« Canadian Macro Performance: Better than the G-7 but... | Main | University Debt II: A Longer-Term Perspective »

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Possible counterargument:

Imagine that the central bank is only capable of producing a new NGDP projection ever six weeks, but they are able to project the overnight rate (based on the quantity of base money) or the quantity of money (based on the overnight rate) in real time. Imagine also that for a fixed quantity of money, the overnight rate may vary wildly from day to day (and vice versa). Finally, imagine that wild swings (from 0 to infinity and back) in the overnight rate are very bad, but swings in the quantity of base money are cheap and harmless. In this case the central bank would be well advised to ignore the quantity of money and only pay attention to the rate of interest.

This argument is K's position, as I understand it (apologies if I've misunderstood).

Alex: I think that might be a sensible counterargument to my final paragraph. But it rests on an empirical (or possibly a normative) assumption: that short term volatility in interest rates "matters" but that short term volatility in the quantity of money doesn't. That's not obvious to me. Does it really matter for any real saving or investment decision if the interest rate is 4% on even-numbered weeks and 6% on odd-numbered weeks? Does it matter any more than if the quantity of money is higher on odd-weeks than on even weeks?

The answer might depend on what is causing the volatility. Some volatility might be desirable. If people don't want to work nights and weekends, it might be desirable to have prices higher at night and weekends, which means that *real* interest rates ought to fluctuate on an hourly and daily basis.

Even if people use paper, not gold, as money, the quantity of money will be endogenous under the gold standard, where the central bank targets the price of gold.
Under a gold standard, gold is the medium of account, notes are the medium of exchange operating under a convertibility constraint (the conversion rate, call it g). The price of gold is therefore the inverse of the price level (well, g/P; but g is fixed). So, you are claiming that a central bank under the gold standard targets the price level.

Surely the point of the gold standard is not having to worry about the price level, since that is anchored by gold via the convertibility constraint. Central banks during the Great Depression sure did not look like they were targeting the price level. It looked like they were targeting gold stocks to preserve the convertibility constraint. But the conversion rate is not the price of gold; gold has no specific price under the gold standard except the inverse of the price level.

Also, I wonder if Scott Sumner puts more emphasis on the MoA role of money than you because he's written a book on gold standard era economics?

Nick,

I think that when people say "money is endogenous" this is usually meant to say that active monetary policy cannot influence the economy. This is spoken by people describing themselves as Post-Keynesians and Modern Monetary Theorists. The usual argument is that there is no "money multiplier" as taught in textbooks. They argue that banks are not compelled by some structural "money multiplier" to increase loans when the CB increases reserves. And as long as commercial banks are free to NOT increase loans when the CB adds new reserves, then there does not exist a "monetary transmission mechanism."

So, when monetary economists like Sumner say that monetary economics does not need to include banks because only the supply and demand for base money matters, such individuals are outraged because they argue that by fully understanding how banking works in the "real world" that you will see that CBs cannot "force" the economy to do anything. Therefore, according to their view, all changes in AD are caused by forces within the economy separate from the CB and that the CB therefore cannot move AD whenever and wherever it wants.

In other words.... the CB cannot control the money supply because it has no control over the money multiplier. Therefore, all this talk about monetary policy moving AD is meaningless.

Here's a nice post expressing this view.... http://socialdemocracy21stcentury.blogspot.com/2013/04/endogenous-money-101.html

Lorenzo: Let Pg be the price of gold in terms of paper, and let Pc be the price of (say) consumption goods in terms of paper. So the price of consumption goods in ounces of gold will be Pc/Pg.

Under inflation targeting the central bank holds Pc fixed (strictly, makes it grow at 2%) and lets Pg vary endogenously.

Under the Gold Standard the central bank holds Pg fixed and lets Pc vary endogenously.

They would only be the same if Pc/Pg never changed. But things like the discovery of gold in Australia would cause Pc/Pg to rise, because it makes gold cheaper to produce relative to consumption goods. This would make Pg fall under inflation targeting, but Pc rise under the gold standard.

Dunno if Scott's emphasis on the Medium of Account function of money is related to his work on the gold standard. Might be. I think it's just because he doesn't understand how incredibly important the MOE function is! ;-)

JoeMac: "I think that when people say "money is endogenous" this is usually meant to say that active monetary policy cannot influence the economy."

