Do Keynesians believe their own models?

I ask this as a quasi-Keynesian myself. I'm not (merely) trying to score points.

[Update 2. Paul Krugman weighs in. Like Brad DeLong, and Scott Sumner (in comments), he questions my assumption that the central bank can keep the interest rate fixed, without buying all the outstanding bonds.]

Take the standard ISLM model straight off the shelf (or straight out of the second-year textbook). Draw the LM curve horizontal; either because the central bank conducts monetary policy by setting a rate of interest, or because the economy is stuck in a liquidity trap where money and government bonds are perfect substitutes. Now let's hit it with a shock.

The shock is that all the bond rating agencies downgrade the rating on government bonds. Specifically, the rating agencies say that whereas the previous default risk was zero, there is now a 1% chance per year that the government will renege on 100% of all promises to pay money on its bonds. (Or a 10% chance per year of reneging on 10% of repayments, whatever.) And assume that everybody believes the rating agencies. Further assume that the central bank holds the interest rate on government bonds constant, by offering to buy and sell unlimited amounts of government bonds at prices commensurate with that same interest rate. [Update: but is this assumption reasonable? See comments by Brad DeLong, Andy Harless, Scott Sumner, and me, below]. Further assume that fiscal policy stays constant.

What happens?

Any good economics student who has completed intermediate macro and really understood the material and has the capacity to think (rather than just memorise results for the exam) ought to be able to take a decent crack at this question.

My answer to this question is the same as the standard textbook answer to the question where the shock is an increase in expected inflation by 1% (one percentage point, strictly).

The IS curve is drawn as a function of the expected real interest rate, because desired savings and investment depend on the expected real interest rate -- the nominal rate minus expected inflation. And the LM curve is drawn as a function of the nominal interest rate. Because the demand for money is a function of the nominal interest rate, or, in this case, because the nominal interest rate is what the central bank sets.

There are three ways to handle a 1% increase in expected inflation in the ISLM diagram: first, put the nominal rate on the vertical axis, and shift the IS curve vertically up by 1%; second, put the real rate on the vertical axis, and shift the LM curve vertically down by 1%; third, put both interest rates on the vertical axis, but put a vertical 1% wedge between the IS and LM curves to the right of the intersection point. I prefer the third, "wedge" method. But all three ways give exactly the same answer. The 1% increase in expected inflation causes the ISLM equilibrium point to move to the right, to a higher level of real income and lower real interest rates. The increase in expected inflation causes the Aggregate Demand curve derived from that ISLM model to shift right in {price level; real income} space. This is standard second-year stuff.

And if you: did believe in the ISLM model; and did believe that the economy was stuck in a liquidity trap; and did believe the economy needed an increase in Aggregate Demand; and did believe that fiscal policy either should or could not be used, for whatever reason; you would say that this increase in expected inflation is a good thing.

And, at first sight, the answer to the question "what is the effect of a 1% increase in perceived risk on government bonds?" is exactly the same as the answer to the question "what is the effect of a 1% increase in expected inflation?". At least, in this model, given my assumptions. Both result in the same 1% fall in the real risk-adjusted rate of interest on private saving and investment and the same rise in AD and real income.

And the ISLM model is no straw man. Simple, yes. Crude, yes. But straw, no. No model that has survived 70 years is a straw man. Paul Krugman likes it. So do I, sort of. And it's not really that different from the canonical New Keynesian model, at least for this question.

So, why aren't US Keynesians cheering the possibility of budget impasse and the downgrading of US bonds? Or did I miss it?

That's not necessarily a rhetorical question. There may be answers. But they aren't in the model.

1. Yes, a downgrading of US government bonds would be a good thing in the short run, because it would enable the economy to escape recession, but it would be a bad thing in the long run by raising the costs of rolling over and financing the debt when the economy gets back to normal.

2. Yes, a downgrading of US government bonds would be a good thing in the short run, because it would enable the economy to escape recession, but it would be a bad thing in the long run because bond prices would fall when the economy gets back to normal and this might cause some banks to get into trouble. (But so would increased inflation).

3. A promise is a promise, and it's bad if you break your bond even if good things happen.

4. ?

Is anybody out there celebrating this rather large silver lining? Or don't Keynesians believe their own models?

Comments

[Determinant]

I believe the problem is the same as the last Credit Crunch. Here goes my explanation from the back of the envelope.

1) The current money supply is multi-layered and US Treasuries form a good deal of the base.

2) An increase in the perceived risk of Treasuries means that banks, which create the functional money supply the economy uses to conduct real trade in goods and services, would need to increase their capital *before* increasing loan volume. Existing interest rates would rise but loan volumes would fall. This would cause a contraction of the real usable money supply.

3) That contraction is yet another example of "tight money". In a monetary economy where all transactions feature money, the money supply is the slate on which all economic transactions are written. (2) means the slate gets smaller and we will feature deficient Aggregate Demand.

4) The net result is a lower volume of transactions in real goods and services due to sticky prices. Sticky Prices come from fixed contracts, the information content of prices, the usual suspects.

Bottom line, a lot of people are going to get hurt.

[Andy Harless]

I think there is a plausible case that a US Treasury downgrade could be beneficial to the US (and even that the short-run benefits might be sufficiently persistent to outweigh the long-run costs). In fact I've made the case myself that a decrease in market confidence in US Treasury securities (which is not quite the same thing as a downgrade by a rating agency) would be a good thing. On the other hand, the main thing about a US Treasury downgrade is that the results are highly unpredictable. If you suddenly take the recognized standard risk-free asset and formally declare that it's actually riskier than a lot of other assets, the sum of all the technical and micro-level problems that are created in the short run might outweigh the macro fundamentals. It's like when a company replaces an existing piece of software with one that behaves completely differently. The new software may be better than the old software, but it's going to create havoc.

