« University Debt II: A Longer-Term Perspective | Main | Flea Market Economics »


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

PPS Yes Nick, I think you've nailed the major issue very accurately. This has been an extremely refreshing experience for me, and--pardon the pun, but--I feel indebted to you for your contribution here!

PPS I think you've nailed the major issues in a very helpful way here Nick--which has been an extremely refreshing experience for me. Pardon the pun, but I now feel indebted to you--in the best sense of the word!

Nick Edmonds:

I need to "Fisk" you!

"I suspect where we differ is whether the increase in spending is permanent or temporary. I think it's temporary. The borrower can't keep up his spending. He financed it with the loan. He's blown that now and has just got debt service payments. And if his spending has fallen, other people's income will have fallen as well. We end up closer to where we were before."

Nope. The original borrower has spent what he borrowed, so his increase in spending was temporary, but if he spent it buying a bike the bike seller also had a temporary increase in income. He now had a temporarily higher income, and holds more money than he expected to hold, so will (presumably) spend that excess money, and so on...

"But what about the new money created? Isn't that going to keep going round until we satisfy the money demand equation Md = kPY ?"

Yep. You anticipated my objection. Why can't the new money keep on going round and round (like a hot potato), expanding PY and expected PY as it does, until the new money is willingly held. In other words, why can't the excess supply of money cause the quantity of money demanded to expand until its no longer an excess supply, rather than the excess supply of money simply refluxing to the banks that issued it?

"Well, if M was fixed by the amount of loans, then maybe there might be something in that. But if money is just the checking account part, then that's not the case."

There you've lost me. If people are still borrowing from the banks, M will keep on expanding. Even if they don't keep borrowing, but don't repay their existing loans (and just roll them over) M stays permanently higher. I'm going to make two responses, based on my guess about what you are saying:

1. Simple intro textbook story where: there is a stable currency/deposit ratio; there is a stable reserve/deposit ratio; central bank holds its monetary liabilities (currency+commercial bank reserves held on deposit at the central bank) constant. Then yes. This process stops. It doesn't even get started unless C/D changes, R/D changes, or MB changes, or someone makes a mistake. And if someone does make a mistake, it goes right back to where it started. (Though maybe one bank could expand while another bank contracts so nothing changes in aggregate.)

2. My story: this process stops when the central bank wants it to stop because PY is getting too big for the central bank's inflation target (or whatever target the central bank has). So the central bank restricts the growth in base money/raises its interest rate/tells people it is going to do that in future/whatever to make it stop.

So, is this process temporary or permanent? Well, it depends on what the central bank is targeting, and on how long it takes to hit the central bank's target.

"But provided the assets are good enough so we can ignore those risks, it is only the liabilities side that matters in the money-creation process.

So in my view:

A bank buying an IOU
A bank buying a computer
A bank buying a meal at a restaurant for its staff to celebrate Christmas
A bank giving money to charity
are all the same, in terms of creating money, and their effects on the liabilities side, though they will have very different effects on the asset side."

I agree with that. Alternatively, you can interpret it as to what is happening on the liability/equity side as per my explanation earlier. The creation of money as a result of asset acquisition of any type always increases bank liabilities at the margin. The question then is what happens to the equity account in conjunction with that. That goes immediately to the marginal solvency impact.

In a loan/deposit asset swap, the equity account doesn't change (although any subsequent deterioration in loan quality will result in a write-down of equity).

In buying a computer, the equity account declines immediately - IF the computer is expensed from an accounting perspective, which amounts to viewing it as having zero realizable value as an asset externally, which is separate from its unique ongoing utility inside the organization. That's pretty consistent with it having zero value for purposes of meeting the demands of a bank run.

On the continuous vs discrete time question: To keep it simple for myself, I just assumed discrete time, and that it took exactly one "month" for people to make a plan, borrow the money, spend that money, and then look at their income over that month and money stocks and revise their expectations and plans. If people revise their expectations and plans more quickly, one "month" will be too long. In principle we could (well, someone could, but my math isn't up to the job) model this as a continuous time process, where people see continuous fluctuations in their incomes and money holdings, but don't know if these fluctuations are transitory or permanent, and so it takes some time for people to realise that there is a permanent change.

Thanks Steve! Yep, I feel this discussion is going well, and maybe getting us somewhere. (It has been hard for me over the years trying to get my head around all this. Money is hard.)


I appreciate your "fisking" - I've learned a new phrase there as well.

There are one or two comments I might make, but I'll limit myself to explaining the bit where I lost you.

You said that what you meant by money was not overall bank liabilities but just the balance of checking accounts. That being the case, the statement:

"Even if they don't keep borrowing, but don't repay their existing loans (and just roll them over) M stays permanently higher."

is not correct because, as we already discussed, M (using your definition) can easily be reduced by people simply transferring balances from checking accounts to time deposits.

