Endogenous Versus Exogenous Money, One More Time

Few debates are as central to the heterodox-mainstream divide as endogenous and exogenous money theories. Neoclassical economists side with the exogenous ‘money multiplier’ idea, which says the banks receive reserves from the central bank, which they then lend out. Endogenous money proponents – generally post-Keynesian – side with another story, which says that banks create loans ‘out of nothing’ first, then the central bank more or less passively accommodates their demand for reserves.

In a final bid to demonstrate to economists that there is a difference between the two theories (and that theirs is wrong), I’m going to go through the age old scientific method known as ‘falsification,’ analysing each prediction of endogenous and exogenous money, and asking whether or not it corroborates with the data.

#1: exogenous money predicts reserves, or base money, would move first, followed by broader, credit based measures of the money supply as this was ‘lent out.’ Endogenous money predicts reserves would move last.

The relevant data was, ironically, put forward most conclusively by none other than the Real Business Cycle theorists Kydland and Prescott:

There is no evidence that either the monetary base or M1 leads the cycle, although some economists still believe this monetary myth. Both the monetary base and M1 series are generally procyclical and, if anything, the monetary base lags the cycle slightly.

By some measures the broader money supply moves many months before the monetary base. This is in keeping with the endogenous theory of money.

#2: exogenous money predicts that the central bank can control the quantity of base money at will. Endogenous money predicts it must play the role of passive accommodation, otherwise the economy will implode.

As John Kenneth Galbraith put it, “Milton Friedman’s misfortune is that his economic policies have been tried.” Monetarism was tried in the early 1980s: in the UK, the US and Chile. But the central banks consistently undershot their money targets and interest rates went wild. The policy certainly succeeded in increasing unemployment and therefore taming inflation, but the attempt to control the money base failed completely. Again, this is evidence in favour of endogenous money.

#3: exogenous money predicts that reserves factor into bank’s lending decisions; endogenous money predicts that they do not, and decisions about reserves are taken after loans are made.

Anyone from the real world – bankers, lawyers, accountants – will tell you banks do not consider reserves when lending. They use double entry bookkeeping to simultaneously create an asset (your loan) and a liability (your deposit, in which the loan is ‘stored.’) The books are balanced; if they need reserves they will borrow them from other banks or, failing that, the central bank.

That the central bank could say no means some might interpret endogenous money as a policy issue, but it isn’t – as I explained in my previous post, it reflects the reality of capitalism: banks must make lending decisions based on endogenous activity, and the central bank must accommodate this. If it does not, the credit markets will simply not work, and credit crunches will ensue.

#4: exogenous money predicts that only the distribution, not the level of private debt matters

In the exogenous story, banks act as intermediaries between savers and borrowers. Money is deposited; the bank lends this out. It does this until it either does not want to lend out for whatever reason, or until it hits the limit of how few reserves it can hold. Private debt simply represents a distribution from one person to another; the level alone should not have much macroeconomic impact.

In the endogenous story, banks create purchasing power out of nothing by crediting their customer’s accounts. This implies that an increase in private debt will add to nominal aggregate demand, and hence that private debt expansion will precede nominal growth, and also be correlated with other economic variables:*

We might expect some correlation between economic activity and private debt in the neoclassical model. But a correlation as robust as this, with private debt moving first, strongly supports the endogenous money theory.

#5: exogenous money predicts that increases in the money base will have an impact on bank’s lending and other economic indicators (with some qualifiers). Endogenous money predicts this will be minimal (though reserves may ‘oil the wheels’ of the system somewhat).

Ben Bernanke’s unprecedented doubling of the monetary base obviously did not lead to a massive surge in lending. Neoclassical economists do have some explanations for this, namely that the ‘money multiplier’ collapsed – in other words, banks do not want to lend out the reserves, or there is a lack of demand. This is believable, but really it’s just a tautology – if reserves will increase lending except when they won’t, economists have told us nothing. We do not have a scientific proposition.

Overall I’ll put this one down as ambiguous, but its certainly not a falsification of endogenous money.

I’ve always sided with endogenous money because it is supported by the evidence. If anyone can offer me contrary evidence about the above or other relevant hypotheses, I’ll be happy to listen. But economist-y special pleading about how, even though exogenous money is wrong, the economy behaves as if it is right, or about how I’m not allowed to refute ‘centuries of theory,’ is simply not enough when the evidence is this strong.

*If you are wondering what the ‘credit accelerator’ is, it is the change in the change in debt (acceleration), which is what matters when you do the basic calculations. For more, see here.


, , , ,

  1. #1 by Ron Ronson on September 22, 2012 - 5:45 pm

    The way that I understand how the system to work is as follows:

    – The CB sets an interest rate that private banks can borrow from it at
    – private banks lend as much as they want to. They will need reserves to fund these loans
    – They can if they choose borrow these reserves from the CB
    – The CB will create whatever reserves it needs to fund this lending
    – if the CB wants to change the amount of lending it can change the rate of interest in the appropriate direction
    – Only at the zero bound can it truly be said that the CB has no control,over the amount of lending.

    I think this system has many of the features you describe as being true for endogenous money while in my opinion being really exogenous since the CB is setting the overnight rate.

    Much of what you say about lending is true: But it is the CB that has orchestrated that expansion of lending. Private banks could not have done it on their own with a CB to provide the reserves. I suppose you could say that private banks somehow control the CB and make it do their bidding but that seems more like a conspiracy theory than serious economics.

    • #2 by Unlearningecon on September 23, 2012 - 2:43 pm

      You are absolutely correct but the point is that the amount of lending is generally endogenous, and the CB responds after it takes place. Both neoclassical and post-Keynesian theories have endogenous and exogenous elements, though – I guess I wouldn’t read too much into the names.

  2. #3 by Ron Ronson on September 22, 2012 - 5:47 pm

    ” Private banks could not have done it on their own with a CB ”
    should say
    ” Private banks could not have done it on their own without a CB”

  3. #4 by AMT on September 22, 2012 - 6:30 pm

    As a functional matter, how would a bank operate under the exogenous money proposition?

    The way I read the mechanism, the bank would have to acquire reserves from the Fed and pay interest while holding them, waiting to make loans to consumers. The bank would want to hold as few Fed loans as possible to lower its own costs, and by limiting its stock of held reserves it would be vulnerable to any spike in demand for loans. I’d expect to run into many situations where loan approvals were pushed back a day or two until the bank could acquire more Fed reserves overnight.