That might be what they really mean. But why do they say it that way? Because even an extreme Neo-Wicksellian who thinks that the quantity of money is determined by the quantity of money demanded at the rate of interest set by the central bank and that the textbook money multiplier story is BS, and thinks that he can safely ignore the quantity of money in his model, still recognises that the central bank's monetary policy matters. And central banks are full of people like that. You might say that's the "orthodox view", if anything is.

In short, whether the quantity of money is exogenous or endogenous doesn't determine whether the central bank's monetary policy does or does not affect AD.

I skimmed that post you linked, but I didn't see it say that monetary policy doesn't matter. Did I miss it?

Nick,

The post I linked does not say that. I had that post bookmarked and without thinking linked to it. I apologize.

When you say, "In short, whether the quantity of money is exogenous or endogenous doesn't determine whether the central bank's monetary policy does or does not affect AD"...

This is the idea I have always had difficulty grasping. I hope you will discuss it in future posts.

Nick,

If I am reading this correctly, exogenous variables are similar to control variables in any experiment. Endogenous variables are similar to resultant variables in the same experiment.

"Even if you thought of monetary policy as a central bank adjusting a rate of interest, and not as adjusting the money supply, you need to remember that the rate of interest it chooses will not be exogenous."

The interest rate that a central bank chooses to lend at is an exogenous variable in it's own model of the economy (it has sole control over that rate). If you or I were coming up with a model of the economy, we would treat the central bank lending rate as an endogenous variable (neither of us are central bank members).

The U. S. central bank has the ability to do two different things:
1. First federal reserve act (circa 1920) - Set interbank lending rate by decree
2. Second federal reserve act (circa 1930) - Set the price of existing marketable U. S. government bonds via open market purchases / sales

In the first case, the interbank lending rate is an exogenous variable in the central bank's model of the economy. This rate is set by decree.

In the second case, in the central bank's model of the economy, the price that the central bank wants to buy and sell the bonds for is exogenous ( interest rate target ) but the price the bonds are actually bought and sold for is endogenous. The central bank has the ability to hit it's interest rate target because of the Primary Dealer legal arrangement.

JoeMac: "This is the idea I have always had difficulty grasping. I hope you will discuss it in future posts."

I don't see the problem. Suppose the CB is on the gold standard, so the quantity of money is endogenous. If the CB decides to double the price of gold, all prices (and hence NGDP) will also approximately double in the long run as a result (as will the quantity of money). It's just halving the value of its paper money.

Frank: "If I am reading this correctly, exogenous variables are similar to control variables in any experiment. Endogenous variables are similar to resultant variables in the same experiment."

I think that analogy is about right.

"The interest rate that a central bank chooses to lend at is an exogenous variable in it's own model of the economy (it has sole control over that rate). If you or I were coming up with a model of the economy, we would treat the central bank lending rate as an endogenous variable (neither of us are central bank members)."

Interesting point. Short version: "In my model, my actions are exogenous, but your reactions are endogenous"

Strangely though, it's not quite true of the Bank of Canada's model (or wasn't, in the past). Because the Bank of Canada also needs to model people's expectations of the Bank's future decisions. And the Bank knows that people know that the Bank will adjust the rate of interest in future in response to changing information in order to keep inflation at 2%. So the Bank modelled its own future behaviour as an endogenous variable.

There are 2 common threads to the issue of endogenous money.

The less controversial one is simply that the money supply is not directly (exogenously) controlled by the Fed, but is endogenously controlled through response functions. The nature of such response functions then becomes a subject for investigation.

The other thread is the way the usual macro equations based on NIPA accounting hide the role of bank money creation. If a bank creates new deposits, when these are spent for value added purposes, GDP, GDI and GDE all instantly simultaneously increase by the amount of the expenditure, because the equations are accounting identities. Of course, if you write an equation involving Y(t) and Y(t+1) you have a very different situation. Unfortunately the two are often confounded.

You can do this and get correct results in a model without money, because you can smuggle money creation in by some side door - perhaps in an expectation function. You can see, however, how some people might feel this is a touch perverse.

The question comes up when you ask - what effect did the increase in credit have on the boom, and what effect did the reversal of this flow have on the bust? How do you formulate it?

Peter N: "The less controversial one is simply that the money supply is not directly (exogenously) controlled by the Fed, but is endogenously controlled through response functions. The nature of such response functions then becomes a subject for investigation."

Let's take an analogy: "The speed of my car is not directly controlled by me, but depends on lots of things like whether my car is going uphill or downhill". In a sense that's true, but it won't get you off a speeding ticket. If I want to, I can keep the speed of my car almost exactly at 100km/hr by watching the speedometer and varying the gas pedal to compensate for things like hills. This analogy does require the central bank to have good data on the quantity of money (a good speedometer). And whereas my car has a limit on horsepower and brakes, so it's only true within limits, the central bank does not run out of paper and ink, usually, and can also reduce the monetary base to zero (provided its assets are not less than its monetary liablities).

"The other thread is the way the usual macro equations based on NIPA accounting hide the role of bank money creation."

You lost me on that bit. But if NIPA accounting hides commercial banks' creating money, it's doing a very bad job of it. Because all(?) first year macro texts teach that commercial banks create money, as well as teaching NIPA accounting.

Maybe you are referring to that Steve Keen equation that goes (something like) "Aggregate Expenditure = Aggregate Income + change in money stock"? Where other post-keynesians nailed him for violating national income accounting identities? In my opinion, Steve is a bit muddled there. What he is maybe trying to say is (something like): "Planned aggregate expenditure = expected aggregate income + change in money stock - change in money demanded". And if he said that, he would be talking about the hot potato effect, saying something important, and he wouldn't be violating national income accounting identities. I did a post on this once.

OK,

I borrow $1000 from a bank and buy a snow blower. GDP GDI and GDE have all increased immediately by $1000 and will continue to increase in (conceptual) lockstep as the money circulates.

As a monetarist, you have to agree that fractional reserve banking creates money, and this money enters the economy.

Keen is guilty of confusing (or allowing people to confuse) an equation in which time appears explicitly with the identity in which it doesn't. The symbols are similar, but the meanings are different.

He's not really dealing with the national identities any more than an equation involving Y(t) and Y(t+1) does. Y(t) needn't equal Y(t+1), at least I hope not.

"Planned aggregate expenditure = expected aggregate income + change in money stock - change in money demanded" comes close. If you expressed it as expected Y(t+h) = Y(t)+M(t)-M(t-h)) or the like... and then took the limit as h goes to 0, you'd be even closer and the difference between expected and actual would vanish, since the future would be now.

You'd then have to deal with the discrepancy between NIPA accounting and actual flow of funds.

I'm not saying he's got this exactly right (in fact he seems to be rethinking things a bit), but, I think he's got a good idea.

And -- Of course the central bank has ultimate control, but there are delays, overshoots side effects and everything else control theory so generously provides. Provable bang-bang control obviously isn't enough.

Good post, Nick. Regarding the previous few comments, what's a good definition for "response function"?

JP:

Continuing Nick's car analogy, Nick is controlling the car by varying the depression of the gas pedal, which varies the flow of gasoline into the engine, which varies the... you get the picture. The "which varies the"s are the response functions (unless I completely misunderstand them). Granted, in the macroeconomy the response functions are not so (ahem) mechanical, but it's the same idea.

JP:

Continuing Nick's car analogy, Nick is controlling the car by varying the depression of the gas pedal, which varies the flow of gasoline into the engine, which varies the... you get the picture. The "which varies the"s are the response functions (unless I completely misunderstand them). Granted, in the macroeconomy the response functions are not so (ahem) mechanical, but it's the same idea.

Peter N: OK. I've just finished writing a post on "What Steve Keen is maybe trying to say". Which isn't about what he actually said, but which gets around all the NIA complaints and makes sense. From what you say, I don't think I'm in a totally different ballpark. But I'm hesitating about posting it. I don't need the aggro.

JP: Thanks! I figure that Alex has probably got it right. Or is it a term from control theory, about the feedback rule?