So, taking M = kPY, it is far from clear that M (again, your definition) is fixed and determines PY. In my world view, PY is principally determined by other factors including overall assets and liabilities, and PY then determines M (although, as with anything there is feedback).

Put another way, I think how much people spend depends on their overall asset / liability position, not just their "ready money", and this determines how much of their wealth they wish to keep in checking accounts.

Yes that's a pretty good statement of it Nick, and it's comparable to the arguments put here (that someone on Twitter just reminded me of):


From my dynamic systems perspective, the phenomenon we're now both describing--and its difference with the conventional view--is the growth factor in what mathematicians call a "dissipative system". The conventional position, derived as it has been from a static model of macroeconomics, has unwittingly instead described what mathematicians call a "conservative system"--and in this case what is being conserved is the amount of money, and also (except for variations in velocity) the amount of demand.

The creation of money by banks means of course that the amount of money in existence is not conserved but expands during booms (and can contract during slumps), as does the amount of demand (though not necessarily by precisely as much, for reasons including the one you gave initially about desired holdings of cash).


I am baffled by your use of gross domestic private non-financial sector debt in those tables.

Suppose someone defaults on their no-recourse loan, and so gross outstanding mortgage debt falls by $400,000. Is domestic private non-financial purchasing power really $400,000 less than before the default? Or is it exactly the same? In what way does the default take money out of the domestic private non-financial sector?

Suppose the government redeems from me a treasury bond. Does "AD" rise, fall, or remain the same? My cash resources have increased, but it looks to me like "AD" doesn't change-- there has been no change in GDP or in domestic private non-financial gross borrowing. Now suppose I take the cash and pay down debt owed to a bank. My cash resources fall back down to where they were, but domestic private non-financial gross borrowing falls. So "AD" falls as a consequence?

What am I missing? I guess somehow your "AD" also counts changes in the money supply? Is the "debt" component of "AD" something other than gross domestic private non-financial sector credit market debt outstanding?

for stephen gordon, on the other thread

8th attempt

Trade Union Density


Hi DR,

I'm not about to claim that my definitions are precisely accurate when applied to the aggregate data, but they have to be pretty accurate to generate the correlations you can see there. There could be instances where the reduction in debt doesn't reduce demand--though in your example (over to you JHK) I presume there could be a transfer from bank equity to bank capital to make up for the loss that would therefore reduce money in circulation and hence demand (to some degree).

I also start from measuring changes in debt rather than changes in money since (a) from the loans create deposits perspective, the change in privately generated money is 1:1 with the change in debt and (b) money is notoriously hard to measure whereas debt is easily measured in the Flow of Funds. So the debt component of AD as I measure it is precisely "gross domestic private non-financial sector credit market debt outstanding". As noted, some fine tuning could be necessary, but the correlations I get with economic data--especially since they go against conventional expectations--imply that I'm roughly right rather than precisely wrong.

Quick clarification: it's the change in the level of "gross domestic private non-financial sector credit market debt outstanding", not the level itself of course.


A mortgage default reduces bank equity capital. Banks lend based partly on capital adequacy (not central bank reserves). So to the degree that capital is diminished, there is less lending capacity at the margin. And if fewer loans are being made, then there is less aggregate demand to the degree that loans are a source of spending power. And the money supply will have been reduced by comparison to the counterfactual in which those loans might have been made from a stronger capital position. This is all at the margin of course, but it was the reason for TARP during the US crisis. And that should all be consistent with Steve’s work.

The redemption of a government treasury bond (net) at the margin reflects a marginal surplus – the bond is redeemed by taxes at the margin. (“Net” assumes no bond refinancing). If you are an MMT follower, that takes net financial assets (NFA) out of the private sector. That is a marginal reduction in private sector wealth. The bond holder now has cash instead of a bond, and therefore is net neutral from an asset value perspective, but some taxpayer is worse off (compared to the counterfactual of bond refinancing instead of taxation), for a net contractionary effect and a probable reduction in aggregate demand. I think Steve plans to integrate public sector effects on aggregate demand into his model later on if I’m not mistaken.


Don't get me mixed up with SK. I am concerned with net lending. SK is the one using gross.

Respecting the treasury bond, the monetary authority is part of the government too. There's no reason the Fed cannot monetize the debt, rather than raising taxes. What? Suddenly the government needs to raise funds in order to purchase something from the private sector? I thought that was antithetical to MMT.

But are banks capital constrained, or no? Do they have just enough equity that they can create additional loans if I don't default, but insufficient equity to create additional loans if I do? That's a pretty tight squeeze, don't you think?


The key word there is correlation, no? It hardly defies convention to say that as demand (proper demand, for currently-produced goods and services) slows/falls, private lending slows/falls. Even if you consider demand for assets as well, if demand falls, and purchasing power is in excess (or at least, there is less need for additional purchasing power.

