what purpose the myth of financial constraints on the state is supposed to serve. Warren Mosler likes to say that the government can’t run out of money anymore than a referee can run out of points. This is true — but no game would work if the referees handed out points without limit, they only work if points are treated as scarce.”
I don’t think is remotely accurate. I’ve seen a lot of discussions of the purposes of these myths from the “founders” of MMT. Further all of them have talked about the real resource costs of acquiring money (most significantly, working) and how what the token is “given away” for determines how valuable the token is. This is an essential part of the JG.
They all emphasize the real purpose is to extract real resources
Well, I’ve read their stuff too, and I’ve also spoken with most of these guys. And I don’t think they are interested in the idea a coherent capitalist class interest that is served by the mythology of financial constraints on the state. But maybe I’m wrong.
I think you meant NPV over 100 years? Otherwise it’s closer to 100 trillion, not 10 quadrillion.
Either way, this is maybe a less-than-perfect exercise in the inverse of the stylized fact in Capital in the 21st Century, where if g>r everything explodes to infinity.
OK, let’s see. ∫x^1.03 from 1 to 20 is 215x, more or less. Current government spending is $4 trillion, times 215 is $860 trillion … so, yes, you are right. Oops. 30 years would get us 2 quadrillion, which I think strikes the right balance between vaguely reasonable sounding calculation and obviously nuts conclusion.
I agree it’s kind of a silly exercise. Why did I do it? First, as a reductio of the arguments for counting all kinds of liabilities in the debt. Second, to make the point that the future payments explicitly recorded on balance sheets are just a small subset of all payment commitments. (This is why I’m a bit more tolerant of accounting inconsistencies than some of my brothers in heterodoxy like Ramanan here or David Rosnick.) And third, to challenge the argument from incredulity that debt of is obviously unsustainable.
That number is still off by a factor of 10, but we can agree on the general point. The NPV of future commitments to spend money enforces an intertemporal budget constraint. Some of the commitments are pretty much just as “rock solid” as debt repayment and therefore put just as much pressure on this constraint.
That’s an interesting point, though I’m not sure it is more effective in marshaling a case for your preferred understanding of our finances (I hope I’m not misconstruing your preferences) or for the Simpsons and Bowleses of the world.
To me, the overall financial situation of Americans (summing private and public indebtedness) is the most persuasive argument in your favor. This is consistent with your dig at the Time Magazine cover, which looks at one very small part of the overall picture – what about the assets?
Crap, you’re right. How did I screw this up twice?
I’m not sure it’s effective either. If I waited til I was sure, I’d never post anything.