    I’ve only taken loans from a bank, and never worked for one, but unless they’re dressing up the process to hide that they’re waiting to get reserves to cover my loan, or routinely keep enough reserves on hand to cover walk-ins wanting credit, it seems like the exogenous model doesn’t track with experience. To me, it appears more likely that banks do just create reserves out of thin air, then fill that air with real reserves from the Fed.

    If we assume that banks have a big say in how the Fed operates, the endogenous model seems to be the one they’d prefer. Were I running a bank, I’d want a system where I make loans based on simple algorithms to calculate risk and the appropriate interest rate, then buy Fed reserves at the end of the day to meet statutory requirements. The alternative, exogenous method seems to imply that as a banker, I’d have to play a double game of assessing a customer’s creditworthiness while also trying to gauge how much I could afford to loan based on the Fed funds I had sitting around.

    I may, however, be completely misunderstanding the dynamic, so feel free to correct me where I’m wrong.

    • #5 by Unlearningecon on September 23, 2012 - 2:52 pm

      No, you’re right – see also Neil’s comment on building societies below.

  4. #6 by Mathieu Dufresne on September 22, 2012 - 7:14 pm

    The credit accelerator is not the change in debt. As Keen uses it, it’s the acceleration of debt (the change in the change in debt) divided by GDP.

    I don’t know if you read that post but it provides additionnal evidences against the the liquidity trap hypothesis vs endogenous money and shows that adding the change in debt to aggregate demand doesn’t involves double counting.


    • #7 by Unlearningecon on September 23, 2012 - 2:41 pm

      Yes, sorry. Original version had two unfinished sentences and this error – that’s what I get for rushing.

  5. #8 by LuisMartos on September 22, 2012 - 7:27 pm

    but really it’s just a tautology


    All neoclassical economics is a giant tautology and more an article of faith than a real science
    nothing to see here

  6. #9 by Cara on September 22, 2012 - 8:19 pm

    Thank you so much for this. I just finished taking an introductory economics course through CGA Canada. As I was reading through the chapter regarding the money multiplier I felt like it wasn’t really explaining the real world, but I couldn’t really explain why.


    Cara Galka

    On Sat, Sep 22, 2012 at 11:20 AM, Unlearning Economics wrote:

    > ** > Unlearningecon posted: “Few debates are as central to the > heterodox-mainstream divide as endogenous and exogenous money theories. > Neoclassical economists side with the exogenous ‘money multiplier’ idea, > which says the banks receive reserves from the central bank, which they > the”

  7. #10 by ivansml on September 22, 2012 - 9:22 pm

    #1: Kydland & Prescott work is far from being “most conclusive” (correlation !=causation, their method may be sensitive to choice of smoothing parameter in HP filter, they don’t deal with statistical significance, etc.).

    #2: yes, targeting of monetary aggregates was not success, but it hardly meant collapse of capitalism.

    #3: OK, whatever.

    #4: see below.

    #5: rise in monetary base didn’t lead to rise in deposits because of zero lower bound – this was not a surprise, since it’s what mainstream theory non-tautologically predicts.

    Point #4 seems key to me. On one hand, level vs. distribution is just semantics – after all, total world net debt is zero, by definition (any creation of debt in double-entry bookkeeping creates corresponding asset), so (from global view) only distribution can matter. On the other hand, “banks create purchasing power from nothing” and “banks are not intermediaries” are claims that don’t automatically follow from previous points (see also my comment on previous post), yet seem to play key role in heterodox theories, so it would be nice to justify them in more detail.

    About the graphs with debt – how do you know that causality doesn’t go the other way? In bad times, banks are more reluctant to provide loans and firms are less likely to take out loans, so naturally debt goes down. That change in debt leads change in unemployment can be caused by other factors (e.g. banks are forward-looking and limit credit from the beginning of recession, while it may take several months for unemployment to rise).

    • #11 by Unlearningecon on September 23, 2012 - 2:49 pm

      #1 Well is there much evidence in favour of exogenous money?

      From what I understand HP filters generally suffer from a ‘confirmation bias’ problem, in that you can tweak the parameters untyil you get roughly the conclusion you want. That it came from two primary RBC theorists suggests to me this wasn’t the case.

      #2 it didn’t collapse capitalism because they overshot their targets! It still caused recessions, but if they had well and truly stuck to the M0 target it would have been much worse.

      #4 It completely follows from point 3 – you go into a bank, the guy gives you a £1000 loan – where did this £1000 come from? It adds to AD.

      The fact that private debt consistently moves first suggests causality goes in the other direction. The expectations-style argument you are using can be applied to pretty much any contradictory evidence to try and get rid of empirical problems – I could then say that expectations of credit drying up cause the unemployment, and we go into an infinite regress.

      • #12 by ivansml on September 23, 2012 - 3:30 pm

        Kydland and Prescott wrote that paper mainly to support their line of research, which explains economics fluctuations as a result of exogenous technology (or more generally, supply-side) shocks, abstracting from money (and debt) altogether. Probably not the kind of theory you’d endorse.

        It’s starting to feel like running in circles. OK, guy in bank approves your loan and gives you money in your deposit account. At this stage, the new deposit could have been created from nothing. But you take loan to do something with it, so you typically either withdraw cash or make an interbank transfer. For this, bank does need reserves, and those reserves can ultimately come from either other deposits or central bank (and the precise timing, whether reserves have to come before or after issuing debt, is irrelevant here). If it comes from other deposits, then owners of those deposits are not using them for purchases at the moment – this is the whole point of fractional reserve banking. So no, not every deposit adds to aggregate demand.

        Situation would change if bank’s deposits would circulate directly and universally as medium of exchange, without having to be converted in cash or reserves. Essentially, the bank would function as its own central bank, with capacity to create medium of exchange at will. Presumably, this is the model Keen has in mind, but (and I’m repeating myself) I don’t see why this would be a realistic description of the world.

      • #13 by Unlearningecon on September 23, 2012 - 4:40 pm

        Perhaps not every loan adds to aggregate demand, but the question is whether the ones that do are statistically significant. Again, Keen’s evidence would suggest they are.