[Andy Harless]

Determinant,

But can't the Fed offset the problem that you cite? Sure, Treasuries are quasi-money and they'll turn into non-money, which will constitute monetary tightening. But the Fed (which surely realizes this problem exists) can still exchange actual base money for this quasi- or non-money. The problem right now, arguably, is that the Fed has nearly reached its limit in the ability to create money, because all it can do is exchange one form of money for another. If you declare that the second form is no longer money, you open up a range over which monetary policy is once again effective. You should be able to get back to where you started out by taking full advantage of that range. (And if you do so, you should actually end up better than where you started, for the reasons that Nick argues.)

[Nick Rowe]

Determinant: but the central bank keeps the interest rate on government bonds constant, so the price of government bonds (treasuries etc.) constant too. But since government bonds are now riskier than before, the price of private bonds would rise (interest rates on private bonds would fall), since private and government bonds should yield the same risk-adjusted and inflation-adjusted rate of return. So the fall in private interest rates means there's an increase in desired investment and a fall in desired saving, at the old level of real income, so real income rises until desired saving and desired investment are equalised again.

[Nick Rowe]

Andy (I see your reply to Determinant is about the same as mine):

"On the other hand, the main thing about a US Treasury downgrade is that the results are highly unpredictable. If you suddenly take the recognized standard risk-free asset and formally declare that it's actually riskier than a lot of other assets, the sum of all the technical and micro-level problems that are created in the short run might outweigh the macro fundamentals."

Maybe. I would like to see that spelled out more. It does sound a bit like "The confidence fairy might get angry at us" ;-)

[Andy Harless]

To expand on my first comment: one thing about the technical issues is that they may result in direct impacts that are paradoxical. It would seem natural that a downgrade of US Treasuries would result in a narrowing or inversion of the spread between other high-quality bond yields and those of US Treasuries. But people who actually work with these things aren't sure that is what will happen. In fact, many of them think the opposite will happen -- that spreads will widen, and people will actually be willing to pay a larger premium for the US Treasury's downgraded bonds than they were willing to pay for today's AAA Treasury bonds. This may contradict the assumption that "everybody believes the rating agencies," and maybe that assumption is the real problem. It's not clear that credit ratings for the most important sovereign issuer in the world -- about which everyone has easy access to the same information as the rating agencies and a lot of people spend a lot of time thinking about it -- add any value at all. Rather the US AAA rating is just a convenience which helps the system work more smoothly. Take away that convenience and all you do is screw things up -- in a way that has no impact on market judgments of the US Treasury's creditworthiness but that has a substantial impact on the actual creditworthiness of other issuers.

[david]

"Don't reason from a price change"? An increase in expected inflation is generally modeled as exogenous, executed or at least tacitly accepted by the independent central bank. Debt default is generally not. Fiscal policy is not, in fact, going to plausibly remain constant under conditions of debt default.

Aside from that, don't real-life sovereign defaults result in short-term chaos rather than short-term benefits? Monetary tightening can just as easily result from an increase in uncertainty, and in the case of debt default, someone is going to be revealed to be much less wealthy than they thought - it's just a matter of finding out who - and so cue flights to security, etc. in the form of durable goods under your bed as a real possibility. Central banks cannot very well carry out OMO when maintaining civil order is itself problematic. Argentina, Russia, etc.

[Brad DeLong]

I think you have gotten this wrong. It is true that on the IS curve the same nominal interest rate now corresponds to a 1% lower real interest rate on the goods-bonds margin--that with the nominal interest rate on the y-axis the IS curve shifts up by 1% point. But something also happens to the LM curve: the same nominal interest rate is now a !% lower opportunity cost of holding cash, so the LM curve would also, I think, shift upwards by 1% point--except for the zero nominal bound...

At the same nominal interest rate, therefore, you hold more cash and also are more eager to borrow and spend on currently produced goods and services. It doesn't look to me like an increase in expected inflation at all...

If the government were going to default on its bonds *and* default on its cash--well, that is what a rise in expected inflation is, isn't it?

[Nick Rowe]

David: but monetary policy could also be argued to be non-exogenous in the event of a change in expected inflation. Yet we do sometimes ask about the effects of a change in expected inflation. For example, we might say that a change in the future money supply will change the future price level, and change current expected inflation. But we often distinguish between the effects of the change in the future money supply from the effects of *news* about the change in the future money supply. I see this as similar. The population receives news about the future likelihood of default. What is the effect of that news?

You might argue that this news would also affect people's expectations about the position of the future IS curve. And this might also affect the current equilibrium. OK. That could change the analysis.

[Nick Rowe]

Brad: Welcome!

I love the smell of really having to think about macro, early on a Monday morning!

I see the logic in your answer, but is it the *Keynesian* logic of the ISLM with horizontal LM (or equivalently the New-Keynesian/Neo-Wicksellian logic of Woodford et al) where the central bank chooses the nominal rate of interest??

If we put the nominal interest rate on government bonds on the vertical axis, will the central bank be able to prevent the LM curve shifting vertically up by the full 1%? If the bank can do this, there will be an expansion of AD. If it cannot, AD will stay the same.

"If the government were going to default on its bonds *and* default on its cash--well, that is what a rise in expected inflation is, isn't it?"

Agreed. (With *default* on its cash in scare quotes, since cash is irredeemable, and not a promise to pay anything; so a "default" on cash would be like repudiating existing notes, not enforcing counterfeit laws on them, and printing up a new batch of different notes that people had to buy). In that case it's exactly like expected inflation. It's like people learning that holding government bonds *and* holding cash causes you to catch cooties. People want to substitute away from cash and government bonds into buying goods.