I don't get it. It seems like causality is jumbled up and demand grafted together with purchasing power. It would help if there was clarity on what "AD" is supposed to measure. Adding production and changes in debt? Are we looking at GDP here as income, and so "AD" is some sort of purchasing power? Then what is the connection to demand? If "AD" is some sort of quantity demanded, why do we include government and foreign sources of demand for current production, but include neither in considering debt?

How does any of this work? What's the underlying behavior? If debt levels increase, and you argue this increases private non-financial purchasing power, then to the extent that the increased purchasing power increases demand for currently-produced goods and services, this is already reflected in the GDP component of "AD". Seems more likely that as demand rises, agents increase borrowing. Are you saying that domestic purchase of goods in excess of production is financed by borrowing? How is that different from the balance of payments? "Excess" purchases (on net) already shows up as a current account deficit/capital surplus.

So many questions...

Honestly, though. I'm not expecting to get sorted out on this stuff. Sadly, my priors are pretty heavy on the "nonsense" end of the spectrum, having been reinforced through previous "discussion" on this matter.

My main goal jumping in to this mess was to address the misrepresentation of Bernanke.

Hi D R,

Thanks for the caveat! Probably the best thing to do in that case is follow Hicks's lead and "Let us have recourse to a diagram"--or in this case a pair of dynamic models written in my system dynamics program Minsky. One sets out a monetary model of "Loanable Funds" (lending from "Patient to Impatient", following Krugman's terminology), the other the same of "Endogenous Money" (lending from bank to non-bank). The differences between the models are the minimum possible: where loans originate, where interest is paid, and the initial conditions.

In the former no matter how you change the lending and repayment parameters (tau_L & tau_R respectively, GDP remains unmoved as does the money supply); only variations to the rate of turnover of production (tau_T) has any impact. Changes in the debt ratio have no relevance to macroeconomics.

In the latter, changing the lending and repayment parameters drastically alter the rate of growth of GDP and of the money supply.

To download Minsky: https://sourceforge.net/projects/minsky/


[Steve: I had to fish this comment out of our spam filter, which has been playing up a lot recently. So if you don't see your comments appear quickly, that's why. NR]


I don't have any idea why playing with your models would help. I wouldn't know what to look for as I'm not even clear on what you're trying to say. Can you tell me what your "AD" is supposed to measure?

Anyway, I don't know if this is the place. If you are genuinely interested in helping me understand, I'll be happy to contact you by email.

Nick said: "If I sell my computer to my bank, the money supply expands. If I then buy that computer back from my bank, the money supply contracts.

Similarly: If I sell my IOU to the bank (if i take out a loan), the money supply expands. If I then buy that IOU back from the bank (if I repay the loan), the money supply contracts."

ProfSteveKeen said: "This is one reason that I have come to appreciate the importance of using double-entry bookkeeping to explain the actions of banks: your first example involves transactions only on the liabilities & equity side of their ledger which transfers existing money but does not create or destroy it; the second involves an increase in assets and liabilities when the loan is made that increases money (and demand) while the repayment reduces assets and liabilities (and demand)."

JKH said: "If the computer is capitalized on the balance sheet as an investment, the computer appears as an asset, equity is unchanged, and the money supply increases by the amount of the deposit created by the bank’s purchase of the computer. The balance sheet change is an increase in an asset matched by an increase in money supply.

If the computer is expensed on the income statement, the computer does not appear as an asset on the balance sheet, equity is reduced by the cost of the computer, and money supply increases by the amount of the deposit created by the bank’s purchase of the computer. The balance sheet change is a reduction in equity matched by an increase in money supply."

I'm going to try the accouting here.

Start a little unrealistic, but it will make the point.

Start bank A by selling $1,000 in bank stock and transfer from another bank. The equity account gets marked up, and the demand deposit(s) get destroyed.

Assets A = $1,000 in central bank reserves
Liabilities A = $0
Equity A = $1,000 in bank stock

Buy the computer for $1,000.

Assets A = $1,000 in central bank reserves plus $1,000 computer
Liabilities A = $1,000 in demand deposits
Equity A = $1,000

Expense the computer so the asset goes to zero reducing equity.

Assets A = $1,000 in central bank reserves
Liabilities A = $1,000 in demand deposits
Equity A = $0

Overall, $1,000 in demand deposits destroyed and recreated. Net change = $0.

ProfSteveKeen said: "It's more valuable to focus on is the point Nick concludes with that "What's special about banks is not what they buy with the money they create, but that they create money", and his original conclusion that my "effective demand is income plus the change in debt", at least when translated into his "Aggregate planned nominal expenditure equals aggregate expected nominal income plus amount of new money created by the banking system minus increase in the stock of money demanded", doesn't violate " any national income accounting identity", and therefore potentially indicates a disequilibrium source of additional demand to that identified in the conventional formula (and the reflux view)."