Seems to me this is all about conditional versus unconditional forecasts. For example, we see the issue arise all the time in deficit projections. If you want to accurately predict the deficit, you need an unconditional model that incorporates the endogenous future policies the legislature will enact. If you want to assess the deficit impact of a specific policy, then you need a conditional forecast in which the future legislature does not change policy after it is enacted. Policy is exogenous in this conditional forecast. Moreover, the unconditional forecast may well be a more accurate prediction, but it is still the wrong with which to assess policy.

This point makes me super frustrated when I argue with people who argue endogenous money theories in a debate about what policies the Fed should take. You can't argue the Fed should change policy by assuming they will change policy.

Thanks Nick, that clarifies it. Your language is different from David Glasner's.

Your point about shifts in gold stocks versus shifts in output is obviously right; hence the deflationary period from 1873-1896 and, after the Kalgoorlie and South African gold strikes, the inflationary period from 1896-1913. There is, however, a presumption there that the demand for gold has a stable relationship with output (so shifts in the supply of gold are what matter). If you like, using G as stock of gold, G[Pg]=Py, where Pg is fixed (and g should be a subscript, but I am not that html clever). So if G rises faster than y, P rises and if y rises faster than G, P falls.

One might be even characterise the role of the BoE as primus inter pares central bank and manager of the system as ensuring a stable relationship between demand for gold and output. (Hence suspending the cover ratio constraint as soon as it looked like it might be binding.)

But the presumption of a stable relationship between demand for gold and output is a bit like Friedman's presumption of stable velocity/demand to hold money; not something to be relied upon for policy purposes. The 1928-1933 debacle came from central banks driving up the demand for gold by hoarding gold and restricting note issue (forcing more use of gold), thereby driving down the price level of goods and services. I am not sure this can be characterised as "targeting the price of gold".

"Or is it a term from control theory, about the feedback rule?"

Control theory talks about transfer functions. When looked at in the frequency domain, each block in a linear system control diagram has a transfer function (a polynomial in the frequency parameter s usually... of a Laplace transform). Then you talk about the open loop transfer functions (you can have several... depending on where you start and end) and the closed loop transfer function.

Of course I suppose economies are very non-linear (and time-varying), and thus this frequency domain stuff is out the door... even if it's multivariate (and still linear) w/ multiple ins & outs it's easiest to go to a state space model (time varying is OK there too) and you get a transition matrix (in the discrete time).

But you guys are probably talking non-linear controls and I don't know too much about that. All I recall from my school days is your luck if you get stability most of the time... and there are ways to analyze that. The phrase "Lyapunov Stability" comes to mind, but that's a pretty fuzzy memory to me now.

Of course there's always the possibility of linearizing a non-linear system around a nominal operating point. My experience with this is in regards to building Kalman filters or unscented filters ... for state estimation... usually not the whole control system. But I think that can be done as well.

Also, the box labeled "controller" in a classic feedback control system is said to implement the "control law."

http://en.wikipedia.org/wiki/Control_theory

Here's my own take on applying classic feedback control theory diagrams and terminology (no math!) to inflation targeting... you can skip most of the words... just look at figure 2 at the bottom:

http://brown-blog-5.blogspot.com/2013/06/inflation-targeting-as-feedback-control.html

Here I depict it as a discrete time control system with a high sample rate inner loop (the FFR targeting loop) and a lower rate outside loop (the inflation targeting itself).

... also for the super nerds out there, rather than a DEMUX (one input, two outputs) I should probably depict the switch operating a MUX (i.e. two inputs, one output).

"If the money supply curve is vertical (perfectly inelastic) with respect to any other variable in your model, then the quantity of money is exogenous. "

OK. Never happens. The government can force a minimum money supply, but seems incapable of reducing the maximum money supply intentionally. Witness "money market funds". The supply will often increase to meet demand.

But then the supply of money can also change in a sudden, discontinuous fashion: witness the results of the sudden "demonetization" of money market funds in 2008, which was disastrous for the world economy.

" But just because a variable is endogenous doesn't mean you should ignore it. "
Well, duh! It means the variable *has* to be analysed properly! It can't be treated as a mere parameter.

The comments to this entry are closed.

Search this site

  • Google

    WWW
    worthwhile.typepad.com
Blog powered by Typepad