        It’s not really about whether a model is ‘realistic;’ it’s about whether the simplifications are scientifically defensible. Keen’s model can easily have a central bank incorporated, with the only role is plays as setting the interest rate. Exogenous interest rates are also compatible with other post-Keynesian models such as Sraffa’s.

      • #14 by Mathieu Dufresne on September 23, 2012 - 3:44 pm

        I wouldn’t argue that the causality is running only in one direction. The process we’re talking about is a positive feedback loop, for example the slowdown in the change in mortage debt causes house prices to fall and the fall in house prices causes the change in debt to slowdown, the causality is circular. Altought the slowdown in debt usually precedes other correlated variables at the turning point of a bubble, a careful analysis of leads and lags over different time periods shows unstable leads and lags, which is what is expected in a complex dynamic system involving a lot of feedbacks.


        That said, if banks needed reserves before they can lend, the monetary base should never lag the cycle. While hoping to prove the opposite, Kydland and Prescott confirmed the results of several studies in post-keynesian litterature, the most important of those beiing the work of Basil Moore.

      • #15 by ivansml on September 23, 2012 - 5:03 pm

        Problem is that in the story I described, purchasing power = cash+reserves (at least if I understand how you use the term), so it doesn’t make sense at all to say that loans add to aggregate demand.

        But anyway, I got you to invoke “Friedman’s defense”, so I’m satisfied for the day 😉

      • #16 by Unlearningecon on September 23, 2012 - 10:35 pm

        Whoah whoah, Friedman said ‘don’t open the black box as long as empirical evidence is OK’ (incidentally, Keen’s models should appeal to you if you endorse Friedman’s methodology). Anyway, what I’m saying is completely different – having a pure private banking system is a neglibility assumption. Central banks can be added later.

        The CB provides the purchasing power in name – on this you are correct. But the credit expansion is really a product of endogenous decisions, with some costs set by the CB.

    • #17 by moiracathleen (@moiracathleen) on September 24, 2012 - 2:23 am

      Do you understand how commercial banks use credit default swaps to evade bank capital adequacy requirements? If you understand how credit default swaps function, then there is no way to defend your claim that the central banks control lending by controlling reserves.

      • #18 by ivansml on September 24, 2012 - 2:38 am

        You’re aware that reserve and capital requirements are different things, right? The first one deals with asset side of bank’s balance sheet, the second one with liabilities/equity side.

  8. #19 by Julia on September 22, 2012 - 10:18 pm

    Much thanks for this. I question whether or not “fiat currency” truly is the root of all evil – as many austrians claim – and that going back to gold (or silver) as currency would solve most of the problems our society faces today. It’s an easy scapegoat invented by people who don’t truly understand how the money system works.

    • #20 by Unlearningecon on September 23, 2012 - 2:50 pm

      Austrians are also generally resistant to endogenous money because it collapses their ‘central banks cause recessions’ view. Having said that, some Austrian economists like Lachmann and Schumpeter accepted it. Here is a blog by a Lachmannian.

  9. #21 by Eric L on September 23, 2012 - 1:49 am

    So I’ve never understood why it makes a big difference whether banks get deposits first then loan or make loans then find deposits. Even making the deposit first, their decisions will be constrained by what rates they expect to be able to get deposits at. That said, they need not be constrained by reserve ratio requirements, as they can turn loans into securities at which point they are not backed by any reserves.


    “after all, total world net debt is zero, by definition (any creation of debt in double-entry bookkeeping creates corresponding asset), so (from global view) only distribution can matter.”

    I’m not convinced “only distribution can matter” follows from “total world net debt is zero.” A bank with $1,000,000 in assets and $900,000 in liabilities is not in the same situation as a bank with $10,000,000 in assets and $9,900,000 in liabilities, and likewise for the economy as a whole. The issue is one of stability. I don’t think this has anything to do with exogenous vs. endogenous money, but this is one area heterodox economists are right to stress. The question is one of whether, when the economy doesn’t do exactly as expected, does it self-correct, or does it spiral out of control? The more leveraged the economy, the more closely it most match what lenders expect to stay in equilibrium, but high levels of debt do nothing to increase the clairvoyance of investors or otherwise stabilize the economy, so with enough debt the economy will transition from a self-stabilizing one to a self-destabilizing one.

  10. #22 by Neil Wilson (@neilwilson) on September 23, 2012 - 6:08 am

    “As a functional matter, how would a bank operate under the exogenous money proposition?”

    As a building society did for generations. It gets deposits from its members ‘short’ and loans them out ‘long’.

    The charge on the loan is the profit of the building society, part of which pays the return to the members and part of which is held in reserve against bad loans.

    If you get too many bad loans then member’s deposits are hit pro-rata.

    In the UK we used to have a restriction that prevented banks with reserve accounts from providing mortgages. And that meant you had to go to a building society to get a mortgage. They would have very limited amounts of money each month to lend out, and the lending criteria were very strict indeed. Often you’d have to have been a member of the society for some time and built up a good deposit before you’d be considered for a housing loan.

    After deregulation the more aggressive UK building societies were desperate to get hold of a reserve account at the Bank of England – because they knew that turns the lending game on its head.

    • #23 by AMT on September 23, 2012 - 4:31 pm

      Interesting, thank you for that.

  11. #24 by kkalev on September 23, 2012 - 8:09 am

    As Minsky said, ‘everyone can create liabilities, the problem is to get them accepted’.

    The mechanisms of private bank credit creation are rather simple actually. A bank expands its balance sheet by creating an asset (loan) and a liability (deposit) constrained only by capital adequacy rules. In a system of no required reserves (like Canada, required reserves are actually mainly used to smooth the interbank rate during the reserve maintenance period) it does need any reserves for its operations.

    Reserves are actually only required in order to settle transactions wth other private parties. These are interbank payments and currency withdrawals. The central bank sets the price for these reserves and through it, the ultimate liability cost for banks.

    Neoclassical theory, when suggesting that the issue is only one of redistribution, ultimately starts by thinking that money can really be regarded as a gold coin, an asset without a corresponding matching liability. The fact that money is created through credit means that debt repayments does not move monetary resources from debtors to creditors but rather destroys them. The original debt and deposit are destroyed and interest (over and above the original loan amount) is paid to the originating bank. Bank ‘investment’ is increasing their capital base through retained earnings in order to be able to provide for more loans (in an expansion) or handle NPLs (in a downturn).