Do effective demand, nominal expenditure, and nominal income include financial assets or not?

D R, JKH, ProfSteveKeen

"But are banks capital constrained, or no?" Surely the distinction between individual banks (at least non-TBTF ones) and the banking system as a whole is central here. When a single middling US bank has too many defaulters it gets wound up by the FDIC. When US banking is threatened by an industry-wide crisis of defaults the result is TARP, relaxation of mark-to-market, ZIRP and so on. It's the lending of the banking sector as a whole that (the thesis seems to say) can't effectively be restrained by the central bank through reserve requirements (or what have you) without bringing the house down, not that of individual mavericks: viz. the old joke about a conservative banker being one who goes bust at the same time as everyone else.

Too Much Fed, yes they do--in the last 3 decades the main purpose of additional debt has been to finance asset purchases rather than production or consumption (though the dynamic between rising debt and rising demand applies in their absence as well, as Schumpeter argued), and a monetary macroeconomics has to integrate finance rather than treating it as a separate field.

D_R, The models illustrate the causal role of additional debt in financing an expansion of monetary activity--but you're right that I should provide more explanation. I will have to leave that for a while though--I'm off for a bit of tourism here in Ecuador in the next couple of days, but I'll try to return to this thread early next week.

Thanks again Nick for starting it.


Thanks, but I'm hopefully not going to be lurking around this page when you get back. So if you do choose to respond here, I'll probably never know. Hope you can help someone else.

Steve Keen,

If these are the same models that I asked you on a few weeks back, then I think it is worth making a point here.

Those models included the assumption that patient agents base their spending on their holdings of non-capital claims on banks, but ignore similar claims on non-banks. If you make that assumption, then you are obviously going to get a result where expansion of bank balance sheets matters and expansion of non-bank balance sheets doesn't.

All your models do is illustrate the consequences of that assumption (which is fine - that's all any model should be expected to do). However, I think the issue here is whether in fact that is a good assumption to make.

In my view, this is unrealistic. I think people take into account all their assets, not just their claims on banks. I'm quite prepared to believe that people place more emphasis on assets the more money-like they are, but that's quite different from drawing an arbitrary line between assets to be factored in and assets to be ignored.

This is an important issue for policy. If you believe that only bank lending matters and you try to control only bank lending, you will simply force disintermediation and you will be left wondering why your old relationships no longer hold.

Hi Nick (Edmonds). I'd be happy to generalise the model to include spending based on other factors. But while including (for example) spending based on share holdings would result in agents spending more, the question to me is whether they would have more to spend.

With an increase in bank deposits via more bank loans, they do; with an increase in other assets, there are some transactions where they do--such as one company buying another with its shares--but in general they don't.

Back to Nick Rowe's original point in this post: the disequilibrium phenomenon of taking out more bank loans and spending them adds to demand directly. That's the same point that those models illustrate--albeit intentionally in a very simple way, so that the difference between Loanable Funds and Endogenous Money is made as obvious as possible.

Hi Nicke ive managed to derive this same result with a very simple capital model and accounting identities - and it produces a very Wicksellian result


Andrew: I took a quick look at your post, and it seems to me you run into exactly the same NIA problem that Steve's original formulation had. If you say "AD = Income + change in debt":

1. Unless we are talking about an economy (like Cuba) where there are shortages of goods, then people actually buy what they want to buy, so AD = actual expenditure = actual income.

2. We can imagine an economy where I borrow $1,000 from Steve, and I plan to spend $1,000 more than my income, Steve plans to spend $1,000 less than his income, so AD=Y, but debt increases by $1,000.

My formulation gets around those two problems because: 1. I make the distinction between aggregate actual income and aggregate expected income; 2. money is different from other goods, so debt that creates money is different from other debt.

Steve (Keen): "Back to Nick Rowe's original point in this post: the disequilibrium phenomenon of taking out more bank loans and spending them adds to demand directly. That's the same point that those models illustrate--albeit intentionally in a very simple way, so that the difference between Loanable Funds and Endogenous Money is made as obvious as possible."

I'm going to restate what Steve just said in that last sentence in my words:

In the ISLM model: the IS curve shows the Loanable Funds theory (I=S); and the LM shows the Liquidity Preference theory (Md=Ms), and the whole point of the ISLM model is to show that *both* LF and LP theories are *partially* true, and that you need both LF and LP to get the whole truth. But what we are talking about here is a case in which the economy is "off" *both* the IS and LM curves, so that *neither* LF or LP theory is true.

It's "off" the IS curve, because the IS curve assumes that people make plans for current spending based on knowledge of their actual current income (which can only work if they know other people's spending plans), and we are talking about people being surprised about their current income. So their *actual* saving is greater than their *desired/planned* saving. And it's "off" the LM curve, because we are talking about people being surprised to discover they are holding more money than they wanted to hold.