Suppose instead that people learn that holding government bonds causes cooties, but holding money doesn't cause cooties. So people want to substitute away from bonds into holding money and buying goods. And that's different, at least initially. But what happens next depends on how the central bank responds. If the central bank responds by swapping money for bonds to keep the price of bonds constant, there should be no excess demand for money in terms of bonds, so the net effect should be an increased demand for goods, just like in the above case.

In the case of Greece, sure, I can see that the central bank of Greece is unable to prevent the interest rates on Greek government bonds rising by the full amount of the risk of default. Default on Greek bonds and default on Greek money are very different things. Because Greek money isn't Greek. But in the case of the US, (or Japan for that matter), can the Fed prevent the interest rate on government bonds rising by the full 1%?

[Nick Rowe]

Yep, I think that's right. Brad is, in effect, questioning my assumption that the central bank will be able to prevent the LM shifting up, and says it must shift up by 1% -- that the Fed won't be able to prevent nominal interest rates rising by a full 1%. It may depend on how you see banks' reserves at the Fed. Are they a government bond, since they are a promise to pay by a government entity?

[Andy Harless]

Brad,

I think your argument contradicts Nick's assumption, which I think is a reasonable assumption:

"...assume that the central bank holds the interest rate on government bonds constant, by offering to buy and sell unlimited amounts of government bonds at prices commensurate with that same interest rate."

The shift in the LM curve is therefore offset by a contrary shift deliberately induced by the central bank. So the opportunity cost of holding cash is unchanged: it is always the nominal interest rate.

If the central bank feels that it is constrained by the zero lower bound, then it is likely to induce such a contrary shift, since it would have preferred a lower interest rate in the first place. If it does not feel constrained, it will still induce a partial contrary shift so as to offset any contractionary impact from the downgrade. (Obviously if the central bank is unconstrained, then under certain assumptions about its reaction function and the SRAS curve, any exogenous event on the demand side becomes irrelevant, so the whole argument is moot.)

[Nick Rowe]

Andy: I now think the key question is whether my assumption is reasonable or not. I can imagine a world where it isn't, so that the central bank will be unable to prevent the nominal interest rate from rising 1%. Though its' hard to square that intuition with the "horizontalist" idea that the central bank can lend as much as it likes at 0.25%.

[Nick Rowe]

Or maybe I've just been reading too much MMT and Neo-Wicksellian "cashless" stuff ;-)

[Scott Sumner]

Well that's certainly some unpleasant fiscal arithmatic! Needless to say I love this post. The only question is which part of the Keynesian model is wrong. Is it the liquidity trap assumption? Or the assumption that interest rates are the transmission mechanism for monetary policy? Why not both?

But let me play the devil's advocate. As I recall when you are in a liquidity trap the expected short term return on all government bonds is zero. (I'm using the expectation's hypothesis.) So if the default risk increases, then the expected return goes negative at a given nominal rate. In that case the Fed buys the entire stock of government debt. And risk adjusted interest rates on private debt don't fall at all, as the risk free rate on Treasury debt stays at zero. Note that $100 bills are now the only "Treasury debt." So I don't see the magic bullet here. But I'm probably missing something, as IS-LM is not my forte.

[Nick Rowe]

Scott: It's good to work through these thought-experiments. Push everything to the limit.

I think you are a good advocate for the Keynesian "devil"! If we literally believe in the zero lower bound liquidity trap, and if we believe that there's something called "cash" that people can hold and stays as good as before, then I think the Fed would have to buy up the entire national debt in order to keep the nominal interest rate at 0%.

[Morganwarstler]

The problem is a central bank that is allowed to choose to catch cooties. They should be screaming sell! into the phone.

[Nick Rowe]

Morgan: Who should be screaming "sell", to whom? And what do they sell? And what do they buy or hold instead?

Arnold Kling also has an interesting response: http://econlog.econlib.org/archives/2011/07/models_vs_hand-.html

[Andy Harless]

I would say the reasonableness of the "horizontal LM" assumption (at least with respect to the Fed today) hinges on the term structure of interest rates and the Fed's hesitancy to use unconventional policy. It's just implausible to me that the Fed would allow short-term Treasury rates to rise significantly in the face of a downgrade. (The behavior of T-bill yields in recent weeks suggests that the market agrees with me.) On the other hand, in the immediate time frame, the Fed is not likely to resist increases in longer-term rates, and subsequent action is unlikely to fully offset any increase in such rates. Thus (unless the expectations hypothesis holds perfectly) there would effectively be some net shift in the LM curve, but my guess is that the shift would not be enough to offset the improvement in the "wedge."

I continue to think that the more problematic assumption is that everyone believes the rating agencies. I think it would be more accurate to see market confidence in the US Treasury as a separate variable from the Treasury's ratings -- one that has near-zero correlation in the very short run (except when a discontinuous event such as a default happens). Reduced confidence in the Treasury is good for the US economy, but a loss of the AAA rating is bad.

[JW Mason]

This is very smart.

Seems to me it's even more relevant as an answer to the question, "Why hasn't default risk affected the price of Treasury securities?" than as an answer to the question of how higher default risk might affect output and income in the future. We've already gotten news about a higher probability of default. If DeLong's argument were correct, wouldn't we already have seen a fall in the price of bonds?

[Scott Sumner]

Nick, You said;

"If we literally believe in the zero lower bound liquidity trap,"

But now I'm confused. I thought we HAD to believe that for your example to work. If we aren't in a liquidity trap (as I believe), then risk free real rates, inflation, NGDP, etc, are all determined by monetary policy, and hence this interest bearing debt problem could not affect AD. Did I misunderstand your original example?

[Scott Sumner]

I should add that I was assuming zero default risk on cash, which seems reasonable to me.