  12. #25 by marris on September 23, 2012 - 2:22 pm

    Any explanation of money that says that the CB either has no control or plays a purely accommodating role for the banks is going to need *another* explanation of while CB’s seem able to target the price level and inflation. For example, the Bank of Canada has empirically maintained a 2 percent inflation target for 30 years. Coincidentally, this matches the policy target posted on their website.

    Are the individual Canadian banks somehow all magically coordinating their loan issuance to magically hit the 2 percent bound? Or is there some structural property of the Canadian economy that says profitable loans appear at a rate of 2 percent per year? And that structural thing has not changed over 30 years?

    I think that’s all bunk. Just because banks don’t need reserves today to lend today does not mean that over the long term, the reserves don’t play a role. The CB’s target announcements certainly have effects on how banks perceive risk in the economy. It changes how they issue loans. If that’s the case, then the post-Keynesian *theory* outline above is wrong (or at the very least, incomplete).

    • #26 by Unlearningecon on September 23, 2012 - 2:40 pm

      Nobody ever said the CB has no control: it has control over interest rates. This is why it can influence inflation (as well as its control over the base rate).

    • #27 by marris on September 23, 2012 - 3:50 pm

      Well, if the CB *is* controlling the price level, then this means that it *is* controlling credit creation. After all, the CB is not directly handing money to people and businesses. Those guys get their money from private intermediaries or fiscal spending.

      Since the price level matches the CB target, the causality *does* seem to run from CB policy to CB actions (reserve creation through asset purchases, interest on reserves, and commitment) to bank actions. The CB does not merely provide reserves “as needed.”

    • #28 by Unlearningecon on September 23, 2012 - 4:26 pm

      I disagree with your first sentence. Control over the base rate, for example, reduces demand and hence increases unemployment, lowering wage inflation. That isn’t necessarily about credit creation (and from the data the base rate seems to a be a key tool used by CBs to control inflation).

    • #29 by marris on September 23, 2012 - 6:02 pm

      OK, let’s start even more basic. Do you think that people can get money (which they can use to buy stuff) through credit? [they can get it in other ways as well, but do you think this is one of the ways?] I believe all macro schools, including the various flavors of Keynesians believe this.

      If you agree so far, then do you believe that increases in bank credit creation can create inflation (increases in the price level)?

      If you agree so far, then do you believe that any mechanism that seeks to control the price level must control bank credit creation? Otherwise, credit creation will simply cause the economy to either overshoot or undershoot the price level, right?

      If you agree so far, then given that the Bank of Canada *has* kept the price level on it’s target path, can we not infer that it *has been able* to control credit creation?

      This is all really basic stuff. If you disagree with this chain of reasoning, please tell me *the point(s)* at which you think the argument breaks down. Either a theoretical or empirical counterargument would be OK.

      • #30 by Unlearningecon on September 23, 2012 - 10:38 pm

        I think you are presenting a false dichotomy. That credit creation can create inflation does not mean it has to be controlled in order to control inflation – there are many other mechanisms.

        As I’ve said, I believe the CB has a limited degree of control by controlling interest rates. If they are higher, markets move there, bank’s costs increase, and generally lending will decrease. But that does not mean the CB can exercise close control over M, in any form.

      • #31 by Eric L on September 24, 2012 - 6:21 am

        “If they are higher, […] generally lending will decrease.”

        So the CB does affect how much lending the banking sector does, then? Either I don’t understand what endogenous money means or I don’t understand what the fuss is about…

        Let’s back up to this statement: “the central bank more or less passively accommodates their demand for reserves.” What precisely is the demand for reserves? Is it affected by the CB or is it something that exists independently of it? It seems to me that banks will lend different amounts if they expect to have to borrow at 6% to get reserves than if they expect to have to borrow at 2%, right?

        Is endogenous money simply the proposition that it is more useful to think of lending as constrained by the interest rates banks expect to have to pay to get reserves (which the CB influences) rather than by the amount of potential reserves that exist (which the CB influences)? If so, I would note that the actual role played by the CB is no different in either story — after all, the way the CB influences interest rates is by expanding or contracting the money supply — but would note that the endogenous view seems to be what the CB believes, as they announce interest rate targets and not money supply targets, so one would presume they will create whatever amount of money is needed to keep bank lending at that rate. Is this claim controversial?

        Or is it the proposition that there is a finite amount of demand for reserves determined by a finite willingness to lend, and any attempt to expand the money supply beyond that won’t actually affect the amount of money in circulation? I find this claim plausible, provided it is understood as applying when at the zero lower bound. However if you are claiming this also describes how things were working in the 80s and 90s I don’t buy it; obviously banks would have lent more if they could have borrowed for less, so the CB was not accommodating all lending that could have been done, and so long as enough lending was being done to keep the economy close to full employment they had no need to.

      • #32 by Unlearningecon on September 24, 2012 - 10:10 am

        I don’t think we even disagree. The point of endogenous money is that credit moves first, then reserves do, contrary to the typical ‘money multiplier,’ story taught in economics textbooks. It also claims that loans, by definition, add to aggregate demand and so private debt a key component of growth, rather than a simple ‘redistribution’ between borrowers and savers.

        Both your second points are true: banks do not ‘lend out reserves’ and as such increasing them will not increase lending. If the CB tries to control the number of reserves, it will cause serious problems in the credit markets; but if it controls the price, then it can afford to screw over a few bond traders to stay on target. It’s sort of a functional rather than logical distinction, but as Keen and I have both argued, the endogenous accommodation of loans reflects the income streams in capitalism.

        Policy implications are: don’t try to control the money supply; watch private debt (which needs to go into actual investment); monetary policy is almost useless at the ZLB. I guess many reasonable mainstream economists would not disagree with these.

      • #33 by Eric L on September 25, 2012 - 6:17 am

        “Both your second points are true: banks do not ‘lend out reserves’ and as such increasing them will not increase lending. If the CB tries to control the number of reserves, it will cause serious problems in the credit markets; but if it controls the price, then it can afford to screw over a few bond traders to stay on target.”

        If the CB wants to expand reserves, it does so by printing money and buying bonds. If the CB wants to push down rates, it prints money and buys bonds. It only has one dimension of control over the economy. So when not at the ZLB, it makes no sense to say that the CB can’t increase lending by expanding reserves but can by lowering rates; whichever way it tries to increase lending it will do the same thing and have the same effect.

        “Policy implications are: don’t try to control the money supply; watch private debt (which needs to go into actual investment); monetary policy is almost useless at the ZLB.”

        I agree with all this but the reasoning that has lead me to these conclusions has nothing to do with whether banks lend first and then find reserves or vice-versa. So I’ve never understood the energy heterodox economists put into this point given that it is now sounding to me like it is a much more trivial point than they make it sound like and it serves mainly to distract from more important discussions.

        Perhaps an analogy can help to show how the theories may be dual theories: Let’s say you have a business, and one of the inputs for your product is copper. Will the quantity of available copper affect how much of the product you produce? Absolutely. So are you going to look at the quantity of copper to decide how much to produce? No, you will most likely look only at the price of copper; it is much easier to make business decisions on that basis. So let’s say a government body could create copper by fiat, and it wants to stabilize your production. It could do this by varying the amount of copper that exists. However since your business decisions are based on price, it would create a lot more certainty for you and thus more easily influence you by announcing a price target for copper and then creating the amount necessary to meet that target. Now let’s say in this world copper became as abundant as dirt and as cheap as dirt. Your production will increase, but it will still be constrained by other economic concerns. Would doubling the amount of copper available increase your production much more? Probably not.

      • #34 by Nathanael on September 25, 2012 - 12:31 pm

        Marris: the important point is the *time sequence*. It matters what *order* things happen in. Having a false story about what *order* things happen in leads to subtle errors which are very hard to track down.

  13. #35 by bond guy on September 23, 2012 - 3:16 pm

    I think we need to be careful about the definition of “neoclassical” or “mainstream” economics. I’m a market practitioner, and not an expert on academic tribalism. However, my understanding is that DSGE-style (micro-founded) modelling is the state of the art “neoclassical” model, and everything done before that is viewed as internally inconsistent verbal hand waving. In other words, “neoclassical” economists, if they want to be consistent with their theoretical framework, should have no opinion on the subject.

    To the best of my very limited knowledge, there are no full(*) DSGE models that incorporate a realistic banking sector. “Money” in the models is base money, M3 (or M2 or whatever) does not exist as a concept.

    These models require that the central bank follow some sort of rule, like a Taylor rule, to complete the equations. I believe the market monetarists think the central bank can guide expectations for the growth path of the monetary base while superficially appearing to have nominal interest rates as an *operational* target (based on what Nick Rowe writes). This can be done because the theoretical simplicity of the model structure allows one to convert the interest rate rule into a rule on the monetary base (I assume this has been done; I don’t know the literature).

    However, it’s all hand-waving if you try to extend these conclusions to an economy where a banking sector exists. If we wanted to model the banking system in a microfoundation framework, such a rule would require us to model this optimisation problem of banks: how should I price a particular loan in the current period (t), if the central bank is actually targeting the aggregate reserve position of ALL banks in the period ahead (t+1)? As noted above, I’m not an expert on the literature, but my guess is that the problem is intractable, and no such model exists.

    (*) In practice, central banks use “DSGE-like” models for the economy, which includes things like banking sectors, However, those models consist of internally-consistent “DSGE” cores, along with internally-INconsistent stuff tacked on in order to make the model generate something resembling useful outputs for forecasting. So yes, these models have banking systems, but they do not qualify as being state-of-the-art models.

    • #36 by Unlearningecon on September 23, 2012 - 4:28 pm

      Great comment. There is indeed a massive trade off between internal consistency and empirical validity with mainstream macro, something Simon Wren-Lewis talks about a lot.

      NGDP targeters don’t really have a ‘model’ – to be honest all they seem to have is the QtoM and references to hyperinflation to ‘prove’ a central bank can always engineer inflation.

  14. #37 by Frances Coppola on September 23, 2012 - 11:53 pm

    Can I just point out that.although loans create deposits, drawdown of those deposits requires reserves. Central banks inflation control is in effect done by influencing the cost of settlement funding. Indirectly this influences credit creation, since all loans have to be drawn at some point and those drawings have to be funded.

    How it SHOULD work is that as more credit is created and the demand for reserves increases, central banks oblige by producing more reserves (M0) but charge more for using them. It is assumed that banks will respond by raising rates to borrowers, which discourages borrowers and hence reduces the rate of loan creation.

    However, transmission of central bank policy through to commercial interest rates can fail – both when demand for credit is very high, when raising policy rates may not be sufficient to encourage banks to raise rates because they are making more money from high volumes than from interest spreads, and when demand for credit is very low, when banks’ interest margins can be seriously squeezed even if the policy rate is near zero. It can also fail due to risk aversion by weak banks: banks that don’t want to lend, perhaps because they already have too many non-performing loans, may continue to charge high rates to borrowers, particularly risky ones such as SMEs, even when policy rates are on the floor.

    Conventional economic models fail to take into account the possibility of policy transmission failure. They effectively treat banks as passive intermediaries, which is very far from the truth.

  15. #38 by kkalev on September 24, 2012 - 6:28 am

    @ivansml #12

    “For this, bank does need reserves, and those reserves can ultimately come from either other deposits or central bank (and the precise timing, whether reserves have to come before or after issuing debt, is irrelevant here).”

    1) The fact that reserve requirements are binding ex post and not ex ante is not irrelevant. In reality they are binding a few months after credit creation. For example, maintenance periods of the ECB refer to deposit data 2 months old: http://www.ecb.int/press/pr/date/2011/html/pr110520.en.html

    2) Reserves can only come from the central bank. The latter will make sure that there are exactly as many excess reserves in the system in order to achieve its target rate without any problems in interbank payments (which are helped by overdraft facilities such as daylight overdrafts and marginal lending facility).

    The monetary base is an endogenous variable which changes based on the needs of the interbank market and currency demand by the general public. How much of the bank assets will be transformed into the MB by the central bank does not play any role in lending. Before the 2007 crisis excess reserves of US banks were of the order of $10-15bn while daylight overdrafts peaked around $150bn without any effects on bank lending.