If the ISLM model is talking about the "short run" (prices are sticky, but people are not surprised about their current income and money stocks), we are talking about the "very short run". How long the "very short run" lasts, is an empirical question.


"I'd be happy to generalise the model to include spending based on other factors. But while including (for example) spending based on share holdings would result in agents spending more, the question to me is whether they would have more to spend."

It seems to me that people choose how much of their wealth to hold in the form of bank checking accounts. Just as a general observation, I think people who have significant shareholdings, for example, hold enough ready money or to have sufficient lines of credit to cover what they want to spend. I'm not saying nobody ever has a liquidity problem, just that it's not a general feature of the economy.

So when you ask whether they would have more to spend, I say they already have enough (ready money) to spend.

The other important point is the distinction between ready money and total bank liabilities, when talking about the relationship with loans. This was the distinction I have been questioning Nick Rowe on.

Assume a simple bank balance sheet with loans (L) on the asset side and checking accounts (M) and time deposits (D) on the liability side. Ignoring equity L = M + D. The total of M + D is determined in the loan market. When banks lend they increase M and no matter what the non-banks do, they cannot get reduce M + D without paying down loans. But they can reduce M to whatever they like by switching between M and D.

So you have to decide which you think is determining spending.

If you decide it's M, then you have to accept that loans do not determine M. People can vary how much M they hold by varying D.

If you decide it's M + D, then you have to explain why time deposits at banks count and things like fixed income and money market funds don't.

The point here is that if claims on non-banks go up and that "requires" more M to support more spending, that extra M can easily come from a switch in the balance of M and D; it doesn't require more L.

If you're interested, I recently posted on my own blog setting this out in a little more detail.


@Nick:Banks buying computers vs. IOUs. "the principle is exactly the same."

Doesn't seem right. Because:

Bank buys my computer, I use the money to buy a computer. Change in the stock of computers in the real sector: zero.

(The bank has no use for my computer; see Frank Restly, above: "Farmer borrows from cab driver using farm tractor as collateral. Both default on loan - cab driver gets a farm tractor and farmer gets a taxi cab."

Bank buys my IOU. I buy a computer. Stock of computers in the real sector: +1.

The only way to make them the same in principle is to say that I'm buying borrowing now by selling future borrowing.

Nick's post said: "1. Unless we are talking about an economy (like Cuba) where there are shortages of goods, then people actually buy what they want to buy, so AD = actual expenditure = actual income.

2. We can imagine an economy where I borrow $1,000 from Steve, and I plan to spend $1,000 more than my income, Steve plans to spend $1,000 less than his income, so AD=Y, but debt increases by $1,000."

a) AD = actual expenditure = actual income and AD = Y.

If I'm remembering correctly, Nick once told me income is national income which is basically GDP (Y). ProfSteveKeen told me it includes financial assets. I didn't think national income/GDP/Y included financial assets. If so, problem there?

b) Debt increases by $1,000 but change in debt is unchanged ($1,000 saved is negative debt)? If ProfSteveKeen saves $1,000 in demand deposits and Nick borrows $1,000 in demand deposits, then the total number of demand deposits is unchanged. So what should the terminology be here?

@JKH and Ramanan:

I don't think your accounting objections wash. I think you're saying, basically, "Demand can't equal GDP+Change In Debt because demand equals GDP. Accounting violation!"

Nick will be the first one to tell you that demand doesn't equal GDP. GDP for a period tells us, roughly, where the demand and supply curves met during that period. There's no national accounting measure that will tell us what the demand was for that period. So I think Steve K's also wrong, but not for accounting-consistency reasons. Rather because he's also seeking to use (two) simple national-account measures to suss out what that elusive economic concept, demand, was or is -- how much people wanted, would have liked to buy (at that period's price point). You won't find that number in the national accounts.

GDP+Change in Debt seems to be an extremely useful measure for sussing out the state of an economy; cf Steve's rather eye-popping correlations. But labeling that sum "demand" doesn't work, I think. It has an important relationship to demand, tells us important things about demand, but the relationship between that measure and "demand" is more complex than that presentation.

Demand is not directly and simply observable (independent of its supply interaction) in national account measures. It's an economic concept, not an accounting measure.

I could say that real-sector AD for a period is best represented by GDP + Bank Lending to the Private Sector + Government Deficit + Trade Surplus.

(Don't parse that too deeply; I just threw it together without much examination, for example.)

You can't say that that's accounting-inconsistent. You can only say that it's not a useful measure, or that it's conceptually unwieldy or misleading -- that the model underlying it and in which it is employed is not useful or predictive or accurately representative of how economies work.