[Nick Rowe]

Scott: but the question is: is my assumption that the central bank keeps the nominal interest rate constant consistent with one possible rationale I gave for that assumption - namely a liquidity trap?

[Nick Rowe]

Andy: I'm leaning towards something like your last comment. Unfortunately, the ISLM, (and most simple New-Keynesian models) only really have one interest rate, rather than a whole term structure.

JW. Thanks!

JW and Andy: there is something problematic about default risk on bonds that promise to pay money you can print yourself. Default *can* happen, if the fiscal authorities cannot overrule an independent monetary authority, or if the fiscal+monetary authority decides that the consequences of default are worse than the consequences of printing whatever it takes to prevent default. But it is hard to imagine that default on bonds is independent in practice from resort to the inflation tax.

[Tom Powers]

How different is this from the fiscal theory of the price level? (Not that I totally understand it, but those guys always seem to be talking about how worsening of the government's financial situation is equivalent to inflation.)

[Matias Vernengo]

How about because the problem of the budget impasse is that the Fed will not be able to monetize debt (and not the downgrading of the debt), and maintain interest rates constant, forcing a massive cut in spending and a brutal collapse of demand. How would anybody think that there is a possibility of an increase in inflation with that scenario?

[Nick Rowe]

Perhaps, what I should be attacking is not Keynesians per se, but that version of Keynesianism which says "Forget money; what matters is only the choice between bonds and newly-produced goods, and recessions are caused by an excess demand for bonds relative to newly-produced goods." Because in that case, anything that makes people dislike holding bonds will cure the recession. What matters, IMHO, is not bonds vs newly-produced goods, but money vs newly-produced goods. The cure for the recession is to get people to believe that holding money gives you cooties.

[Andy Harless]

Oh, dear, I think Scott is right. I take back what I wrote before (when I hadn't seen Scott's comment yet). If we are talking about the Fed today and we think the short-term interest rate is the relevant one, then AAA-rated money would dominate AA-rated T-bills, which would simply disappear into the Fed's portfolio. And if longer-term rates are what matter, then Nick's assumption is implausible given the way the Fed actually appears to behave. So, no, an increase in Treasury default risk is not expansionary in the Keynesian model, given empirically plausible assumptions about the Fed's reaction function.

[Nick Rowe]

Tom: FTPLers lump together money plus bonds. It's all just "government liabilities". Money is just a slightly different form of bond. Most FTPLers assume perfectly flexible prices and full employment, but lets relax that, and assume we start in a recession. In that case, a fall in the PV of future surpluses, causes an excess supply of government liabilities, at current prices and interest rates. That would have to spillover into an increased demand for private investment and newly-produced goods, which would cure the recession. Or, if prices were perfectly flexible, into an increased price level. So yes, sticky-price FTPLers should welcome a worsening of the governments' fiscal situation, I think.

Matias: would the Fed be unable to monetise debt if Congress fails to reach an agreement on the debt ceiling? (That doesn't seem likely to me, but I haven't really been following the US laws on this.)

[K]

Great post, Nick. Really fun!

Nick: "But it is hard to imagine that default on bonds is independent in practice from resort to the inflation tax."

Exactly.  And a real default, with actual debt repudiation, will only ever happen for a sovereign issuer in conjunction with hyperinflation (Like the Ruble crisis) where there are no longer any available palatable outcomes.  Absent that, payments may be deferred by a dysfunctional fiscal authority, but they *will* be made. The reason for that is that even deferral will turn out to be *extremely* painful, and there is not a chance in hell that it will continue indefinitely. And deferral by a few months is of no economic consequence at the zero bound (in terms of interest rates - it *will* cause some Y2K havoc in the financial sector, and deferral of payments in government programs will be economically disastrous which will cause a major treasury bond rally).  So even ignoring the question of whether the CB can keep long term yields constant as default risk increases, is the question of whether it's possible to convince the market to price *any* default risk *at all*.

Currently the market for USA CDS is around 55 bps, up from 40 a few months ago.  Lest anyone should think that means the market thinks there is a real chance of taking real economic losses on a US default, what this really means is that there is a very high probability of a trigger of USA CDS, but with extremely high recovery (e.g. CDS will settle on some low coupon bond issued in Dec '08, currently trading at $97, or whatever the price of the cheapest-to-deliver). This is what happened when Fannie/Freddy "defaulted." The fact that USA CDS is not trading at zero, therefore doesn't indicate that treasury yields incorporate *any* expected loss risk whatsoever. They don't.

[Nick Rowe]

Andy: hang on. I've just had a thought! If the Fed's portfolio is full of AA rated T-bills, and its liabilities consist of AAA rated cash, and it bought the T-bills at par, don't we have to worry (in a really good way, for AD) about the Fed's capital? If we really are in a ZLB liquidity trap, where there is no seigniorage revenue for the Fed, the Fed is now bust. The Real Bills Doctrine (the Equity Theory of Money) now starts to kick in. If all the Fed's assets are AA [edited to fix typo] and the face value of its total liabilities equal the total face value of its assets, then its liabilities must also be AA, not AAA. Cash must be AA too, which means the Fed *can* keep nominal interest rates at 0%. ???

Which ties back in with K's points.

[JW Mason]

Somewhat OT, but did anyone see the items in today's FT Alphaville about what seems to be deliberate use of settlement failures as a form of unconventional financing in secondary markets for AAA securities? Not directly related, but it is consistent with the intuition of this post that default risk is a natural alternative margin in debt contracts when nominal interest rates are stuck.

Also, an alternative response to Scott S.'s point is that there are significant costs to holding cash. (And the Fed can always tax reserves.) Of course that then begs the question of why we're so sure the Fed can't target a negative rate.