  16. #39 by Mathieu Dufresne on September 24, 2012 - 2:43 pm

    I would recommend not taking a condescendant tone when you’re clearly ignorant of non-linear correlations, which can have negative r squared. And the visual correlation appear positive because unemployment is inverted…

    • #40 by Unlearningecon on September 24, 2012 - 5:57 pm

      pontus, go away and never come back. I might have responded to your comment if it weren’t so condescending and full of insults, but to be honest you just demonstrate a complete lack of capacity to grasp that the causal mechanics you yourself describe and agree with:

      (a) Mean that debt adds to aggregate demand, because banks expand credit before the central bank expands monetary policy.

      (b) Mean that mainstream textbooks – which often say something different to what you describe, something I have documented on this blog – are wrong in their description of the mechanics of banking. As are DSGE models like Krugman’s recent paper, who assume banks are merely intermediaries.

      Have a tantrum elsewhere.

    • #41 by ivansml on September 24, 2012 - 6:09 pm

      @Mathieu: econometric refresher: in a linear model R^2 has its precise definition and can never be negative (in univariate case, it’s equal to squared Pearson correlation coefficient). There are various pseudo-R^2 measures for nonlinear models (like probit) which can sometimes be negative, but that’s irrelevant here. Sorry, but whoever labeled those graphs is clueless about basic definitions.

      • #42 by Unlearningecon on September 24, 2012 - 6:28 pm

        There is no problem with a negative r^2 in a non-linear correlation calculation. God this is so simple.

      • #43 by ivansml on September 24, 2012 - 6:32 pm

        I’ve taken several courses in econometrics and yet this is news to me, so feel free to explain. Please start with defining “nonlinear correlation”.

      • #44 by Unlearningecon on September 24, 2012 - 6:41 pm

        A correlation that isn’t linear?

        I’m tired and can’t be bothered to argue with people like you and Pontus. I guess I’d say ‘let’s agree to disagree’ and you can go back to using exogenous money models where private debt level doesn’t matter and CB can control money supply, whilst simultaneously saying you don’t do that. Neither of us with ever convince the other – the debate is fruitless.

      • #45 by ivansml on September 24, 2012 - 7:00 pm

        Correlation has precise mathematical definition. R-squared (or coefficient of determination) in a linear regression model has also precise definition (I could quote textbooks, but really Wikipedia has it all). The definition implies that 0 <= R^2 <= 1, for any data. Are we now arguing about mathematical facts?

        You're probably right, the debate is pointless because you don't even bother to reply to our arguments. Instead you keep repeating your claims, over and over again, as if that made them magically true.

      • #46 by Unlearningecon on September 24, 2012 - 7:03 pm

        I’m not arguing a squared number can be negative, I’m arguing Keen’s use of R^2 is not a squared number.

        And I repeat my claims because you never respond to them. You have not convincingly responded to my 5 evidence points above, because endogenous money is right. All you can do is special pleading and muddying the waters.

        Just stop commenting here. It’s the same every time.

      • #47 by Mathieu Dufresne on September 24, 2012 - 11:45 pm

        After verification, this is really a correlation coefficient R, not R^2, so there’s a mistake in the legend. Where did you found those graphs?

      • #48 by Mathieu Dufresne on September 24, 2012 - 6:41 pm

        Ivan, I don’t checked out how Keen did his calculus but I would never expect a linear correlation between a 1st and a 2nd derivative. Actually, that’s right it’s a pseudo-R^2, it’s not something squared. I guess the term R^2 is used because it’s more a coefficient of determination.

      • #49 by Oilfield Trash on September 24, 2012 - 7:49 pm

        If you are a student of Economics and what drives you is “getting it right” and not about “being right” I would encourage you to follow the link


        Keen put on his blog all of his lectures and it does not take long if you have a some post graduate level economic exposure to understand his Macro theories, and his criticisms of current neoclassical thinking. It also should give you a base line on how Keen prepared the graphs above.

        The fact that he found a correlation between a 1st and 2nd derivative over 12 years in his data should if anything get your attention and motivate you to a least figure out if he is right or possiblity why he is wrong.

        Also on this blog is a presentation where he unveils a mathematical proof of his AD= Income plus the change in debt.


        I do not want to spoil it for you, but the issue of ex post and ex ante impacts on aggregate thinking is worth the time.

      • #50 by Unlearningecon on September 24, 2012 - 9:59 pm

        Thank you. I am positively exasperated at people who have taken a glance at Keen’s models and made superficial criticisms.

      • #51 by Oilfield Trash on September 25, 2012 - 6:34 pm

        As a heretic you will require the patience of Jobe in your efforts to move heterodox thinking to orthodox. As the president said it is diffcult trying to make changes from the inside, but I like your grit.

  17. #52 by yorksranter on September 24, 2012 - 6:44 pm

    I’m not sure I really get the importance of the pre-funding vs. post-funding thing. I can see that it would be important in a once-off lending decision, but it’s a repeated decision. Even if the system is post-funded, after I (speaking as a representative loan officer through the marvels of praxeology) have made a loan and funded it, what do I do next? Probably I open the next loan application in my inbox, at which point the cost of the last lot of funding is bound to influence my reasoning.

    I mean, in any kind of economic model it *must*. RatEx – it’s information, in fact it’s price information, the only sort you care about. Adaptive expectations – you adjust your expectations based on the last lot of information. Myopic expectations (aka 50% accurate weather forecasting) – your forecast is always the actual price from last time. If you’re using some sort of empirical model to estimate risks (like an actual loan officer), the current interest rate for a comparable risk is obviously going to be a key input.

    • #53 by Unlearningecon on September 24, 2012 - 6:48 pm

      So the banks consider the cost of borrowing instead of constraining lending based on the amount of reserves they can get. Economists don’t seem to think there is a difference as they toboggan up and down linear curves, but there clearly is. They are different things.

      • #54 by yorksranter on September 24, 2012 - 9:35 pm

        OK, consider a prefunding model where the prefunding is specific to the loan (which is just disaggregating a model with bulk prefunding).

      • #55 by Unlearningecon on September 24, 2012 - 10:00 pm

        I’m not sure I understand where you are coming from…?

    • #56 by Nathanael on September 25, 2012 - 12:35 pm

      OK, here’s one scenario where the pre-funding vs. post-funding matters: suppose the bank is overly optimistic, makes a bunch of bad loans, and then suddenly finds that its balance sheet looks really really ugly and the central bank doesn’t want to backstop this.

      This looks really different from the case where the reserves are doled out first!

      And crucially, this is *exactly the situation which happened in the banking collapse of 2008*, so it’s an important situation to get right.

      • #57 by Eric L on September 26, 2012 - 4:18 am

        But this isn’t what happened. Whether a bank makes a bad loan before it gets the deposits or after makes no difference; it will be just as bad for the bank a year down the road when the payments stop coming.

      • #58 by Nathanael on October 1, 2012 - 4:19 pm

        Wrong, Eric. It makes a big difference. Sure, it’ll be just as bad *a year down the road*, but that’s a year down the road, and the people running banks don’t look past the next quarterly statement.

        If the bank makes loans before deposits, then what happens is that the depositors look at the book of bad loans, refuse to deposit money, and the bank goes to the government for a bailout. *THIS IS WHAT HAPPENED IN 2008*.

        If the bank (in the style of a English Building Society) has to get the deposits first, the bank is limited in the degree to which it can overextend itself.

        I suppose it could round up a huge amount of deposits and *then* go on a spree of bad loans, but the depositors would wonder why the bank was sitting on cash, and stop depositing. (Also, earnings would be terrible.)

        Barring that scenario, the *rate* at which the building society could make bad loans would be limited to the rate at which it could attract deposits, which would give the depositors a chance to pull the plug well before the massive over-extension of credit.

        In contrast, when loans come first, the credit is already overextended, and the banks are demanding government bailouts, before the depositors have even gone to look at whether the bank is a safe place to deposit money.

        So, Eric, you should see that it is *crucially* important what *order* things happen in. Building societies — which had to get deposits before making loans — have a record of having far fewer catastrophic failures due to loan over-extension than banks — which make loans before getting deposits — do.

  18. #59 by Nathanael on September 25, 2012 - 12:26 pm

    By the way I’ve been archive-diving. Unlearningecon, you have compiled what is probably the best collection of resources on *actual* economics *ever*, and you write very clearly. Book contract? The factors-of-production article is perhaps the most useful outline of the state of *proper* microeconomics I’ve seen.

    Get the entries in this blog into the right order, and rather than a “heterodox” book like _Debunking Economics_, you may well have a book which surveys the *actual* state of knowledge in economics, as distinct from the nonsense which dominates the schools. (The real state of knowledge is more like early-19th century physics, which had lots of little theories which worked for particular situations — optics vs. waves vs. mechanics — than it is like the grand unifications which happened much later)

  19. #60 by Hedlund on September 25, 2012 - 12:31 pm

    I always figured he listed r^2 as negative to signify that he’d inverted one of the two graphs, as per the right-hand label. I don’t think this is standard econometric practice, though. (It may also be an artifact of whatever software he uses, but who knows.)

  20. #61 by Unlearningecon on September 25, 2012 - 3:21 pm

    Hedlund and Mathieu,

    Thanks for doing the research. I think it’s obvious this was just a superifical problem and Keen did not actually think he had obtained a negative square number. Btw, I got it from a blog post not by him I can’t find. The inverted version is here. Perhaps somebody else changed it.


    Thanks, and good points.

    Eric L,

    Nathanael made a good point about the importance of causality. Also bear in mind that nowhere in mainstream macro does it say banks create debt out of nothing which adds to nominal demand. I know I keep repeating this but it is important, and economists don’t believe it (as Krugman demonstrated in his debate with Keen).

    • #62 by Eric L on September 26, 2012 - 4:39 am

      “Also bear in mind that nowhere in mainstream macro does it say banks create debt out of nothing which adds to nominal demand.”

      I don’t see what this has to do with endogenous money.

      Are you trying to tell me that if I put money into a bank at 3:00 and it gives you a loan at 4:00, our actions cancel each other out from an aggregate demand standpoint, but if a bank loans you money at 3:00 and I deposit it at 4:00, the value of the loan gets added to aggregate demand?

      I don’t think that’s what you mean, but if it isn’t what you mean, then you seem to me to be missing the point as much as Krugman. Because it doesn’t matter which happens first, either way you’re gaining the ability to spend at no perceived loss to me. There’s also the issue as I mentioned last thread that I have reasons to save regardless of whether I’m making interest and if you decide against the loan at the last minute that won’t cause me to spend instead.

      • #63 by Mathieu Dufresne on September 26, 2012 - 1:26 pm

        I have to say I never saw how the vision of credit being only a transfert of spending power was consistant with the money multiplier model. I tend to agree with Eric on that, if you would properly model the transmission mechanism of monetary expansion according to the money multiplier model, the change in debt would still add to aggregate demand.

      • #64 by Unlearningecon on September 26, 2012 - 3:59 pm

        Are you trying to tell me that if I put money into a bank at 3:00 and it gives you a loan at 4:00, our actions cancel each other out from an aggregate demand standpoint, but if a bank loans you money at 3:00 and I deposit it at 4:00, the value of the loan gets added to aggregate demand?

        You are assuming the deposit comes! The question is, what if the bank cannot acquire it? It must come from the central bank.

        Maybe the issue is one of framing. In endogenous story, banks create debt and the central bank sells them any needed reserves at the agreed price. In the exogenous story, the banks get the reserves first and lend them out. Both additions to aggregate demand ultimately depend on the central bank, but if the central bank is simply accommodating endogenous demand for loans, this is only true in name.

      • #65 by Eric L on September 28, 2012 - 6:59 am

        “You are assuming the deposit comes! The question is, what if the bank cannot acquire it? It must come from the central bank.”

        I’m assuming the deposit comes because that is the interesting case. As you note in your next post, exogenous money proponents have no trouble dealing with the fact that printing money adds to aggregate demand. But debt is several times the money supply, which tells you that the vast majority of times banks do not need to go to the CB for reserves. Do the loans that are matched to deposits matter from an AD standpoint, or are they just a transfer of purchasing power from savers to debtors? See above for my thinking on that.

        Both additions to aggregate demand ultimately depend on the central bank

        No, no, no. Loans can change aggregate demand even if the CB doesn’t print a dime and every cent loaned is covered by deposits. This is the point I’m talking about when I say you’re missing the point as much as Krugman.

      • #66 by Unlearningecon on September 29, 2012 - 2:13 pm

        But banks do not receive deposits and ‘lend them out;’ loans create their own deposits, though there is indeed some debt in the economy that is simply redistributive. Exogenous money can possibly deal with the increase in AD, but that doesn’t mean its right. As I say, the evidence points towards endogenous money.

      • #67 by Nathanael on October 1, 2012 - 4:22 pm

        But the case where the deposit *doesn’t* come is critically important, because it’s the case which causes banking crises!

        Exogenous money — which can be created by strict, harshly enforced banking laws, such as those which applied to Building Societies in the UK — is more resistant to banking-crisis explosions.

        (The downside is that an exogenous money regime, a bit like a gold standard regime, makes depressions and long periods of high unemployment last longer and makes them harder to get out of.)

      • #68 by Eric L on October 6, 2012 - 5:06 am


        “Loans create their own deposits.”

        It is true that when a bank creates a loan it creates an equal deposit. However most loans are quickly spent and unless they happened to be deposited in the same bank as a result those loans will need to be backed by other deposits. It’s not really wrong, but it obscures more than it illuminates, and in particular it handwaves away the interdependencies that loans create that make a highly leveraged economy more fragile. In this sense it is just as true and just as unhelpful as neoclassicals noting that net debt is zero.

        In the more general discussion, let’s separate this into three independent propositions:

        1. Banks create loans, then find deposits.
        2. Loans are added to aggregate demand.
        3. Loans are added to aggregate demand because banks create loans, then find deposits.

        It is #3 that I claim is wrong, and this is a very important point independently of the validity of the other two, because if you’re getting this wrong then you aren’t fully understanding your own theories and you’re putting a lot of energy into making arguments that others are rightly finding unconvincing. And that’s a shame, because point #2 is important.


        Bank runs don’t happen because banks make loans and the deposits never come, they happen because the deposits that were covering the loans leave. You can make only good loans and match every loan to deposits and have a reserve, but if your depositors all leave at once you’ll be looking for a bailout. If anything the endogenous money perspective does worse here, because endogenous money proponents tend to ignore depositors and treat banks as unconstrained by them, but bank failures happen precisely when they are a crucial constraint.

      • #69 by Unlearningecon on October 7, 2012 - 1:29 pm

        Banks create loans, then find deposits

        You mean reserves, right? Loans create deposits, as you’ve said.

        I feel like you might be confusing capital and reserves. Deposits add to a bank’s capital base and as such they do improve its position. Reserves are equity, which is different.

      • #70 by Eric L on October 7, 2012 - 8:20 pm

        You find reserves by finding deposits.

        Let’s say you create a $1,000 loan, simultaneously creating $1,000 in deposits in the account of the loan recipient. You have no reserves backing that $1,000 deposit, so you will need to find an additional $111 in deposits if you want a 10% reserve ratio. If the loan is then spent, the $1,000 deposit you created most likely leaves for another bank, so you need to find another $1,000 in deposits. In either case you find the deposits the same way.

  21. #71 by y on September 25, 2012 - 10:02 pm

    To Marris’ point that the CB controls inflation.

    They don’t control asset price inflation, i.e. house price inflation. That is not counted in their inflation measure. That is where the increase in AD shows up, with inflated asset prices. This is what Steve Keen talks about. Of course people don’t take out huge loans to go and buy CPI goods, in general.

  22. #72 by Frances Coppola on October 8, 2012 - 12:47 am

    UE, can I correct a statement you made earlier about deposits and reserves? You said deposits add to the capital base, whereas reserves are equity. This isn’t true.
    1) Deposits – wherever they come from – are debt, not capital.
    2) Reserves are the cash balance in a commercial bank’s current (settlement) account at the central bank. In the UK, Canada and Denmark (inter alia) this balance does not currently have to be greater than zero once all transactions have been settled, but other CBs do impose a positive reserve requirement, which is in effect a minimum cash balance in the current account. That cash balance is a liability of the central bank and an asset of the commercial bank, but it is most likely to have been funded by borrowing so there will be a corresponding liability on the commercial bank’s balance sheet. In the extraordinary event that the commercial bank had NO DEBT, that balancing item would be equity. I hope this makes sense.
    Equity is shareholders’ funds in various forms and it makes up the capital base of the bank.

    Eric #57: unless a loan is very large (a syndicated aircraft loan, for example) it would be very unlikely that a bank would match fund with deposits as you suggest. Deposits created by lending are indistinguishable from “real” customer deposits: drawdown of either creates a funding requirement at the central bank. Funding of deposit drawdowns is generally handled on an aggregate basis and forecasted daily. Whether there are in practice sufficient reserves to settle drawdowns without borrowing depends on the pattern of deposits and withdrawals on a particular day. Funding gaps are met through interbank lending or, as a last resort, overnight repo at the central bank, not by attracting additional customer deposits – banks don’t really have any influence over those on a day-to-day basis, though obviously over a period of time the attractiveness of the rates they pay to savers will influence the volume of deposits and the rate of drawdown. But the point is that banks lend without considering reserve availability as such, though they do consider cost of funding and that will influence their willingness to lend – or, more accurately, it will influence the willingness of borrowers to borrow at the rates the banks want to charge.

    I guess what I’m saying is that the money multiplier is way too simplistic and doesn’t remotely resemble how banks actually work in practice. Models are all very well but if they depart too far from reality, their usefulness becomes questionable.

    • #73 by Unlearningecon on October 8, 2012 - 9:03 am

      Thanks, Frances. Perhaps I am still somewhat confused about the details myself, but I do think I have it broadly right.

      I am confused about why people such as Eric L (not singling him out) seem to be insisting that the world behaves as if the money multiplier is right, when clearly the causal mechanics are wrong.

  23. #74 by Jamie on October 18, 2012 - 2:08 pm

    I think there’s a typo in the article text:

    “The policy certainly succeeded in increasing unemployment and therefore taming inflation, but the attempt to control the money base failed completely. Again, this is evidence in favour of endogenous money.”

    – I think “money base” was meant to be “money supply” – Yes/No?

    • #75 by Unlearningecon on October 18, 2012 - 2:16 pm

      They were unable to control both the money base and the broader money supply IIRC.

  1. Endogenous Versus Exogenous Money, One More Time « Unlearning Economics « Economics Info
  2. Exogenous and Endogenous Money: Room for Reconciliation? « Unlearning Economics
  3. Debunking Economics, Part XII: Keen’s Minsky Model « Unlearning Economics