"Nick will be the first one to tell you that demand doesn't equal GDP. GDP for a period tells us, roughly, where the demand and supply curves met during that period."

Doesn't prove the equation "aggregate demand = GDP + change in debt" or other versions such as "aggregate demand is income plus the change in debt" is right.

Go to Grasselli's blog linked here and find that the equation is gone. See his last equation.

"aggregate demand = aggregate income"

I won't write it that way but still it is different from what the original supposedly revolutionary equation was.

It has been 18 months since Keen has presented it as something revolutionary which even Post-Keynesians haven't understood.

He owes a public apology to Paul Krugman for this.

Funny thing is how simple things like purchase of computers is confused around here by none other than Keen himself.

To do theory in physics one needs basic mathematics and not confuse things like the derivative of exponential is the exponential itself, derivative of a constant is zero and so on.

Similarly if your interest is writing accounting models of the economy (all concepts, debt, income etc are accounting concepts) you don't go around and muddle simple stuff.

Sorry to say it this way. Something's definitely wrong.

And there are two things here: empirical and theoretical.

If you want to say something empirical - you may want to refer to this nice Bank of England paper:


"Financial Stability Paper No 10: Growing Fragilities? Balance Sheets in the Great Moderation"

it uses concepts such as breaking net lending of a sector into net accumulation of financial assets and net incurrence of liabilities and its importance (and refers to work of PKEists) and is far more detailed.

From a theoretical viewpoint, one needs to write down behavioural equations but it is critical to start with self-consistency. The bad thing about accounting models is that great care is needed to achieve self-consistency and simple errors can lead to totally strange results.

And so while aggregate demand and debt are important, you don't really need slogans such as "aggregate demand = gdp plus change in debt" to push this point.

The growth before the crisis was led by private expenditure rising faster than private income and that things such as gross borrowing etc are as important as net borrowing but it is just that these things have to be looked at more carefully.

So entrepreneurs may make production plans based on sales and sales trends and decide how much to borrow and by what means. They will also need to decide how many to hire and so on.... This itself has an effect on demand and we have an income-expenditure model. And detailed models of such kinds already exist.

So let's not pretend to invent the wheel.

And nothing of that means "aggregate demand is gdp plus change in debt" which has anyway changed but maybe not.

So let's first get simple things right. This is because the controversies have existed since more than a year on such simple matters and repeated insistence from the other side and its supposed supremacy.

Nick Rowe has sidestepped these issues and taken a different route - which is good. I however see the same old issues appearing in the comments section by others (not initiated by NR)- as if these matters are crucial to get right before moving on to where Nick Rowe wants the discussion to happen.

My own attitude is similar but it's just that others who are insistent that their points about some accounting identities are right and crucial are themselves wrong.

Steve Roth,

I don’t think aggregate demand is the same thing as GDP, thanks. I think that it’s some sort of function that can produce different predicted values for GDP at a point in time, depending on the inputs to the function. That seems parallel to me to Nick’s distinction between supply and demand versus quantity supplied and quantity demanded. Supply and demand are both functions. But the intersection is a realized value or a predicted value based on those functions. So AD can be interpreted ex ante and ex post as the intersection of a bunch of influences. And I think ex post it should be the same thing as GDP as the accurately predicted value for the function. Which is why I think it’s a bit redundant to look back and try and figure out what AD was at some time in the past. The realized result is what it is – apart from that, you’re trying to figure out what the form of the function was in the past and what all its conceivable values might have been using different inputs at the time.

But the ex post view is not the only view. Which is why AD is not necessarily “the same thing as GDP”. That doesn’t contradict the idea that its ex post realized value if correct should be the same thing as GDP. And that should hold in continuous time as well.

So whatever the form of the function is, it shouldn’t include GDP as a sub-expression within it – not unless it’s written in lagged time regression form with prior rather than current values of GDP as part of the argument along with changes in debt, etc. And such a time lag should hold even if written in continuous time, because otherwise I do think you have a contradiction when you move to continuous double entry accounting along with that (which is the only consistent thing to do from a measurement perspective).

With continuous double entry accounting, realized AD should equal GDP continuously, so the “correct” value of the equation as a continuous predictor should be capable of generating that. And it can’t do that if the instantaneous value of GDP is augmented with other things in attempting to predict instantaneous AD.

The financial system can always be represented as a construction of double entry accounting, and nothing changes without being able to record it as a change in accounts, and those accounts must balance, and that has to hold in continuous time as well. This should be recognized as a constraint governing the viable form of behavioral economic equations, which is what an AD equation is I think.
The expression AD = GDP + debt is at its core a behavioral one. It’s potentially powerful but very crude. It won’t take flight in continuous time, because the accounting also becomes continuous in order for the entire measurement framework to be on consistent footing. I think that’s a point that was missed in the (earlier) version. An accurate predicted value for instantaneous AD shouldn’t be greater than the actual realized value for instantaneous GDP – according to the accounting recognition of behavior.

