Scott Sumner  

Missing the big picture

My Simplistic Theory of Left a... A snapshot of Portugal...

I've done scattered posts discussing the flaws in alternative approaches to monetary economics. Here I'd like to try to show some underlying themes in these critiques. I'll briefly discuss 4 common myths:

1. The trade view
2. The fiscal view
3. The banking/finance view
4. The Cantillon effects view

In my view all four of these approaches miss the big picture, because they don't focus on how changes in the supply and demand for base money drive NGDP through the hot potato effect. At the same time I understand why people are drawn to these fallacies, as the hot potato effect is really hard for most people to understand, and indeed even many economists don't really get it.

1. Recently I argued that the Japanese could raise their NGDP by 20% if they depreciated the yen. One commenter discussed the policy from a trade perspective---looking at exports and imports as a share of NGDP, trade elasticities, etc. But these miss the big picture, and confuse a change in the real exchange rate reflecting non-monetary forces, with a change in the nominal exchange rate engineered by the central bank.

When the central bank changes the nominal exchange rate from its previously expected path, it also changes the long run expected path of NGDP and the price level by an exactly equal amount. This is because money is neutral in the long run and hence monetary policy has no long run effect on the real exchange rate. If you depreciate the nominal exchange rate by 20%, it will cause the price level to rise proportionately in the long run, as the real exchange rate will be unaffected. Thus in the long run, any currency depreciation caused by central banks affects the price of domestically produced goods by exactly as much as the price of imported goods. In contrast, a currency depreciation caused by more domestic saving can affect the real exchange rate in the long run.

2. Another common mistake is to assume that monetary policy becomes much more important if it is used to "monetize the debt." If you go back to the period before 2008, the base was about 6% of GDP. Thus a $1 increase in the monetary base would raise NGDP by about $16. Most estimates suggest that an extra dollar of government spending would only raise NGDP by $1 or $2. (Even assuming no monetary offset.) That's not to say that fiscal policy can't have any effect, but the total effect on aggregate demand will be almost the same, regardless of whether the government does or does not accompany a monetary injection with an equal fiscal injection. Most people put way too much weight on things like helicopter drops. It's monetary policy that determines NGDP; helicopter drops don't solve any fundamental problems associated with the zero bound--such as Krugman's "expectations trap." That's why Japan's NGDP kept falling despite big increases in Japan's budget deficit, financed by printing money.

3. Another common mistake is to assume banks play an important part in the transmission of monetary policy. That's wrong for several reasons. First, far less of the new money goes into banks than most people assume (except when rates are zero.) Before 2008, more than 90% of new money went into cash held by the public, and even that figure was artificially held down by required reserves, which was just a tax on banking and in no way central to the process. The excess reserves were less than 1% of the base. If reserve requirements had been abolished, then reserves would have been a tiny portion of the base, while cash would have been 95% to 98%. Even if the entire banking system did not exist, the Fed would have conducted monetary policy in much the same way. They would buy bonds from bond dealers, and pay for them with currency. That's actually pretty much how they ran things prior to 2008, except that the new money would spend a few days as bank reserves, before going out and circulating as cash. They could target NGDP just as effectively with that system, as with the current system.

Nor does the fact that banks hold bank reserves make them special. Reserves are just base money. Drug dealers also hold lots of base money, but it doesn't mean they play an important role in the monetary policy transmission mechanism. If drug dealers demand more money, the Fed will usually accommodate that demand to keep prices stable. Ditto for if banks suddenly demand more reserves.

The best argument for banks being special is that their demand for reserves is highly elastic at the zero bound. So the share of the base held by banks soars at this point. But this fact has nothing to do with bank deposits and loans, which are the mechanisms by which most people think banks are important. Another argument is that banks are important because bank deposits are a medium of exchange. That's true, but it only impacts the price level to the extent that it impacts real demand for base money. If creating a banking system caused the demand for base money to fall by 20%, then it would cause the price level to rise by that amount. Except that the Fed would almost certainly offset the effect with a reduction in the base. Similarly, if drug legalization caused the demand for base money (cash) to fall by 20%, then it would cause the price level to rise by that amount. Except that the Fed would almost certainly offset the effect with a reduction in the base.

Of course banks are very important and provide valuable services to the economy, as do electric power companies. It's hard to imagine life without banks or electric power companies. But those are supply side benefits, and as we've seen in Zimbabwe a central bank can dramatically boost NGDP even if the supply side is in horrible shape. Thus if you are looking at factors that impact NGDP growth, you need to focus on the supply and demand for base money. And since central banks have virtually unlimited ability to change the base supply in such a way as to offset shifts in base money demand, you want to focus on central banks, not commercial banks. Don't confuse money and credit.

Note, I'm not saying that monetary stimulus can't indirectly cause more (real) lending to occur. It can, just as it can indirectly cause more electric power plants to be built. But those are secondary effects resulting from a combination of rising NGDP and sticky wages.

So why do central bankers think banking is so important? Because they use a flawed targeting procedure that accidentally makes banking more important. Because real world central banks target nominal interest rates rather than NGDP futures prices, a banking panic can lead to a tighter monetary policy. But fundamentally the central bank is causing the tighter policy; it just doesn't understand that fact.

4. The same sort of mistake is made when people focus on Cantillon effects. When the Fed injects new money it generally buys bonds. The effect of the extra money (if permanent) is profound, whereas the effect of the bond purchases is trivial. If they used a different procedure such as paying public employee wages with the new money, the only difference would be that bond dealers would lose out on a bit of revenue from selling bonds to the Fed. But the commissions are so infinitesimal in the (highly liquid) bond market that the secondary effect is trivial compared to the direct effect of a permanent increase in the monetary base. It doesn't matter "who gets the money first." Cantillon effects only become important if the central bank buys something wasteful like bananas, which quickly rot.

Does any of this change at the zero bound? Not much. The bond commissions become bigger because more bonds are purchased. Maybe the Fed buys MBSs issued by GSEs, but even these are backed by the Treasury. They are risk free for the Fed. The Fed now pays interest on reserves, which makes the newly injected reserves much like government debt. QE is essentially swapping one form of debt for another. And the addition of interest on reserves means the Fed now has two tools to impact base demand (IOR and reserve requirements) but nothing fundamental has changed.

The base is called "high powered money" because permanent increases in the base have a massive impact on the nominal economy. All the other issues (trade, fiscal, banking/finance and Cantillon effects) are minor sideshows by comparison. Keep your eye on the big picture when analyzing NGDP.

And when analyzing living standards in the long run, even NGDP becomes a minor footnote. In that case it's supply-side factors that matter. If you want to focus on the impact of banking or fiscal policy, you'd be much better off ignoring their interaction with monetary policy, and looking at their effect on productivity--the supply-side forces that matter in the long run.

PS. There are of course many other fallacies, such as the idea that deflation is caused by bad demographics, or overproduction, or international trade, or less lending, or any number of other non-monetary factors. I'll address those in a future post.

Comments and Sharing

COMMENTS (27 to date)
Kevin Erdmann writes:

In a regulatory regime that leads to much lower mortgage levels (higher equity proportions), how do you see that affecting monetary policy? Pre-HUD banks look like they held more treasuries. Was this related to the low mortgage levels of the time? Or was this due to other factors? I wonder if, lacking several trillions of dollars of mortgages that would have been issued if we had a stable regulatory regime, the banks would revert to holding treasuries again, and that excess reserves are basically a substitute for that. I assume that more equity ownership might lead to demand for deposits on the banks' liability side but not for credit on their asset side.

Am I out of my element here?

Market Fiscalist writes:

I very much liked the article.

I have a question on:

'It's monetary policy that determines NGDP; helicopter drops don't solve any fundamental problems associated with the zero bound--such as Krugman's "expectations trap." That's why Japan's NGDP kept falling despite big increases in Japan's budget deficit, financed by printing money.'

I'm not getting why the newly printed money that financed the deficit did not create a HPE and drive up NGDP. Was it becasue the demand to hold money was so high that all the new money was held? I assume that sufficiently large budget deficits funded by printing money would drive up NGDP eventually.

Effem writes:

Good post. However, I think you are too quick to bypass political ramifications. Sure, the Fed could theoretically perform only helicopter drops to Republicans as one way of expanding base money. And if money is neutral it wouldn't matter. However, the Fed would also lose its independence in a nanosecond.

The Fed would be wise to distance itself from the banks in my opinion.

D. F. Linton writes:

You have debunked Cantillon effects in the same way that Keynes debunked Say's law: Call something completely different by the same name and debunk that.

Unless you believe that monetary manipulations equilibirate in zero time or that path effects are completely irrelevant, those who get the new money first must benefit. Cantillon effects don't imply that M*V can't be manipulated, but rather that the economy that exists after a money injection is not the same economy that existed before but with rescaled prices.

Vaidas Urba writes:

Scott, a bit of nitpicking:

1. Suppose cash did not exist, and only the banks could hold the monetary base. Well, nothing substantial would change in this case. Your cash story is an interesting teaching device, but as monetary policy in cashless economy would have the same effect, maybe cash is a bit of a distraction?

2. Sticky wages are not the whole story. Sticky loan terms could be an important part of the picture too. More real lending can be a primary effect of monetary stimulus.

bill writes:

Questions about helicopter drops.
If it's simply the Fed buying T Bonds with new cash, why isn't it just called open market operations?
And what would it be called if the Fed printed up some money and just threw it out the window (or dropped from a helicopter, or just mailed every American ten hundred dollar bills)?

Doesn't IOR paid to banks reduce the HPE? In that sense, that it's only banks that can earn IOR (the Fed won't pay me 25 bps on cash I give it), doesn't the banking sector matter?

CMA writes:

"The effect of the extra money (if permanent) is profound, whereas the effect of the bond purchases is trivial."

The effect of bond purchases is not trivial IMO. OMO's affect interest rates. Injections of MB without purchasing bonds through emoney heli drops expand MB while generating less downward force on rates because demand for bonds doesn't increase and because loanable funds don't increase as funds enter accounts hosted by fed which commercial banks cant access.

Stimulating at higher rates will result in less finsector and debt stimulus which will positively affect RGDP and stability.

Lorenzo from Oz writes:

Nicely clarifying, thank you.

On monetary fallacies, it is amazing how many discussions of the Price Revolution (from the late C15th to early C17th) try to blame non-monetary things -- such as population growth.

The fivefold increase in Central European silver mining (from new technology such as water pumps and lead-silver smelting) followed by the looting of the Aztec and Incan empires followed by the Potosi silver mountain obviously drove the upward increase in European prices, given that silver was the dominant monetary metal. But folk often don't know about the Central European silver surge, so point to the price rises beginning before American silver and say "more silver cannot be the cause".

Slightly more sophisticated analysis will talk about rises in demand to hold silver counteracting increases in the supply of silver (apparently, silver has a very weird demand curve). Or expansion in the use of credit instruments (which led to less demand to hold silver, but that would not have had much effect if the supply of silver wasn't continually increasing faster than output). And so on.

The supply of silver (the main monetary metal) increased faster than output. Continually. For almost two centuries, Of course prices rose. How hard is that?

Lorenzo from Oz writes:

I must not be fully awake. Rise in use of credit instruments would have had no systematic effect on prices, because they were denominated in silver.

James writes:


You assert that it does not matter who gets the money first but give no reason to believe this. Tomorrow, the fed could increase the money supply buy purchasing only bonds currently held by people living in New York, or by purchasing only bonds held by people living outside of New York. Do you seriously claim that it wouldn't matter which group of people the Fed bought bonds from?

Your understanding of open market operations seems to miss the big picture. If the open market desk were to buy bonds in quantities large enough to affect NGDP, that means some investor somewhere has to increase his cash position and reduce his bond position. If asset markets are efficient then that investor is at his target asset allocation to begin with so the open market desk has to pay a premium to get that guy to deviate from his target allocation.

You also miss the small picture. Even if everything else works as you claim, the hot potato effect only increases welfare if it induces people to engage in transactions with gains from trade. If people aren't willing to spend initially, there might be a policy that induces them to increase their purchases but that doesn't mean the average person is better off. If you believe that there are potential transactions with gains from trade but people just aren't completing those transactions, you haven't given your reasons to believe this.

A writes:

James, I think that you might be referring to a premium paid to account for large transactions. The argument you use could describe any transaction, and would imply that a cash for asset exchange does not occur in an efficient market. But there is such a thing as an ask, and size-dependent ask increases are usually attributed to liquidity premiums. If there is a Cantillon Effect from the process you describe, it flows through prime dealer profit, rather than the order of receiving fungible cash.

The Cantillon Effect would be most likely if OMOs were large and secret, so that intermediaries possess an exploitable information advantage.

Jose Romeu Robazzi writes:

@Prof. Sumner
This was very clarifying, thank you. Now I think I understand your reasoning. In your model, growth in credit is important to the extent that there was demand for money following NGDP growth. And when banks react to easing by the FED in effect what they are doing is increasing base money held by the public, to counter potential fall in money velocity that falling longer rates have indicated. If for some reason that transmission mechanism fails, the monetary authority should just go and directly purchase assets, increasing base money held by the public. If I had to teach this concept, I would say that money velocity is a "mains street" concept, and in general businesses and households react to rising interest rates by increasing money velocity, and react to lower rates by decreasing money velocity. As for the banks, the monetary authority should look at whatever they are doing considering their effect on the base money aggregate. They could amplify or detract from what the monetary authority is doing, but eiter way, the monetary authority can always do something else to achieve its (nominal) goals. Oh, and of course, the "thermometer" is always NGDP growth, not a certain monetary aggregate, or the level of interest rates.

ThomasH writes:


Quite off the actual point of your "big picture" rpost, but I think you might pay some attention to another kind of big picture, the discussion going on about why the Fed seems so hot to raise short term interest rates when the price level is still below what would be implied by a 2% inflation target.

It seemes that your political theory is that the Fed is constrained by an inflation rate ceiling because higher inflation rates are unpopular (people think that inflation = lower real wages). NGDP gets around this by letting the Fed target something that IS popular, income.

[NB: I am talking abut the political economy of ngdp targting, not it's actual benefits vis a vis so-called inflation targeting.]

While your theory is quite plausible, I think you ought to address the Krugman theory (not that the two are necessarily in contradiction), that low short term nominal interest rates are bad for banks* and the Fed is too cosy with bankers in part because bankers are actually on the Fed board and partly for sociological reasons.

This is also Gourevits's point, that the Fed is heavily constrained politically.

In your theory, ngdp targeting is a easy work-around to "populist" opposition to monetary stimulus that results in inflation. In Krugman's theory, the opposition to ngdp targeting would be much harder to overcome.

* DeLong thinks bankers my be mistaken but that's irrelevant to the political economy of their position.

Scott Sumner writes:

Kevin, If that slows real growth, and if the Fed targets inflation, then the effect is to make policy a bit tighter.

Market Fiscalist, It's because the increases were viewed as temporary, and that's because the BOJ chose a inflation target, not a price level target.

DF, You said:

"those who get the new money first must benefit."

No they don't have to benefit. And if they do the benefit is usually so trivial that it's not worth considering, as I explained in the post.

Vaidas, Or if banks didn't exist then retailers might be important, as they hold lots of cash in their registers. My point is that prior to 2008, banks held relatively little cash and most of it was locked up in required reserves due to a regulatory quirk.

Is your second point something that affects NGDP or RGDP?

Bill, Helicopter drop refers to a combined monetary fiscal stimulus. A simple OMO doesn't affect fiscal policy.

Yes, IOR can be important. This post is more about the pre-2008 system, when banking was widely viewed as being important to the monetary policy transmission mechanism, but in my view was not. And even IOR matters because it affects the demand for base money, not because it affects bank lending.

Lorenzo, Interesting. Who are the people who deny that more silver caused the inflation? Are they historians or economists?

James, You said:

"You assert that it does not matter who gets the money first but give no reason to believe this."

Actually I do give a reason, please reread my post.

I also said one can imagine bizarre setups for injecting money where Cantillon effects do matter--and I mentioned bananas.

Suppose the people who get the money first are public employees, who are paid their wages in cash instead of by check. Explain to me how they would benefit. I think it would be a negative, they'd have to walk to the bank once a month, instead of having their check electronically deposited.

You said:

"If you believe that there are potential transactions with gains from trade but people just aren't completing those transactions, you haven't given your reasons to believe this."

Yes I gave a reason, sticky wages.

ThomasH, I have responded to that in various comment sections, and one post, and basically don't think the argument fits the facts. No one can explain why the political power of bankers suddenly soared after 2009, a point in time when they were intensely hated, even by many people at the Fed. Bernanke was extremely upset with the banks. So how do they suddenly have the power to implement a tight money policy?

Why didn't they have that power when inflation was above 2% under the Bush administration? Why didn't hawks complain about easy money under Bush? (I.e. as much as they now complain, when inflation is below 2%.)

Sorry, but I find these conspiracy theories to be totally unconvincing. I get them in my email box quite often.

Kevin Erdmann writes:

It is bizarre to me what a counter-indicator public consensus is. The period with the lowest nominal growth and lowest inflation since the Great Depression is when everyone decides the Fed is engaged in profligate expansion.

The period where the Fed is explicitly uninterested in a 25% collapse in nominal value of the most important piece of collateral on bank balance sheets and banks enter a prolonged period of failures and defaults is when everyone decides the banks are part of a powerful financial conspiracy.

People are nuts. Cantillon Effects are based on the price of securities being pushed above the market price, right? But, bond pries fell during the QEs. Given that people are nuts, it follows that Cantillon Effects would naturally become a concern during the time when the liquidity effect is not dominant.

It's also natural that while everyone is worried about the profligate Fed that is looking out for the banks the Fed is talking about overheating while inflation expectations collapse and real estate lending enters its 8th year of depression level growth rates.

The only thing that would make this all more understandable is if the presidential candidates included a smiling cat, a guy with a big hat, and a rabbit.

James writes:


Re Cantillon effects, I asked you first and gave an example deliberately similar to current practice. You ignore my question (why?) but I'm polite enough to answer you:

To get Cantillon effects requires increasing the money supply. If federal employees are only paid their standard salary by newly printed cash and if the treasury keeps whatever bank balances (from previous taxing or borrowing) would have been used to pay them under more normal circumstances, then there is no monetary expansion yet. When the treasury spends those tax receipts or borrowed funds, they are spending more than they would have in the ceteris paribus world. The recipients of those funds benefit because this is spending that was not going to happen otherwise.

Seriously, why do you think anyone takes the other side of the trade when the central bank buys assets? Is it because they all discover they need to hold more cash and less bonds to reach their target allocations?

Sticky wages do not imply people are foregoing transactions with gains from trade. For all you and I know wage stickiness may be optimal in the labor market. Even if that's not the case, it doesn't follow that sticky wages indicate foregone gains from trade in any other market.

bill writes:

Upon further research:
"helicopter drop" does not equal helicopter drop

Not being a smart aleck.

Bob Murphy writes:


Suppose next Tuesday the Fed sells $2.4 trillion worth of Treasuries, and says they will only roll over the remaining amount for the rest of time; they will never accumulate more Treasuries.

They don't extinguish those dollars, however, but instead send a check for $8,000 to every man, woman, and child in the United States.

They further explain that they are still committed to their dual mandate; it's just that they are switching from buying Treasuries to instead doing helicopter drops as a way of hitting their targets.

Are you saying the yields on Treasuries would only move a few basis points because of this policy shift?

Dikran Karagueuzian writes:


This was an excellent post, because it explicitly acknowledged that the effects described are real, and noted that we have data which show that they are small.

Particularly good: "demand for reserves is highly elastic at the zero bound. So the share of the base held by banks soars at this point". I would have liked to see an analysis of what would have to be true here to make the banking concern valid. No matter how ridiculous the scenario.

This post, and the responses, are probably frustrating for you. But I hope you can carry on.

Lorenzo from Oz writes:

Mostly historians.

Sadly, not only historians -- see this useful review article.

D. McCloskey also tackled the issue in a 1972 book review.

"PS. There are of course many other fallacies, such as the idea that deflation is caused by bad demographics"

There is a non-fallacious version of this which is gerontocracy: older people are more anti-inflation than the young because of a mix of monetary illusion and self-interest. So, if they have more political power, expect lower inflation or even deflation.

Matt writes:

Since I think I might be the "one commenter" being discussed here, I should probably weigh in. : )

First, I think that there's an asymmetry here between myth #1 and myths #2-#4. There's a common thread running through the latter set of myths: they’re all qualitatively possible but quantitatively irrelevant - so that the persistence of these myths is in large part a consequence of innumeracy. (This is distinct from myth #1, which I’d argue is only partly a myth. But more on that later!)

Take myth #2. As you point out, when the base was 6% of GDP, a $1 increase in the monetary base would increase NGDP by about $16 (assuming constant V), whereas a $1 increase in spending would increase NGDP by $1 or $2 at the very most. So if we create a dollar and use it to finance government spending, the “monetary” impact is much larger than the “fiscal” impact.

But it’s even starker than this, because the base growth (and effect on NGDP) is, at least until we actively decide to reverse it, permanent; while the increase in spending is only temporary. So the monetary impact of this policy is multiple orders of magnitude larger than the fiscal impact - and it’s crazy to pay much attention to the latter.

Or take myth #3. People often think that the banking sector is important in the transmission of monetary policy because it faces reserve requirements, and looser monetary policy makes these requirements less binding. Sure. But quantitatively, this is almost irrelevant. Prior to 2008, required reserves accounted for less than 0.5% of the balance sheets of depository institutions in the US. Suppose that a bank raises its marginal funds in proportion to the mix of its existing liabilities. A 1% increase in interest rates will directly push up the cost of marginal funds by around 1%; the additional effect working through the opportunity cost of required reserves, meanwhile, is less than 1%*0.5% = 0.005%, a trivial addition. Required reserves are tiny and not a major cost for the banking sector.

Or people think that the banking sector is important because it holds excess reserves after the Fed engages in QE. But it’s not clear why the world we live in is really so different from a hypothetical world where non-banks could hold demand deposits at the Fed too. (In fact, as the Fed uses reverse repo to implement interest rate hikes, that world will become real.) There would be no difference if banks could act as pass-through entities, frictionlessly offering large-scale deposits backed by reserves and thereby allowing non-banks to indirectly deposit funds with the Fed. And while this isn’t quite true, it is close to being true; banks do charge a spread, but the short-term money market rate for non-banks isn’t very far below the IOR rate for banks. We’re talking about a handful of basis points.

Finally, take myth #4. This one is the worst. Suppose the Fed increases the monetary base by $1 by buying $1 worth of assets. The commissions accruing to whoever trades with the Fed is going to be a very small fraction of this - maybe 0.1% of the asset’s value, or $0.001. These are liquid markets, after all. Meanwhile, that $1 increase in monetary base increases NGDP by over $10 each year. Over 10 years, that’s $100 of NGDP. So the gap between the profits for the Fed’s counterparties and the effect on NGDP is a factor of 100,000! Trivial indeed.

Again, the common thread in myths #2-#4 is their innumeracy. They’re not completely spurious, but the effects in question are so tiny as to be effectively irrelevant compared with the dominant effect of monetary policy itself.

Matt writes:

I’d contrast the above with “myth” #1, which is rather different - being more of a conceptual point than a quantitative one.

You argue that if a central bank engineers a purely monetary exchange rate depreciation, then the future path of NGDP and the price level must rise by an equal amount. This isn’t quite true, at least unless the depreciation is permanent; a non-permanent depreciation could reflect a lower expected path for interest rates. (Dornbusch’s “overshooting” captures this effect: when prices are sticky, a 10% permanent shock to money causes a 10% permanent depreciation, and then an additional temporary depreciation reflecting the decline in nominal interest rates during the transition to the new price level, while the real money supply is elevated.)

But let’s grant that it is true, and that the exchange rate depreciation reflects a long-run rise in the path of NGDP and the price level. There’s still the question: how will this affect NGDP today? The answer in the simplest version of the New Keynesian model, with only nondurable consumption and no international trade, would be “intertemporal substitution”: you consume more today in anticipation of some combination of higher prices and consumption in the future.

But in reality, we might not think that intertemporal substitution is strong enough. (For one thing, maybe consumers aren’t that forward-looking.) What other channels are there? One, investment; two, international trade. They both presumably play important roles, but the former can be hard to quantify - how exactly does investment today respond to an increase in NGDP in the indefinite future? International trade is simpler, since in the short run the real exchange rate has weakened substantially, and we can use elasticity estimates to calculate the implied increase in net export demand.

My main point with Japan is that this implied demand increase is so strong - if you believe that the trade literature is even within the right ballpark on elasticities - that it’ll be enough to eliminate any remaining output gap, put upward pressure on price level, and thereby achieve the desired convergence to the long-run NGDP target.

This isn’t to say that the main short-term response will come through trade. Your point here is well-taken: a change in monetary policy sufficient to produce a substantial depreciation in exchange rates must necessarily involve some mix of a decline in expected interest rates and an increase in the long-run price level, and these changes will elicit responses from consumption and investment too, not just net exports. It would be wrong to ignore these other channels.

Still, I suspect that net exports in practice account for a sizable share of the aggregate demand effect; and they provide a floor for the plausible size of this effect. So they’re an important piece of the puzzle, and it makes sense to examine their role in detail.

(Again, this is in contrast to myths #2-#4, where the purported mechanisms are so tiny as to be practically irrelevant -- mythical indeed.)

James writes:

It seems like there is some confusion as to what the world would look like if the Cantillon effects view were correct.

In a world with Cantillon effects, investors would pay close attention to central banks and try to forecast their moves, believing there is a financial advantage to portfolios holding assets similar to those purchased in open market operations. Whether or not this is how our world looks is an empirical question.

Do investors engage in Fed watching?

Do portfolios with short term government bonds benefit when the Fed cuts rates?

Matt writes:

Bob Murphy: "Are you saying the yields on Treasuries would only move a few basis points because of this policy shift?"

Ignoring the signaling impact, they might move 10 basis points or so. The helicopter drop would effectively be adding $2.4 trillion to the government's consolidated debt, and that would be almost a 20% increase over current debt held by the public. In modern experience, Treasury yields haven't been very sensitive to the overall quantity of Treasuries outstanding, but such a large change might cause several basis points of movement. Although the $2.4 trillion increase would diminish the federal government's creditworthiness, if it was a one-off event the risk of default and therefore the risk premium would still be very low. 10 basis points seems the right order of magnitude.

Of course, if this was an actual event we couldn't ignore the signaling impact. I would have to massively update my beliefs about the world if the government ran up the debt by $2.4 trillion in a sudden orgy of rebates to households. I probably wouldn't trust the government's future fiscal rectitude, not to mention its commitment to low inflation and all other traditionally serious-minded goals. Given that, the effect of such a policy on Treasury yields would probably be enormous.

But that's a sense in which I think your thought experiment is a little unfair: one has to distinguish between (1) the direct effects of the policy and (2) all the other consequences of living in the kind of world where the government would do something so sudden and unprecedented. (1) wouldn't be that bad, but (2) would be crazy.

Now, here is the sense in which Scott's point is still quite valid. The effects of

(A) a $2.4 trillion increase in the debt of the consolidated federal government

would be very small compared to the effects of

(B) a permanent $2.4 trillion increase in the expected path of the monetary base (assuming that the Fed wouldn't simply revert to IOR or some other method of implementing monetary policy independent of the size of the base).

The results of (A) would be nothing special - we've seen fluctuations of that magnitude in debt/GDP in the postwar era and have managed them rather smoothly. But (B) would more than double the expected level of NGDP a decade or two hence, which would mean a sudden burst of inflation. (B) would be a whole lot worse than (A).

So Scott's right that if we compare similarly sized "fiscal" and "monetary" interventions, the monetary one is a much bigger deal. This means that if we bundle the two together, the monetary component is the one to really focus on.

If it seems that, to the contrary, the fiscal component is a bigger deal, that's probably because the "monetary" intervention doesn't represent any actual shift in monetary policy. (For instance, issuing more base money to buy assets, but then planning to use IOR or reverse repo to implement the exact same inflation-targeting interest rate rule as before..)

Scott Sumner writes:

James, You missed the point of my example. The thought experiment I contemplated involves the Fed injecting new base money into the economy. So the monetary base would increase in net terms, contrary to your claim. The Treasury responds to this action by borrowing that much less.

You said:

"Seriously, why do you think anyone takes the other side of the trade when the central bank buys assets? Is it because they all discover they need to hold more cash and less bonds to reach their target allocations?"

Are you suggesting that the Fed purchase of bonds would cause bond prices to rise? Then why did massive Fed bond purchases in 1965-81 cause bond prices to plunge? And even if you are correct, the result would be no different from the impact on bond prices if public employees were paid in newly printed cash, and hence the Treasury did not have to issue as many new bonds as otherwise.

Bob, Are you also contemplating the Fed using negative helicopter drops, when needed to hit their targets?

But yes, a helicopter drop is a fiscal stimulus, and the "Cantillon effects" from fiscal policy may be a bit larger than for monetary policy. Still not very important for the US, in my view. But it's the Cantillon effects from monetary policy that are utterly trivial.

I see Matt agrees, see his comment below.

Thanks Dikran.

Thanks for the link Lorenzo.

Luis, Interestingly, the ultra old Japanese just elected Abe in a landslide, promising higher inflation for perhaps the first time in modern history.

Matt, I agree with your comments. Notice that I referred to the yen depreciating relative to the previously expected path of the exchange rate. That was my attempt to finesse the temporary/permanent distinction.

I certainly agree that money is not neutral in the short run, and hence that the exchange rate channel may be more important (initially) than some of the other channels.

Excellent comments, much better than my post!

James, I would argue that asset markets do not respond in the way one would expect if the Cantillon effect theory were true. After more expansionary than expected surprises in January 2001 and September 2007, bond prices fell. After a more contractionary than expected policy shift in December 2007, bond prices actually rose.

Michael Byrnes writes:

"Excellent comments, much better than my post!"

Actually, you have been on a tremendous roll of excellent posts lately, this one included. (But I'll agree the comments were good too.)

Comments for this entry have been closed
Return to top