[JW Mason]

Also, there seems to be a consensus between Nick and all of his critics that increasing the perceived default risk of US bonds is at worst neutral with respect to AD and output. Is that correct?

[Scott Sumner]

Nick, If I'm not mistaken, the liquidity trap view says the expected short term return on all Treasury debt is zero. So if you keep the nominal rate on cash as zero, and add default risk to Treasuries, the bonds all disappear. So in your example there are no T-bonds in circulation as soon as default risk kicks in as everyone sells then to the Fed. And the Fed is forced to buy them, otherwise interest rates rise and we are no longer in a liquidity trap. See also my reply to Mason.

However, I do agree with your answer to Andy that an irresponsible fiscal authority can create expected inflation, because it puts the central bank in the position of being forced to monetize the debt. Now you seem to have suggested a different reason--it makes the central bank broke. So I see that as another aspect of the well-established view that extreme fiscal irresponsibility can create higher inflation expectations. I have no argument there.

JW Mason. Good point. I agree that there are lots of real world problems with my argument, and I agree those arguments beg the question of whether we are really in a liquidity trap at all. The argument cuts both ways, as you suggest. My way out is to be skeptical of the entire liquidity trap argument.

[Nick Rowe]

JW: "Also, there seems to be a consensus between Nick and all of his critics that increasing the perceived default risk of US bonds is at worst neutral with respect to AD and output. Is that correct?"

I think that's correct, unless you bring in some other effect that isn't in the basic model.

I hadn't seen that FT Alphaville before.

[K]

Scott: "If I'm not mistaken, the liquidity trap view says the expected short term return on all Treasury debt is zero. So if you keep the nominal rate on cash as zero, and add default risk to Treasuries, the bonds all disappear."

Reserves yield zero.  T-Bills yield the expected loss rate on T-Bills. So the expected return on T-Bills is the nominal yield minus the expected default loss rate, for a net of zero.  

"So in your example there are no T-bonds in circulation as soon as default risk kicks in as everyone sells then to the Fed."

No.  The expected return is zero, i.e. equal to the expected path of the short rate.

"And the Fed is forced to buy them, otherwise interest rates rise and we are no longer in a liquidity trap."

I think you just need to modify your definition of liquidity trap to "expected return of T-Bills is zero." The only relevant meaning of liquidity trap is that the Fed can no longer influence the market through its target for the interbank lending rate. So they need to target reserve quantity, or term yields.  I don't see how this changes anything.

[Matias Vernengo]

Yep. There was a whole debate about Ron Paul's proposal to destroy the debt held by the Fed in order to create space for additional debt. Here is a link to Dean Baker's discussion of the proposal http://www.tnr.com/article/politics/91224/ron-paul-debt-ceiling-federal-reserve

[Lars Christensen]

Nick, I like this. It is not necessarily ultra relevant in the present situation, but it is get fun and intellectually stimulating.

I have been playing with the same idea, but in a slightly different model set-up. Lets imagine we have a liquidity trap and/or we are trapped in some deflationary spiral combined with a Sargent-Wallace model of “some unpleasant monetarist arithmetic”. Then the policy recommendation would be to appoint Zimbabwean central bank governor Gideon Gono as Treasury Secretary. Hence, everybody in the financial markets would know that a completely irresponsible person with about no credibility would now be in charge of US fiscal policy and the logic of the monetarist arithmetic would mean that sooner or later that would lead to Zimbabwean scale money printing and inflation expectations would spike dramatically and immediately solve the liquidity trap problem and hence would be expansionary in the same way as you suggest.

Judging from Brad Delong’s comments this is maybe his model? So after all Brad it not a Keynesian, but a monetarist arithmetician;-)

[Jacques René Giguère]

A few minutes ago Krugman has commented on his NYT blog about this post.
I think he has the right argument.

[Lars Christensen]

I might add that the problem with the "Gono as Finance Minister"-model is that Japan seems to have pursued the policy recommendation of the model for years without having the expected impact...Maybe Barro can explain why the model has failed in Japan;-)

[Lee Kelly]

Does the Pope believe in his own dogmas? Yes, except when he doesn't.

"When the facts change, I change my mind. I then change them back again when nobody is watching. What do you do, sir?"

Didn't Keynes say something like this?

[K]

Scott weighs in with his own post:
http://www.themoneyillusion.com/?p=10216

And so does Krugman: "But there is still a “shadow” rate, the rate at which private investors would be willing to buy short-term US debt — and that rate can easily go well above zero.

This shadow rate, in turn, is — if I’m getting this right — the rate that feeds into the determination of longer-term rates. So we should expect rates to rise all along the term structure."

But I don't understand why he puts it like this.  Expected loss rate from default can hike treasury rates all the way out the curve.  But it's definitely not observations of short end expected loss rates (i.e. short term CDS rates) that determine long term rates. Long term expected rates of default, are driven  by long term expectations of default which cannot be extrapolated from observation of the short term dynamic.  There is no other way than to look at long term debt distribution and balance sheet dynamics. Anyways, for most firms *by far* it is impossible to observe short dated default risk - all we have is longer term bonds.

And how does the lack of observation of short term rates cause the long term rates to rise, if that is in fact what he is saying? I'm really puzzled.

[Nick Rowe]

This is getting exciting!

My mind is still stuck on the Fed's balance sheet. Normally I steer clear of the "backing theory", but if we really are at the ZLB, so cash is a perfect substitute for bonds, and there's no seigniorage, the Fed becomes merely a money market mutual fund, that has bought assets at $1, that are now, because of default risk, worth only $.99. So I don't see how that default risk can fail, in some sense, to spill over into cash. If the government defaults (and say pays only 99 cents on the dollar) on the Tbills the Fed owns, the Fed's balance sheet deteriorates year after year. Asymptotically, the Fed's assets approach zero, and it cannot withdraw cash from circulation in future even if it wants to.

[Lee Kelly]

Nick,

Yeah, that's where I got stuck at. All else being equal, further reducing the supply of safe and liquid assets would tend to increase money demand, but all else is not equal. If the Fed can't withdraw money from circulation, then the inflation risk on money rises. The question is: does the opportunity cost of holding money instead of U.S. government debt remain constant? If it does, then a default might be expansionary after all, because opportunity cost of holding money and U.S. government debt will have increased relative to everything else.

[Lee Kelly]

In other words, I think I have given a monetarist account of how default can be expansionary.

[Nicholas Grossman]

So far we have these potential problems:
-treasuries aren't cash (Krugman's objection)
-assuming the Fed can and will make all the necessary bond purchases may not be reasonable (DeLong and others here)

An additional one:
IS/LM doesn't consider multiple national economies in a globalized world. If expectations of US default increase, many holders of US treasuries will just shift to German bunds or other assets they see as safer. If that were impossible, then rising default expectations could arguably incentivize potential savers to spend more in the US. But in the internet age, such a transaction is virtually frictionless.

[Nick Rowe]

Lee. Yep, sounds about right.

A beautifully paradoxical way of asking the question: if the Fed is a money market mutual fund, and the default risk on T-bills rises, will the buck break the buck? If so, AD expands.

Gotta go now. Will probably not check back in before tomorrow.

[Jasiek]

Lee Kelly has made a good point. I’d suggest you should just try and read p.202-208 of Keynes' General Theory.

[Lee Kelly]

jasiek,

I don't have the book. I can access it online, but page numbers aren't much use. What part do you mean, specifically?

[K]

Nick: "So I don't see how that default risk can fail, in some sense, to spill over into cash. If the government defaults (and say pays only 99 cents on the dollar) on the Tbills the Fed owns, the Fed's balance sheet deteriorates year after year. Asymptotically, the Fed's assets approach zero, and it cannot withdraw cash from circulation in future even if it wants to."

Well, not year after year. Default only happens once. But...

Like Scott's coin seignorage post or Ron Paul's total balance sheet destruction plan, a sudden increase in default would be a loss in asset value for the Fed which would likely increase inflation expectations. Hard to say if the inflation effect on the dollar is exactly equal to the expected loss on the debt: As long as there is no default, the Fed is, after all, earning the 1% credit spread on the treasuries, so the loss is slowly coming back to them in the form of some kind of seignorage.

[Kaleberg]

I think it's briar patch thinking. If anyone with functioning brain cells advocates anything, the Republicans will oppose it on principle. If Obama and the Dems were in favor of a government downgrade and default, the Republicans would have to oppose it. The only way to get them to go for it is to beg, "Please don't throw me in that there downgrade briar patch", and hope that Uncle Remus was right.

[RSJ]

I think the problem boils down to the institutional structure, and it is very instructive to our current situation.

For example, Canada manages rates by the corridor system. It doesn't matter if the government defaults on bonds or not, a bank will still lend reserves to another bank at a price within the corridor, because it will lose money if it doesn't.

So bank cost of funds will not change, neither will the central bank need to go on a bond buying spree in order to keep rates low. Arbitrage will ensure that the marginal cost of reserves is what the CB dictates, and from the marginal cost of reserves, other rates will propagate.

But if banks hold government bonds as assets -- which they do -- then their balance sheets will change. They will suffer a huge loss of capital, so who pays? One option is for everyone to pay by paying higher (non-risk-free) borrowing rates. But that is not the best option. If the regulators swiftly close the banks and have the bank creditors eat the loss, then it need not result in higher rates on the margin. If history is a guide, regulators will not do that and they will prefer to socialize the losses. That means banks will charge higher rates to their customers and slowly recapitalize themselves, while keeping their own creditors whole. That would result in higher interest rates and lower output.

Therefore whether or not higher rates are the outcome depends on how the resulting balance sheet problems are handled. The central bank can always insure that the marginal cost of reserves is whatever they want, and they do not need to buy any bonds to do this -- it's enough to be willing to lend to banks at some rate, and to pay interest on reserves at some other rate.

And I think that this is what has actually happened. Rather than having mortgage losses borne by those who are supposed to bear them, the losses are being borne by the non-financial sector as a whole. This is equivalent to placing a tax on the non-financial sector, which lowers real output and raises real borrowing rates to a sufficient level to recapitalize the banks from the entire economy, instead of recapitalizing the banks from the existing pool of bank creditors.

[JKH]

ISLM doesn’t incorporate the term structure of government interest rates; nor does it incorporate central bank’s operational capability to modify that term structure through its own intervention. If the government retains the existing term structure on its debt, the market will price that debt according to perceived default risk. Other things equal, nominal interest rates will move up, adjusting for default risk. And the LM curve will shift. But if the central bank buys back all government debt by issuing interest paying reserves, it can control the nominal interest rate, other things equal. So the horizontal version of the LM curve doesn’t have to shift up.

There’s no default risk on “AAA cash” (bank reserves) in that case, because the fact that the central bank is now the main funding mechanism for government deficits removes the threat of operational default or “insolvency”, provided the central bank keeps the nominal rate paid on reserves low enough. That won’t be a problem in the early stages of AD stimulation.

The bonds that the Fed holds and the condition of the Fed’s balance sheet are not an issue for default risk. It is of no economic consequence to net fiscal math that the market would rate bonds held by the central bank AA. It is irrelevant for actual deficit financing. This is internal bookkeeping. Even if the government (treasury) chooses to “default” on these bonds, or to mark them to market for default risk (which it wouldn’t in any case), it is easy enough to recapitalize the central bank by exchanging a capital injection for an internal loan from the Fed to treasury. There is no net fiscal effect due to such internal bookkeeping, given the remittance of Fed profits to Treasury.

I think Krugman is right on the shadow rate as a concept, but he’s wrong on the interest rate implication. There’s no reason for the private debt market to price interest rates differently - whether the reference curve for risk is an actual treasury yield curve that is truly risk free, or a shadow treasury curve that includes default risk with higher interest rates to reflect that risk. Shadow rates for the risky shadow curve would increase, but rates on private debt wouldn’t change – because a contraction in the credit risk difference between private debt and treasury debt accommodates a commensurate contraction in the interest rate differential between them.

[JKH]

P.S. the exit strategy (if desired) for withdrawing cash (reserves) is done via either central bank liability management (e.g. term deposits) or by Treasury restarting its debt issuance (it pays down the internal loan, which extinguishes reserves). Central bank marketable assets are not required to withdraw cash.

[Account Deleted]

Just when I thought I had my head wrapped around IS/LM you throw me for a loop.

Isn't the assumption that a 1% increase in perceived risk on government bonds and a 1% increase in expected inflation will be the same rather questionable？ An change in in inflation expectations will instantly change all real interest rates the same amount. A change in the perceived risk on government bonds will only directly change the rates on treasuries and effect other interest rates in a complex hard to predict fashion.

[Kevin Donoghue]

Lee,
Jasiek's reference is to GT Ch 15, sections (ii) and (iii). But I don't see that it helps much; Nick's IS-LM model isn't closely related to that of Hicks, which in turn is some way from the GT. A discussion of this kind won't go anywhere until we specify the menu of assets more precisely. A typical IS-LM model doesn't distinguish private-sector bonds from government bonds.

[K]

Nick:  

Bottom line: The Fed will *never* default on money. A default on treasuries may debase the Fed's balance sheet which increases risk of inflation. This will devalue future money. But inflation will have exactly the same effect on the value of future treasuries since they are denominated in money. I.e. inflation can't change the relative value of money and treasuries (a change in nominal rates *can*, but not a change in inflation).  Therefore while credit risk will impact bonds, it will not impact money.  Real bills is false.

[Lee Kelly]

K,

The interest rate on treasuries rises to offset inflation. The interest rate on base money is fixed near zero by the Federal Reserve.

[babar]

any foreseeable 'default' at this point would just be a slight delay in coupon payments, and since interest rates are nearly zero, it would be a non-event from a theoretical point of view. this thread is about real default, which isn't within the realm of possibility currently.

[Scott Sumner]

I just noticed that Krugman gave the same answer as I did, but he didn't cite me. :(

[K]

Lee: Here's another way to think of it.  Money is a perpetual zero coupon bond.  The only way it can have value is for it to have some kind of convenience yield. For $1 (the medium of exchange) to have *perfectly constant* value (in terms of the medium of account)  that convenience yield must be exactly equal to the short rate. I.e. currency is equivalent to a perpetual floating rate note with a convenience yield equal to the FF rate. There is no other coherent way to define it. So from the point of view of interest rates or inflation, holding $1 of money is exactly equivalent to holding a portfolio in which you roll $1 of T-Bill and consume the interest. Neither a change in rates, or a change in inflation can change the relative nominal or the relative real value of those two portfolios. But, if the T-Bill defaults, the money is still there. Where's it going to go? If you insist on real bills then the government then taxes back some assets and gives them to the Fed to "back" the money. But it doesn't matter. The Fed's money doesn't need backing because there is no mechanism for it to default since it doesn't mature and doesn't pay interest. Since it is the *numeraire* it merely measures the value of everything. Defaultable T-Bills do not.

[K]

All of which is to say, Krugman is right. Inflation and government credit risk are quite different. Money, inflation, and the term structure of risk free rates exist completely independently of government bonds.  Government bonds are just something the CB may or may not have on balance sheet. The Fed, in fact, holds lots of bonds of non-government issuers of a variety of credit qualities. And if the Fed is being well run, it should be well diversified and have taken sufficient hair cuts against risky assets to assure that it will always have enough assets to prevent inflation (but even if it hasn't, it doesn't matter - see my previous comment). Oh yeah, and I take back what I said to the contrary yesterday.  It was some muddled nonsense.

[ezra abrams]

but are the assumptions accurate ?
If the assumptions break down, then speculation on the model is just navel gazing, not science
so, before you waste a lot of time and mental energy, supply the data that shows the assumptions are accurate enough that the model has some validty to the real world.
I'm guessing you don't have the data, which is what separates economics from science (and even if you had the data, it would be difficult to assert that it applys to the future)

[JW Mason]

A lot of the pushback on this hinges on the option to sell bonds for cash. But why hasn't this option been exercised already? Bonds are necessarily riskier than cash - there's always a nonzero chance that rates could rise before the bond matures, producing a capital loss. Bonds are less liquid than cash, more or less by definition. So why is anyone holding a zero-yield bond rather than cash? And if we don't have a good answer to that question, how can we be confident that they wouldn't continue to hold them even with some default risk?

[Phil Koop]

"So why is anyone holding a zero-yield bond rather than cash?"

Is that a serious question? I can answer it, but I don't think that will cast any light on the ISLM model or on the beliefs of Keynsians.

If you want to hoard a considerable sum of "cash", 100mm or more, it is impractical to do so in the form of pieces of paper. First, because once transaction and storage costs are considered, paper cash earns a negative rate of nominal interest. Second, because the "default risk" of paper cash is positive: despite all your expensive precautions, your vault could be destroyed by earthquake or fire, or be robbed. Third, because although any individual agent might be able to find 100mm or more in paper, there is not sufficient paper to satisfy the aggregate demand of all agents.

A private person might have a computer program break up his 100mm into 400 or so deposits at FDIC-insured institutions, but most large sums are held by institutions (even when the beneficial owners are in fact individuals.) These are not eligible for FDIC insurance. Consequently, even "cash" in the form a demand deposit has default risk in the ordinary sense; on a risk-adjusted basis a demand deposit earning 0% is normally worth less than a T-bill earning 0%. Banks are able to park their cash as reserves, but most institutions do not have that option either and use bonds directly (by buying them) or indirectly (by lending repo and accepting bonds as collateral.)

This use of bonds is one reason why DeLong's sanguine view that the market will just "switch to cash collateral" is unworkable: you can't usefully collateralize cash with cash. (There are other reasons, but they are even more tangential to the subject of the original post.)

[JW Mason]

If you want to hoard a considerable sum of "cash", 100mm or more, it is impractical to do so in the form of pieces of paper.

OK, but this creates its own problems. First of all, for purposes of these discussions, reserves are cash too. Second, it invites the question I asked above, if there are significant costs to holding cash rather than bonds, why are we so confident the Fed can't target a negative overnight rate?

You might then say, the Fed could target a negative rate, but for some reason it won't. In which case a positive default risk becomes attractive as an alternative, and we're agreeing with the argument of the original post. Right?

[Determinant]

I post this to be corrected, but it was my understanding that sovereign borrowers by definition had the highest credit rating in their own currency. They have tax powers, issue the currency and deal with the central bank; private corporations don't. I understand that when a sovereign borrower is downgraded, it takes everyone else in that currency down too. Private borrower credit ratings are not independent of the sovereign rating, which many on this thread seem to assume.

If the US is downgraded this can easily be empirically tested. Warren Buffett and Berkshire Hathaway make a big deal out of Berkshire's AAA credit rating. Buffett uses Berkshire's large balance sheet and creditworthiness to engage in large reinsurance deals. He explicitly uses Berkshire's credit rating as a selling point to do business. So if the US is downgraded Berkshire's rating can be watched to see if it remains AAA or goes down because the US went down.

[K]

Determinant: in principle you should be right since no sovereign issuer will ever really default except in a condition of hyperinflation, which will cause skyrocketing rates which be equally distressing for all borrowers. In practice, though, they will downgrade the US and not corporates, on the principle that even delaying an inevitable interest payment is unacceptable for a AAA issuer, and does not reflect badly on anyone else.

[Phil Koop]

I have already dealt with reserves as cash. The Fed can target any fed funds rate it likes, having the ability to make an unlimited market at its own 2-way price. It can even make the actual traded market rate a small negative one, when the conditions are right. Large negative rates, though, will merely prevent the market from clearing. They can have no bearing on the ISLM model because they have no connection to LM. Krugman is in the right of that: you can make any hypothesis you like on the empty set, but it would be tendentious to assume that your hypothesis is thereby supported.

An aside: it is conventional for you economists to treat small rates, both positive and negative, as equivalent to zero. Is that wise? That is a question you must answer for yourself.

[Phil Koop]

@Determinant, in addition to what K said, another practical example of a sovereign defaulting on its own currency obligations was Russia in 1998. They repudiated their short-term ruble debt and yet honoured their longer-term USD debt. This is the opposite of conventional wisdom, that a sovereign can always print its own currency, but must buy FX with real goods. But the discount rate on the GKOs had become ruinous, and once default became the least painful option, they defaulted.

[Determinant]

Phil, I wasn't saying that sovereigns were perfect credits by definition, my point is that they are the highest (though not perfect) credit by definition in their own currency. The US Government, 12 states or so and a dozen US corporations have AAA credit ratings. If the US gets downgraded to AA, what happens to the other entities credit ratings? Do they go down too, as the theory on credit ratings I have read leads me to believe. If so that would imply a good deal of suffering. Not to mention every other credit, mostly though those BBB or above. Credit ratings are partly a relative definition with the point of reference being the sovereign currency issuer of the debt in question.

[Doc Merlin]

"I'm trying to figure out if this reasoning is also consistent with the (casual) observation that poorer people buy lottery tickets, thereby voting for redistribution that increases inequality." - Nick Rowe on in a comment on the econlog blog

Hey Nick, I know this is off topic, but I got banned from econlog for saying something disparaging about deLong's attitude towards other bloggers.
Poor people have a very, very high effective marginal tax rate due to stacking benefit losses as income increases (it can get over 100%). This can by standard microeconomic models of risk preference make them "risk loving."

[Nick Rowe]

Doc: "Poor people have a very, very high effective marginal tax rate due to stacking benefit losses as income increases (it can get over 100%)."
Agreed.

"This can by standard microeconomic models of risk preference make them "risk loving." "

I'm still trying to get my head around this. I think I can see it for taking risks over *taxable* income. Does it also work if lottery winnings are untaxed (like in Canada?), and you can't write off lottery losses against taxable income??

This might be an important insight.

[Nick Nolan]

Krugman's answer:

http://krugman.blogs.nytimes.com/2011/07/25/default-in-a-liquidity-trap-very-wonkish/

PK:
..I think this is wrong — but in an interesting way.

It’s true that, say, a 1 percent possibility that your bond holdings will become worthless within a year is similar to the expectation that inflation will erode those bonds’ real value by 1 percent over the next year. But inflation doesn’t just erode the value of bonds; it also erodes the value of cash. And that’s why expected inflation can help in a liquidity trap: it makes sitting on cash less attractive. The threat of default doesn’t do that. As far as I know, we’re not talking about a loss of confidence in pieces of paper bearing pictures of dead presidents. And that’s why the threat of default isn’t equivalent — and not expansionary.
..