Have a look at Godley Lavoie again. There is no question from their perspective – no question – that coherent accounting is a full constraint on the viable form of expressions that can feasibly represent coherent economic behavior. That’s what I believe now, and did before even seeing the book.

Continuously yours …


Steve Roth,

In economics profession, there exists a tendency to confuse accounting identities with causalities.

It is not surprising that Paul Krugman was one of the first (but not the first) to catch Keen on this point because Krugman understands this confusion. He himself may probably make it in other places and that's not inconsistent with the previous sentence.

In this case however the story is strange - you want a causality and insist an accounting identity is right. You don't really need such a thing to tell a story of how debt is important for aggregate demand.

Sometimes it's a good thing to wait before replying; I like the turn the comments have taken recently, since they appear to have turned back (a) to Nick's original contribution in this post of providing an expression in which rising debt adds to demand which (b) is consistent with accounting, since it must be and (c) at the same time we are attempting to derive economic principles and should not confuse "accounting identities with causalities".

[As an aside, I argue that Krugman does do that to argue that since assets = liabilities, the level of and change in private debt doesn't matter (outside a liquidity trap), but equally he is also at times quite aware of the mistake of confusing accounting identities with causalities, as he pointed out very well in his Cambridge lecture on the 75th anniversary of the General theory.]

On JKH's continuous time points, the propositions you put there underlie how the monetary "Godley Table" component of the Minsky system dynamics program is designed (and I hereby admit that I learnt a lot about the importance of double-entry bookkeeping in designing the program that I didn't appreciate before I started). Those two extremely simple example files I've linked to above differ only in where loans originate--on the liability side in the "LFvEM" model and on the asset side in the "EMvLF" model--and the continuous time dynamic equations can be derived directly from the program. Both models are 100% accounting consistent--they have to be, otherwise the program flags an imbalance--and in one changing debt levels don't have any macroeconomic impact, while in the other they do.

I think they're the simplest possible statement of the difference between Loanable Funds and Endogenous Money that one can make in an explicitly monetary model. I should be able to spend some time analyzing those equations as part of the course I'm giving down here in Quito, and I'll post a link to them when I've done so.

Nick's post said: "So in my view:
A bank buying an IOU
A bank buying a computer
A bank buying a meal at a restaurant for its staff to celebrate Christmas
A bank giving money to charity
are all the same, in terms of creating money, and their effects on the liabilities side, though they will have very different effects on the asset side."

I think there is a difference.

IOU (I'm going to call that the loan)
probably less than 100% capital requirement

computer, meal, and charity
probably 100% capital requirement

If the capital requirement is 100%, then the bank can’t expand the “money” supply overall.

Nick, I put in an email address that is not mine. I'm hoping I will not be put in spam now.

ProfSteveKeen, I believe this link is relevant here.



Nick Rowe seems to be saying that this only matters in "the "very short run". See his last comment above.

Philippe: I did say that. But I called it "the very short run" only in relation to what the ISLM model calls "the short run". If the economy never hits the ISLM "short run", that's a rather academic definition.

I think I would say it like this: this very short run process lasts until the central bank wants it to stop and makes it stop. And it won't make it stop until Y and/or P rises to the central bank's target, which is how we might define the "long run".

Yes Philippe, that does seem to be what Nick is saying--that this applies in the very short run (though with a caveat as to whether the economy is ever in the long run).

My take of course is that "in the long run we are still in the short run"--that the economy is always in disequilibrium and we therefore have to model it that way. Nick picked up on that in his rephrasing of my effective demand argument of course, and I was very pleased by that--a lot of commentators inadvertently shoehorn my views into an equilibrium framework for which they were explicitly not designed, and that makes it harder to argue my case.

Incidentally Nick, there's one key Hicks paper on IS-LM and equilibrium that I've gone blue in the face trying to get certain authors to read (hint: one of them has the initials PK). But now that we're having a delightfully civil exchange here, can I ask whether you've seen it?:

Hicks, J. (1981). "IS-LM: An Explanation." Journal of Post Keynesian Economics 3(2): 139-154.

If not, I highly recommend reading it on this topic--of whether IS-LM can be used to model the economy at times like 2007/08 when it clearly was not in equilibrium. Here's Hicks's observation on this in relation to his perceived previous disequilibrium year of 1975:

"We know that in 1975 the system was not in equilibrium. There were plans which failed to be carried through as intended; there were surprises. We have to suppose that, for the purpose of the analysis on which we are engaged, these things do not matter. It is sufficient to treat the economy, as it actually was in the year in question, as if it were in equilibrium. Or, what is perhaps equivalent, it is permissible to regard the departures from equilibrium, which we admit to have existed, as being random. There are plenty of instances in applied economics, not only in the application of IS-LM analysis, where we are accustomed to permitting ourselves this way out. But it is dangerous. Though there may well have been some periods of history, some "years," for which it is quite acceptable, it is just at the turning points, at the most interesting "years," where it is hardest to accept it."

Steve: I don't remember reading that paper before. I have read it now. My reactions:

1. Yes. The sort of issues he is getting at there are very similar to what we are getting at here. Uncanny resemblance about very short run vs short run, or weeks (his "weeks" are my "months") vs "years" (ISLM). And the stuff about surprises and plans not being fulfilled. Uncanny resemblances.

2. He makes a real dogs breakfast over the labour market/product market distinction. Why not just assume self-employed hairdressers, where W=P, and L=Y, to keep this stuff simple, so we can concentrate on the difficult stuff.

3. He is very fuzzy on "equilibrium". There is: Qd=Q, Qs=Q, Q=Qexpected. In full equilibrium all 3 = hold true. ISLM assumes Qd=Q and Q=Qexpected for the output market and IS curve. I am dropping Q=Qexpected.

4. Walras' Law just does not work for a monetary exchange economy. He talks about "the money market", which is nonsense, since all markets are markets for the medium of exchange. His "money" isn't really money. My old post here.

Despite ALL that, it's still a good paper. He was a great economist, of course.

"I think I would say it like this: this very short run process lasts until the central bank wants it to stop and makes it stop."

From what I've seen, central banks don't appear to be particularly good at controlling asset-based debt bubbles by mucking around with interest rates. The 2007/8 financial crash is a good example. What did raising the Fed Funds rate achieve? It made lots of debts unpayable and probably helped crash the economy. Is there any other benefit that you can think of?

Yes, I agree it's messy Nick--there are a lot of points I would reject from a complex systems point of view alone. But the wisdom on equilibrium is very valuable. I argue--I'm arguing in my next book anyway--that economists have tended to mistake equilibria for identities, and therefore to insist on equilibrium outcomes when in fact identities (such as supply == demand in every market) apply in or out of equilibrium.

Anyway, I'm very glad that you've read that paper. Please recommend it to some friends!

Cheers, Steve

@JKH: "ex post [aggreagate demand] should be the same thing as GDP"

So looking back, we could not say "demand was higher than the quantity supplied"? Because of supply constraints for instance? People woulda bought more if they coulda?

If I'm remembering correctly, I have seen Nick Rowe, Scott Sumner, and others say for every borrower there is a lender. That means if the borrower repays the loan, the lender should be able to spend. Is that right?

In general Too Much Fed, no it's not. If the lender is a non-bank, then the repayment of a debt lets the lender spend because both debt and loan are on the liability side of the banking system's ledger; but if the lender is a bank, then the repayment of the loan takes money out of circulation (I prefer that expression to "destroys money") because the debt is on the asset side of the ledger. That's the essential difference between Loanable Funds & Endogenous Money, which I'm trying to illustrate in a pair of very simple models that I'll post on my blog shortly--and link to Nick's discussion here.

ProfSteveKeen, is this right?

non-bank: all assets have a 100% capital requirement

bank: some assets have a less than 100% capital requirement

When I see loanable funds, I think a 100% capital requirement.

TMF: No, that's wrong. Capital requirements are determined by regulators and by the prudence of the lenders themselves, depending on the riskiness of their loans. They may be the same or different for banks and non-banks. Loanable funds does not assume 100% capital requirement.

Let me try it this way.

A non-bank is an entity that either can't use leverage or chooses not to. It has an effective capital requirement of 100% for all assets.

A bank or bank-like entity can use leverage and chooses to. Some assets have a less than 100% capital requirement.

"Loanable funds does not assume 100% capital requirement."

I am pretty sure that would be worth exploring in more detail. For example, Bill Mitchell describing Mankiw's version of the gov't borrowing and scarce savings.

TMF: No. I used leverage when I got a mortgage to buy my house, but I am not a bank.

I don't think that Bill Mitchell would say that the theory of loanable funds assumes a 100% capital requirement. If he did, he was wrong.

Here is a post by Bill Mitchell on loanable funds. See Q1.


What do you think of it about loanable funds?

Here is another post by Bill Mitchell on loanable funds. See Q1.


Nick said: "I used leverage when I got a mortgage to buy my house, but I am not a bank."

Let's keep going with that. Nick wants to buy a house for $100,000. He saves $10,000 for a 10% down payment. He needs to borrow $90,000. I save $9,000. Nick will fail to buy the house for $100,000 if he borrows directly from me.

Instead, I start a new bank with $9,000 so that the equity of the new bank is $9,000. Assuming a 10% capital requirement for mortgages, Nick can now borrow $90,000 from the new bank.

How should that difference be described? Is there such a thing as "money" leverage?

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad