« Signs of population aging | Main | Should new editions of intermediate microeconomics textbooks be banned? »


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

This is a great teaching post and covers a lot of ground.

Here are two comments:

1) You are using a static model to look at dynamic adjustments.

I.e. "if the central bank changes r to r'". And I'm not sure how you can do that. For example, suppose that at the current level of savings, the interest rate that maintains price stability is 4%. But really you mean a total return of 4% -- a rational person will look at total return. That total return can be obtained by a 1% interest rate and capital gains of 3%, or a 4% interest rate and capital gains of zero. The capital gains can be obtained by the interest rate being cut.

So a time path of cutting rates can give a time path of total returns so that price stability is maintained. At least until you hit the zero bound. Cutting rates in the dynamic model can be the same as the static model with a rate increase.

Similarly, a time path of raising rates can cause negative total returns and be equivalent, in the static model, to a rate cut.

So there is a inconsistency when thinking of the static problem versus the dynamic problem. But you are using the solution of the static problem to analyze changes in rates or savings demands. I find this really troubling.

2) Why is the LRAS curve vertical? Many east nations have adopted development policies that, at least over near to medium term, greatly increased the LRAS curve as a result of keeping nominal interest rates low. Are you really saying that the LRAS curve is independent of nominal rates? So a nation can set nominal rates to, say, the natural rate + 4%, and this nation will have the same long run output as a nation that sets the nominal rate to the natural rare -4% ? I think it more likely that this nation will have a permanently lower capital stock. Changes in nominal rates are going to affect total output even in a two good model, because total output is aggregated by a price deflator, and the relative prices of capital/consumption goods are going to change so that total output changes.

I think you need a one good model to conclude that total output is independent of inflation.

rsj: Thanks! (That's really what i was trying to do.)

"1) You are using a static model to look at dynamic adjustments."

If we assume certainty, and that the Bank of Canada always gets it right, no problem. Just lend me an n-dimensional chalkboard for the n-dimensional Irving Fisher diagram (or a tame mathematician) and I can solve for the whole time-path.

But if the Bank of Canada ever misses, and is expected to maybe miss in future, that opens up a whole can of monetary worms. Including upward-sloping IS curves. Which is one of the reasons I want us to move away from thinking of monetary policy as setting a rate of interest, even in the short run.

"2) Why is the LRAS curve vertical?"

Actually, if we draw the LRAS curve in {r,Y} space, it probably slopes up. At a higher real interest rate, we want to consume fewer apples this year and more apples next year. Similarly, at a higher real interest rate, we want to consume less leisure this year and more leisure next year. (Some RBC models work on this channel). (I ducked that to keep the pictures simple and standard).

If we draw the LRAS curve in {inflation, Y} space, it will *probably* slope backwards. Because higher inflation screws things up, if there are sticky prices, that people don't want to change, so output supplied falls. Or it *might* slope up, if there is absolute downward sticky nominal wages and/or prices. If you let me rig all the assumptions, I can build a model to make it vertical. But the game isn't worth the candle. Empirically, it *seems* to slope backwards at very high inflation rates. But at least part of that will be reverse-causation (screwed up policies cause low Y and also high inflation). The data are just too dirty. All we've got is theory, plus some indirect data. Too much endogeneity. "Roughly vertical" is roughly defensible.

Re: 1, no that's not what I mean. The central bank cannot get things right because it can only use the short rate where total teturn is the rate. But investment is long term and savings is also long term. If in the long term it wants rates to be higher, in the short term, total return will be lower. The curves depend on total return. Therefore it is impossible to escape a period of time when the CB makes things worse and not better when it wants rates to go up or down. Then you need to measure the length of this time period versus whatever price stickiness you have.


1, 3 and 4 are all the same - I find it very hard to imagine why there would be a subjective time preference that would be in disequilibrium with the marginal product of capital. If the rate of interest does not adjust, employment/output adjust to create the equilibrium. The equilibrium condition of S&I is not a very interesting phenomenon in and of itself, or, the IS curve is most probably horizontal.

You're right to recognise the challenges to loanable funds in an open economy. But I tend to think that this is a relatively simply conclusion from your assumptions - a world where real exchange rates may not be constant is impossible to reconcile with a world where real interest rates are the same. But there's a different challenge, which does not require liquidity preference or an open economy.

Your conception of loanable funds is one of saving - it is the standard conception. But loanable funds, properly specified, is not about saving, it is about financing. Financing is the same as saving only if the credit equations in the economy do not change. As Leijonhufvud details painstakingly (and Claudio Borio does for the modern era and audience), there are numerous processes - financial deepening being the most obvious one, varying the percentage of retained earnings by firms being another - which ends this equivalence.

So even if you have a 'pure' loanable funds theory of interest, it does not necessarily follow that the S-I equilibrium will always be a full-employment one. Money is not a veil, but neither is finance.

My favourite way of thinking about 'the rate of interest' is to imagine a linear dial that says :

time preference/ MEC (income or accumulated flows) ----(LF)----financing (cash flows, instantaneous flow)----(LP)---- portfolio allocation (asses, or stocks).

Loanable funds (LF) sits between time preference and financing. Liquidity preference (LP) sits between financing and portfolio allocation.

The natural rate is given by an interaction of financing and time preference/MEC, or LF. (E.g. a period of easy financing - which does not need to correspond to a period of excess (desired) savings - tends to drive down the MEC/ natural rate without any 'real shock' whatsoever. When asset prices have been high, businesses need even higher prices to invest further.)

The market rate is given by an interaction between LP and LF. Whether the two are in consonance is dependent on the shape of the LM curve.

The standard neo-classical method has been to begin with the natural rate and assume that the market rate must be bounded by this plus transaction costs. As David Glasner shows, this is what Earl Thompson did - his model is basically your option 3. This is the kind of model that leads to the capital controversy debates - about the heterogeneity/ costs of switching capital etc. I believe that we don't really need such a model anymore.

The most satisfying neo-classical model, the benchmark case of showing the 'equivalence' between LF and LP, is the Fischer Black model. He took Fisher's capital theory in its entirety (all input was capital, and human capital was quantitatively the more important type) and took it to its logical conclusion - he collapsed the usual economist's distinction between quantities and prices. Thus, we move focus from the incomprehensible heterogeneity of assets of the economy to the more systematic heterogeneity of its liabilities, because there is an elaborate system in the market economy that tries to price this heterogeneity. Capital is wealth, wealth is capital. When asset prices fluctuate, capital fluctuates.

I like to imagine this as a large, strong bridge being put together between time preference and portfolio optimization, parallel to my linear dial. While early neo-classical theories had a variety of just-so stories that assumed this bridge (with Arrow-Debreu being the crowning moment in the just-so-ness of these stories), Fischer Black is the one who showed what that bridge actually was. The only GE that even remotely makes sense is the CAPM.

(As an aside, I often like to imagine that a very strong pull along this bridge sometimes causes the initial dial to crack open at the cash flow junction. This is 'the Minsky moment'.)

Throughout this, my conception of 'the rate of interest' is the 'cost of capital', not the short rate. That's also the rate of interest in the axis of my IS and LM curves. The short rate is a parameter, a tool in the hands of the monetary authority that could potentially set and move the cost of capital, but sometimes fails to do so.

rsj: As an extreme case of the problem you are talking about, suppose there's a lag in investment decisions. We decide this year on how much investment we will do next year, and that depends on this year's expectation of next year's interest rate. So the BoC must set next year's interest rate this year. It's gotta promise (or lead people to expect) it will set the right interest rate next year, when the BoC doesn't even know what people are currently planning for next year.

No problem in a world of certainty, because the BoC can figure out what people are planning to invest next year, and can promise today to set interest rates next year to keep I=S next year. But yes!

Ritwik: I'm afraid you really lost me there.

Let's take an extreme case. Pure consumption economy where output of the consumption good is exogenous, so the PPF is reverse-L-shaped. You can define an equilibrium interest rate in the Irving Fisher diagram, (the slope of the I-curve at the endowment point) even though there's no capital.

Sure, it's a weird case, but the Irving Fisher diagram can handle it. How would Fischer Black handle it?

Ritwik: "The only GE that even remotely makes sense is the CAPM."

And you really lost me there. I would say that CAPM is partial equilibrium. It's not even a theory of the rate of interest. It's a theory of interest rate differentials that takes the rate of interest on safe assets as exogenous.

It's gotta promise (or lead people to expect) it will set the right interest rate next year, when the BoC doesn't even know what people are currently planning for next year.

This requires time inconsistency on the part of agents. A long term rate will only fall if it assumed that a sequence of short term rates fall. So to credibly get long term rates to fall, the CB needs to promise low short term rates for a prolonged period. But if, in so doing, it shortens the recession, then you are requiring either false expectations on the part of investors -- e.g. tricking them into believing that short term rates will be lower for the next 10 years, even though we exit the recession in one year -- or you are promising short term rates to be too low for the next 10 years, veering away from your inflation target.

But you cannot both adhere to the inflation target and respond to fluctuations in savings demand.

rsj: you have lost me.

Let D1(r1,r2,r3,...rn)=0 mean "excess demand for output in period 1, which is some function of short term interest rates for the next n periods, =0"

Also D2(r2,r3,r4,...rn+1)=0

And D3(r3,r4,.....rn+2)=0

Etc., from now until the end of time.

So we are looking for some path for interest rates r1,r2,r3....etc. that solves that system of equations.

"Loanable Funds theory"

Are you saying an economy is 100% Loanable funds?

TMF: "Are you saying an economy is 100% Loanable funds?"

No. I don't even know what that means.


Yes, and I am saying that there is no solution.


Would it be correct to say that the liquidity preference theory refers to the secondary market for bonds, while the loanable funds theory refers to the primary market for bonds?

rsj: "Yes, and I am saying that there is no solution."

Never? Sometimes? A corner solution? Discontinuous PPF or I curve? Or not convex/concave?

JoeMac: ("Primary" presumably means "newly-issued, to finance new investment"?) Interesting twist. But I don't think so. What about the market for output, where newly-produced output is exchanged for money? If LF is new bonds, and LP is old bonds, what about the output market?

I.e. the set of solutions {r_j} that clears excess demand for D1 intersected with the set of solutions {r_j} that clears demand for D2 intersect with the set that clears D_3, etc. can be the empty set.

This is because the rate that sets D_J = 0 is a function of the average of rates, as we are assuming that investment is a function of long term rates.

To see a simple example, let n = 2. Suppose you need the average to be 1% in period 0, 2, 4, .. And you need the average to be 3% in period 1, 3, 5, ..

Then you have:

(1) r_ 0 + r_1 = .02
(2) r_1 + r_2 = .06
(3) r_2 + r_3 = .02


From (1) we know that

0 <= r_0 <= .02
0 <= r_1 <= .02

From (3) we know that

0 <= r_2 <= .02
0 <= r_3 <= .02

Etc. So we know that all the terms are less than .02 and greater than 0.

But then (2) is impossible to satisfy.

It might be worth putting in giant <blink> tags that (1) if you believe in endogenous money, and (2) that therefore money demand and supply depend on some third factor that itself messes with the real interest rate, then (3) the very first diagram may not represent the locus of points which represent consistent Md, Ms, and r for your pet outlook!

And hopefully that quietens the horde...

rsj: OK. I think I understand. Suppose Summer investment is very productive, and Winter investment is very unproductive. (A period lasts 6 months, and there are only two seasons). And an investment produces output for the following 2 periods. Would something like that work as an example of what you are saying?

A couple of thoughts:

1. I think we might need to look at geometric(?) averages of interest rates?

2. We might get a corner solution with zero (gross) investment in Winter, where all Winter output is consumed.

I might be getting the intuition wrong, but I think rsj's identified dynamic inconsistency can be suppressed if the CB conducted price-level path targeting instead. The inconsistency seems to stem from the CB needing to adjust their interest rate 'again' after their short-term policy has induced inflation consistent with their long-run target.

Nick's post said: "TMF: "Are you saying an economy is 100% Loanable funds?"

No. I don't even know what that means."

A corporation wants to issue new bonds. Borrowing medium of exchange from an individual saver is what I consider 100% Loanable funds. Borrowing medium of exchange from a bank is NOT what I consider 100% Loanable funds.

Does anybody else get that?

Okay, I've probably gotten the intuition wrong. Time to get some paper out.

david: What's a "blink tag"? (I had to edit my comment, because it came out weird when i put those arrow thingies in??)

TMF: re-read the post where I discuss the relation between the Loanable Funds and Liquidity Preference theories.

rsj: another example: Suppose a nuclear power station, once up and running, keeps on cranking out the same amount of electricity day and night, even though the demand falls at night. Night-time electricity might become a free good, if people are satiated. The 12-hour real interest rate (taking electricity as the numeraire) goes from plus to minus infinity.


My understanding is that the IS curve assumes as a default that anything borrowed is automatically spent.

For example, a firm goes into loanable funds and borrows money, and then goes into output market and buys good. When we speak of the IS curve, we first use the loanable funds to explain where it comes from, and then define the curve as being the goods market. It therefore conflates the two processes of the firm. When explaining the IS curve economists do the following. First they show the loanable funds market, then they show how changes there affect the Keynesian cross (output market), and then how that becomes the IS curve.

Technically the loanable funds (primary market) is separate from the goods market. But the IS curve implicitly assumes that any endogenous or exogenous borrowing by firms/households in the loanable funds will be 100% automatically and immediately spent in the goods market as the cross shows. After all, the textbooks always show an immediate jump from loanable funds directly to the KC.

So, when you say "There is an output market where output exchanges for money, and a bond market where bonds exchange for money." the output market includes the primary bond market and market for goods because it assumes borrowings are always spent on goods.

Or is this all nonsense?

joeMac: what you are saying is not nonsense. But that assumption implicitly underlying the IS curve is deeply problematic.

We can all plan to spend more than our income, not by borrowing, but by planning to reduce our stocks of money. Or we can all plan to spend less than our income, not to lend the excess, but planning to increase our stocks of money. And that's what causes recessions. We can also all plan to borrow, not to spend, but to increase our stocks of money.

Hehehehehehehehe. M........M.......

We might not be on the same page. I'm thinking of Loanable funds where people claim borrowing depletes scarce savings and therefore raises the interest rate(s). Is that what you consider the same thing?

Plus, what about borrowing to speculate in financial assets?

Lastly and in the third graph, what happens if the stock of money (I'd rather say medium of exchange) goes down?


I have to think about your summer-winter example, but my initial example is "yes, assuming that firms enter into fixed rate nominal debt contracts at the 2 period rate". And I am thinking that that 2 period rate is the average of the single period rates. I am also assuming perfect foresight, so there is no unanticipated shock or bankruptcy, etc.

If single period volatility is smoothed away in the rolling 2 period average, then the 2 period rate will not be sufficiently responsive to short run changes in investment/savings curves. Over the short run, interest rates are not going to clear the I-S market, but income adjustments will need to clear it. The fact that the rates are an average means that they will always be too high or too low, regardless of how you define average (e.g. weighted arithmetic mean, geometric mean, etc.).

So in your nuclear power plant example, the plant is obligated to make fixed nominal payments out of its profits day and night. That poses a problem if electricity is free at night.

Whereas the social planner might make electricity a free good, the bond markets will not finance the construction of the plant unless it agrees to throw some output away or produce less so that night time sale of electricity earns the same profit as daytime sale of electricity.


Just one more question. Are you saying...

1). No, JoeMac, you do not understand IS curve. It is not making that assumption that you describe.

2) Yes, you are absolutely correct. That's what the IS curve does. But that is a metaphysical problem for reasons X, Y, Z.

Ritwik writes:

1, 3 and 4 are all the same - I find it very hard to imagine why there would be a subjective time preference that would be in disequilibrium with the marginal product of capital.

#3 and #4 are presented by Bohm-Bawerk in Capital and Interest as two of the three determinates of the interest rate (the third being the uncertainty of the future), and Fisher took the position that his (Fisher's) Theory of Interest was largely a restatement of the earlier work of Bohm-Bawerk in mathematical terms.

I'm having trouble grasping your point Ritwik and suggest you read: http://www.econlib.org/library/BohmBawerk/bbPTC34.html#Book V,Ch.V

Though perhaps well you need to start earlier in the treatise to have any hope of understanding the phrases employed--being as they are archaic and roundabout to the modern economist...


1) Not sure what you mean by 'pure consumption economy where the output of the consumption good is exogenous', but a Fischer Black economy with no physical capital still has capital - people are doing the producing and all capital is human capital. (Such an economy might be a pure services economy? )The rate of interest is simply the rate of return on that human capital, the capital value is implied in the asset values of the firm(s) in this economy and Y=rK then gives the rate of interest.

A reverse L shaped PPF reduces the interest rate to the subjective time preference - this time preference gives the rate of interest (and hence capital values). The big difference (and this is why 'finance' talks past 'economics') is that you have to change your frame of reference and *begin* with the capital values rather than from inter-temporal consumption optimization.

Plus, I am not even sure what the thought experiment of a one-good consumption economy helps us achieve - ostensibly we want a theory of interest in a multi-good multi-asset monetary economy?

2) CAPM is a temporary security price equilibrium, but it can be easily re-framed as a general equilibrium - with more risk-bearing societies being able to generate higher rates of growth. If the only issue is the exogeneity of the risk-free rate to private decision making, here's my (rather speculative) take on it - imagine a world where the CB did not even bother setting the short rate, but simply announced its price path target/ NGDP target and was perfectly credible. Let's say that the CB has somehow managed to make money neutral/ super-neutral. This economy would devise a system of repo loans, interest rate swaps, credit default swaps (or other means of splitting out the economic activity of funding from the economic activity of risk taking) and the rate on repo (or other such pure funding) loans would give you the risk-free rate, endogenously. This is my interpretation of the idealized Fischer Black world.

I really think Mehrling's takes on Black and CAPM are indispensable here, and would request you to read them , if only to ensure that I am not mangling his message.



Additionally, John Geneakopolos and Martin Shubik have re-formulated the CAPM as a GE with incomplete markets on multiple occasions, but I am not invoking them here because I am not sure how they handled the exogeneity of the risk-free rate.


Say there are (as in Bohm-Bawerk) 3 determinants - time preference, MEC, uncertainty. In my understanding, Irving Fisher modelled the interaction of time preference and MEC, by first abstracting out of uncertainty, and then next including uncertainty but drawing a sharp wedge between capital quantities ('real capital') and capital values (asset prices), so that the uncertainty element of interest rates feeds only into prices but not into production.

Let's say we turn this around - we remove the sharp wedge between capital quantities and prices, see uncertainty as governing the production process itself (rather than just asset prices), and see time preference/ MEC as two ways of stating the same economic fact (and hence always in 'equilibrium'). Now we've arrived at Fisher Black (or so I think).

Oh, and we add that all input is capital, and that there is no better way of understanding the *aggregate* production process except through asset prices. Then we truly arrive at Fisher Black.

I guess I should acknowledge that in a perfect foresight model it wouldn't make sense for firms to fund themselves by selling longer period bonds, everyone would fund long term investments by rolling over commercial paper, and no saver would buy fixed rate multi-period bonds, they would all buy short term commercial paper, etc.

Understanding why firms sell long term debt to fund long term investment requires a model with risk, in that firms value knowing what they must pay ahead of time, and this value outweighs (to the individual firm) the possibility that ex-post, they will always be paying more or less than the market clearing rate.

But if we just accept that this is how investments are funded, and if we also accept short run variability in both demand/supply then the above argument should show that in general there is not going to exist a sequence of short term interest rates whose rolling average, however defined, will be exactly the sequence of long term rates that clears investment and savings demands in every period.

I *think* this is a good case for the Keynesian view that income adjustments are necessary to clear the savings and investment markets, at least for short run variability in investment/savings demands.

Excellent post!

JoeMac: 1.9 ;-)

Update: what I mean is: there is probably more than one way to rationalise the IS curve. Yours works, but I don't like it. But maybe all other ways are equally bad.

rsj: people invest in sheep, as long as the sum of the values of wool and mutton exceed the costs of the sheep. But the wool dividend might be worthless. Chickens pay dividends in eggs, meat, and feathers. The feathers might be worth nothing, but the market in chickens can still work.

Switching to general equilibrium, and from different commodities to different dates (all the same in an Arrow-Debreu dated-commodities approach): We can imagine a world where food is the only output, and food is so plentiful in Summer it becomes a free good. You could say there is an excess supply of output in Summer, but it's not a "Keynesian" excess supply problem. The Keynesian excess supply problem is where the unemployed want to eat, but can't buy food. Here, it's satiation.

Let me try it another way: you have specified the "investment" part of the loanable funds model, but you haven't specified the "saving" part. If we add in the saving part, we could only get your result (of no solution) if people are satiated in consumption during some periods.

Determinant: Are you OK mate?

TMF: Look at the first picture. If the Id curve shifts right, the rate of interest will rise (unless the Sd curve is horizontal).

If I borrow money from another person to buy assets from a third person, that third person might just buy my IOU from the second person. So it's a wash on net.

It depends on what causes the stock of money to fall; is it a fall in supply or a fall in demand?. If the supply of money falls, the LM curve shifts left. If the demand for money falls, the LM curve shifts right.


This is a very good post.

Although, given your enumeration of "the joys of national income accounting", I’m tempted to ask – is this the same Nick Rowe who used to blog for Worthwhile Canadian Initiative?

“Which is one of the reasons I want us to move away from thinking of monetary policy as setting a rate of interest, even in the short run.”

If “the rate of interest” in question (“natural rate” etc.?) is viewed as a continuously changing rate for a continuously perpetual term, does monetary policy determine (implicitly) the term structure for that rate - as the combination of the policy rate (for some expected term) and a residual perpetual rate? And the policy rate is always overshooting or undershooting and changing everything else in the process as well?

Jon: as you will notice, I carefully avoided using the A-word. My reading of Austrians is like yours: same as Irving Fisher, plus uncertainty, minus the neat diagram. But some Austrians sometimes sound a bit more like 4.

Hmmm. I'm a bit surprised there's still no riot broken out. The LP and Cambridge UK guys must all be sleeping in. (Or maybe gathering their forces for a massive attack on two fronts.)

JKH: Thanks! I didn't really want to write it, but I thought it needed to be written. Someone maybe should have written something like this years ago, to stop a lot of unneccessary arguments, so we could concentrate instead on the bits we (at least I) still don't properly understand (like in my digression, which is where I depart from "mainstream").

Yep. I was surprised I wrote that bit about NIA too. I nearly censored myself. But it's a very "subjectivist" joy of NIA. NIA as a tool to help us tighten our intuition.

You lost me a little on the last bit. Let me take a stab at it:

Assume the central bank always gets it right, and everyone knows this. There is a "natural" 1-year rate, that is moving over time. And a natural 2-year rate, that is also moving over time. (And 3 and 4 etc.). If there were no uncertainty, the pure expectations hypothesis would imply that the 2-year rate would be a geometric(?) average of the two 1-year rates. Under uncertainty, and risk-aversion, that won't work exactly. The central bank can choose any one of those rate (say the 1-year), and try to keep it equal to the natural 1-year rate.

If the central bank gets it wrong, and is expected to maybe get it wrong in future, that will have real effects, and will affect investment, which will affect the future capital stock, which will affect future natural rates, which will affect the current natural long rate,...etc.


Let me play that back in terms of the relationship between natural and actual, and the feedback between the effect of the difference between the two on real activity, with further effect on natural rates.

There is an actual yield curve (and BTW an actual curve of implied forward rates for all terms based on that actual yield curve).

Suppose there is a natural yield curve (and BTW with a natural curve of implied forward rates for all terms based on that natural yield curve).

(The forward rates are a by product in both cases, not really necessary to the discussion.)

The CB sets an actual policy rate, say the ON rate.

That actual ON rate need not be the same as the natural ON rate.

And therefore there is nothing to say that any corresponding points between actual and natural yield curves or between any set of actual and natural forward rate curves be equal.

And deviations will have real effects with consequences for the natural curve itself, with knock on deviations between actual and natural, etc. etc.

IF the actual ON rate deviates from the natural ON rate, then the actual ON rate perturbs the natural yield curve into an actual yield curve which is not the same, and which has real effects with further consequences for the natural curve.

Would a natural curve have a slope?

Or would it be optimized to a uniform perpetual rate?


That is very similar to my favoured way of thinking about natural and market (actual) rates, which I specified in a comment on some other post by Nick. Suppose that we reduce the yield curve into only two assets - money and bonds. Now there are four rates of interest which we want to talk about : money-natural, money-actual, bond-natural, bond-actual.

The CB only sets money-actual. It could fail to move bond-actual. Or, it could fail to clear money-natural at the money-actual rate that it sets.

The most perverse developments happen when the CB 'manages' to make bond-actual clear with bond-natural, without making the money rates clear. We see macroeconomic stability, low unemployment, low inflation, stable NGDP etc. All this while, money-natural can drop a lot and diverge significantly from money-actual, creating a massive increase in money demand because of the simple arbitrage of investing in money-actual. Because the macroeconomy is stabilized, nobody realizes/ bothers about this money demand that is increasing in the background. A entire industry sustains itself on finding new ways to service this money demand. Then, one day, s*it hits the fan and everybody re-discovers their favourite theories of the recession - shortage of safe assets, negative real rates, money demand, NGDP shock etc.

The one big leading indicator in all of this is what Raghu Rajan had been warning about since about '05-06, the ongoing dearth of fixed investment/ hard assets.

JKH @9.48: Yep. Agreed.

JKH @9.56. In principle, the natural yield curve could slope either way, and change slope over time. E.g. demographic changes, which people see coming, will change the ratios of middle-aged savers to young and old dissavers, so the natural short rate will be expected to change over time, and the slope of the yield curve will change with it. Or it could be anything else, not just demographics. In general the (real) interest rate is a relative price of two goods (claims on future apples vs current apples), and relative prices will change over time when supply and demand (tastes, technology, resources) change over time. Only under very special assumptions will relative prices not change over time, and you get a flat (real) yield curve. In principle, there must be a monetary policy that would give you a flat *nominal* yield curve, since monetary policy can (in principle) make the price level do whatever it wants, so by getting inflation and expected inflation to change by just the right amount at just the right time, the BoC could covert a fluctuating sloping *real* yield curve into a flat *nominal* yield curve. (I'm not thinking Operation Twist here.)

Nick, what theory underlies the view that an asset's interest rate (or own-rate. ie. expected rate of price appreciation) is one component of the expected return that said asset must provide in order to be held? ... and that after adjusting for risk and liquidity, all asset yields equal out via arbitrage? Is this Fisher + liquidity preference?

JP: I see that as very different. Loanable funds is a theory of "the" interest rate. And it can be used to figure out a theory of the time-path of "the" interest rate. Same with Liquidity Preference, which is also a theory of "the" interest rate. For interest rate differentials across different assets, Arrow-Debreu implicitly contains a theory of all real interest rates. Any asset is just a bundle of Arrow-Debreu commodities. But since it's a world with zero transactions costs and complete markets, I'm not sure how far I would push it.


I'm sorry, I don't understand your sheep example. In general, the effect that I am talking about has nothing to do with corner cases, it is the general case. The CB is trying to solve an infinite tower of optimization problems, subject to the constraint that all but the first of the short rates used to solve each problem must also be used to solve the problem below it. When I say that there is no solution that simultaneously solves all problems, it is not because the solution to each individual optimization problem must correspond to a special solution of the underlying problem.

Think of it this way -- We are in a coffee-shop economy. And coffee shop owners will not build out a store if they can only secure a month to month lease. They prefer a 10 year lease, with fixed payments. Similarly, owners of land prefer to give a 10 year lease. If those rental payments correspond to the 10 year geometric average of gross rates, which is roughly the 10 year arithmetic average of rates, then in half the periods you would expect the short rate to be lower, and in half the periods the short rate is higher. When the central bank lowers the short rate, the capital rental rate remains higher, and when the central bank raises the short rate, the capital rental rate remains lower.

Now the question is, if the CB knows what the market clearing sequence of rental rates will be each period, can it target the short rate so that the rolling average is the pre-scribed sequence? The answer, in the general case, is no. Demand to rent coffee shop space may be low in even periods, and high in odd periods. But as 10 year leases are being signed, the rental payments will reflect the average demand, not the spot demand. To the spot market, it will appear as if the rent rates on offer are "stuck", in that they are not responding to the demand for spot rentals. In reality, the landlord would rather let the space sit empty and wait for one period until demand is higher, rather than get a lower payment for 10 years because they need to rent the space out now in order for the spot market to clear.

None of the above has anything to do with satiation, corner cases, discontinuous preferences, etc. This is an argument based on incomplete markets.

[Edit: rsj: maybe skip this comment, and read the next one. I may have misunderstood you.]

rsj: "Demand to rent coffee shop space may be low in even periods, and high in odd periods. But as 10 year leases are being signed, the rental payments will reflect the average demand, not the spot demand. To the spot market, it will appear as if the rent rates on offer are "stuck", in that they are not responding to the demand for spot rentals. In reality, the landlord would rather let the space sit empty and wait for one period until demand is higher, rather than get a lower payment for 10 years because they need to rent the space out now in order for the spot market to clear."

Let the "period" be 12 hours. Coffee shops sit empty at night, and are used during the day. The rent would have to be negative at night to keep coffee shops fully employed at night. The landlord doesn't want to collect a negative rent, so chooses to leave the coffee shop unemployed at night. But this isn't Keynesian unemployment. This really is a Zero VMP coffee shop. The unemployment of coffee shops at night is an optimal allocation of resources.

(The analogy with sheep is as follows: a coffee shop/sheep is an asset that produces two goods in fixed proportions: daytime space/mutton; and nighttime space/wool. The demand curve for nighttime space/wool intersects the supply curve at a price of zero (it would be negative if you weren't allowed unemployment/free disposal of nighttime space/wool. But we do allow unemployment/free disposal, so we get a corner solution, on the horizontal axis, where the coffee shop is unemployed at night/the wool is thrown away.)

I'm fine, Nick. Just being a devilish imp and nearly invoking those three letters: M, M, (*dare not say the last one*) that detonate threads.

Since I can't do maths for geometric averages, let me change your assumptions slightly. Suppose the landlord wants the payment in a lump sum up front, and that coffee shops last two years. The rental in odd years is worth $100, and the rental in even years is worth $200.

If the 1-year rate of interest were the same in both odd and even years, the PV of a new coffee shop in an odd year would be $100/(1+r) + $200/(1+r)^2, which will be less than the PV in an even year, which is $200/(1+r) + $100/(1+r)^2.

You would get exactly the same result (PV less in odd than in even) if the lease is 2 years, at $150 per year, but the 1-year rate is r0 in odd and re in even years, and re > ro

In odd years, PVo= $150/(1+ro) + $150/(1+ro)(1+re)
In even years, PVe= $150/(1+re) + $150/(1+re)(1+ro)

PVe>PVo because the second terms are the same, but the first terms are different.

rsj: I'm working on a better example.

Assume workers can only do one thing: build coffee shops. Constant returns technology, perfectly inelastic labour supply, fixed nominal wage, and at full employment they can build 100 shops per year at $200 each. Shops last 2 years then fall down. So at full employment there is a stock of 200 shops. Let the equilibrium rental for a shop (when there are 200 shops) be $100 in odd years and $150 in even years.

The PV of a new shop at the beginning of an odd year is:

PVo= $100/(1+ro) + $150/(1+ro)(1+re) = $200

The PV of a new shop at the beginning of an even year is:

PVe= $150/(1+re) + $100/(1+re)(1+ro) = $200

Those two equations are not linearly dependant, so there must exist a solution for ro and re.

The intuition is that the denominators in the second term is the same, so the BoC can't affect that by making r lower in odd years than in even. But the BoC can affect the denominators in the first term by making r lower in odd than in even years.


No, if someone is selling a 2 year bond, then ro = re.

That's the point.

Today, short rates are at zero, but long rates are not. Someone wanting to borrow long term must pay above zero rates. Why? Because in the future, short rates will be higher, and the long dated borrowing costs today reflect the anticipated higher short rates. Therefore even today, he must pay out a higher coupon, even though the short rate is zero. The borrower pays the same coupon every period. On average, half the time the coupon he pays is higher than the current short rate, and half the time it is lower (assuming pure EH). But he does not pay a lower coupon in one year than in another year. How much he pays for the next 10 years is decided by the long rates when he borrows.

Then you can ask, if I know that the market clearing sequence of long rates in each period is {L_1, L2, .. L_n,...}, can I select a series of short rates {r_1, .., r_n, ...} such that the average of the short rates is the long rate in every period. This is not a convex optimization problem, and there is no guarantee of a solution. The answer is, generally, no, it is not possible for the CB to have long rates be whatever it wants in every period.

It is only possible to do this if each of the long rates {L_1} are "close" together and do not wiggle too much (sorry for being too technical). If they wiggle a lot or are not close together, then there does not exist a sequence of short rates that will cause the long rates to be the ones demanded.

[Edit: this comment is wrong, as rsj points out below. I was muddled with the yield on a 2-year annuity. NR]

rsj: I had (implicitly) defined ro as the 1-year rate in odd periods, and re as the 1-year rate in even periods.

The 2-period rate in odd periods r2o is defined by; 1/(1+r2o) + 1/(1+r2o)^2 = 1/(1+ro) + 1/(1+ro)(1+re)

The 2-period rate in even periods r2e is defined by; 1/(1+r2e) + 1/(1+r2e)^2 = 1/(1+re) + 1/(1+re)(1+ro)

You can see that r2o and r2e are not the same, if ro and re are not the same.

Roughly speaking, the 1-year rate must wiggle twice as much over time as the 2-year rate, and 3 times as much as the 3-year rate.

It's too quiet. Eerily quiet.

I have just argued that modern monetary theory should assume loanable funds, because modern central banks set a rate of interest equal to the loanable funds rate. (OK, at least they try to, but we armchair economists don't know better than the BoC what the LF rate is, so it makes sense for us to assume it.)

Maybe it's Monday. Or maybe they are all plotting something....waiting for dark....


You can see that r2o and r2e are not the same, if ro and re are not the same.

I am assuming EH, in which case r2O and r2e would be the same:

Let re and ro be the short rates in even and odd years.

I can borrow $1 in the first (even) period, and owe (1+re), Then I roll that over again, and owe (1 + re)(1 + ro).

Therefore my 2 period rate, r2_e, is the solution of this problem:

(1 + r2_e)^2 = (1 + re)(1 + ro)

Taking logs, you get that the two period rate, borrowed in an even period is the average under the approximation that log(1+x) = x.

If go through the same argument, but starting in an odd year, I get

(1 + r2_o)^2 = (1 + ro)(1+re)

Because multiplication is commutative, the 2 period rate, borrowed in an odd period, is the same as the two period rate, borrowed in an even period.


Note that counting equations is not enough here to assume a solution exists, as the solution set must be positive, and in general, you have an infinite dimensional problem so knowing that there is an infinite number of unknowns and an infinite number of data points is not enough to know a solution exists. The infinite dimensional aspect goes away in our cyclic case, of course.

rsj: OK. Try my 4.26 comment.

Nick - thanks for this post. It's a great synthesis of the mainstream perspective on interest rates that I've been looking for, so that I can avoid going back to the text books :). I'd love to develop a perspective on this vs. post-Keynesian. I just don't know enough about both yet. Hopefully someone else can chime in on that. RSJ should be able to, or JKH.


"I have just argued ..."

There doesn’t need to be a conflict.

Whether or not the CB is targeting the natural rate with its actual rate, it always has the freedom to depart from the natural rate with its next policy decision – by intention or by error. It sets the actual rate either way. It could be argued that it is constantly groping toward the natural rate, even if it is not aware of that or resists it with short term policy changes. It is free to make errors because it is free to decide.

And the loanable funds framework fits OK provided it somehow incorporates a projection of future I and S and Y (rather than a current, static snapshot), which allows for economic/financial growth and corresponding dynamic paths for I and S over time. A static assumption would be objectionable.

I think it’s primarily a static assumption that your favorite (other) heterodox group objects to.

wh10: Thanks! I'm not sure if any textbook really does a synthesis? All the bits are there, but sort of spread out across macro texts and micro texts, that often discuss interest rate determination only in passing.

I think the *main* Post Keynesian approach would be Liquidity Preference, in *some* variant. But I think there's also a sense that interest rates are an instrument for determining the distribution of income between workers and capitalist/rentiers. Of course, there's a big range of views out there. I don't think there's one single view that you can say represents the main challenger, apart from *maybe* Liquidity Preference.

Nick, re the 4.26 comment:

There is no solution for positive rates r_e, r_o. Just try to solve the simultaneous set of equations.

Geometrically what you have with the equation:

$100/(1+ro) + $150/(1+ro)(1+re) = $200


100 = 150/(1+re) = 200(1+r0)


0.5 = .75/(1+x) = 1 + y


y = -0.5 + .75/(1+x)

For x > 0, you have one leg of a hyperbola. Only a finite portion of it lies above the x-axis (because of the negative -0.5 term).

Similarly, the other equation gives you one leg of a hyperbola, shifted away from this one.

And unless these two hyperbolas are very close -- which means that the two rental payments are close -- one will lie above the other in the upper right quadrant and they will not intersect for x> 0 and y > 0.

I tried to spell out the intuition for this (although badly, I'm sure) when I gave my original example.

rsj: you know I can't solve equations!

re will be positive. I don't know whether ro will be negative. It may be. Nothing rules out negative real rates. It depends on preferences and technology. The Irving Fisher diagram can tell us that. (Costless storage technology would rule out negative real rates, because you could mothball the shop for 1 year so it lasts 3 years).

Wait! I'm wrong. Let me do some checking..

Take the worst case scenario. Suppose the rent is $0 in the odd year, rather than $100. And $250 in the even year, rather than $150. Then (I think, unless I've screwed up the math, which I might have) the solution is: ro=0, and re= 0.25

But if we changed my example slightly, so that shops lasted 3 years instead of 2, then (I think) ro could be negative. Because if you built a shop in an odd year you could only get one even year's rent. But if you built a shop in an even year, you could get two even years' rent. Not sure.

But interesting thoughts, these.

Nick, yes, in a 2 period model you will have a solution with dividends provided one of the terms is not zero.

Since these are nominal rates, the zero bound is a hard bound -- I am assuming that the CB selects the short rates (which are nominal) in order to determine the market multi-period rate, which must then also be nominal. I have to think about this a bit more, as the dividend paying example should not destroy my zero coupon example. It should only add some complications.

Remember that the 6-month real interest rate on strawberries will be negative at times. Strawberry prices rise a lot from Summer to Winter, so the real rate will be negative. Then fall from Winter to Summer, so that real rate will be very high. As long as prices are flexible, and the CB doesn't try to target a price that needs to fluctuate relative to other prices, it can handle negative real rates. (I'm pretty sure real rates go negative in the evening, because most stuff costs more at night than in the day.)

Nick writes: "Jon: as you will notice, I carefully avoided using the A-word. My reading of Austrians is like yours: same as Irving Fisher, plus uncertainty, minus the neat diagram. But some Austrians sometimes sound a bit more like 4."

Yes, that is of course due to Mises's error (or dogma); he accepted only the subjective theory of time preferences as determining the rate of interest (and rejected Bohm-Bawerk). Although Mises never wrote even an essay on origins of the rate of interest, he does refer always to interest as a arising from time preference alone and lectured to that effect...

the Misesian theory of interest depends entirely on subjective time preference, with no influence attributed to physical productivity. ... More specifically, Mises' theory of capital and interest is in disagreement with Böhm-Bawerk's on the following points: ... b. On the role of productivity: As already mentioned, Mises sharply deplored the concessions Böhm-Bawerk made to the productivity theorists. To Mises it was both unfortunate and inexplicable that Böhm-Bawerk, who in his critical history of interest doctrines had "so brilliantly refuted" the productivity approach, himself fell, to some extent, into the same kinds of error in his Positive Theory. There is some disagreement in the literature on the degree to which Böhm-Bawerk in fact allowed productivity considerations to enter his theory. The issue goes back at least to Frank A. Fetter's remark in 1902 that it "has been a surprise to many students of Böhm-Bawerk to find that he has presented a theory, the most prominent feature of which is the technical productiveness of roundabout processes. His criticism of the productivity theories of interest has been of such a nature as to lead to the belief that he utterly rejected them....[But] it appears from Böhm-Bawerk's later statement that he does not object to the productivity theory as a partial, but as an exclusive, explanation of interest".*101 Much later Schumpeter insisted that productivity plays only a subsidiary role in what is in fact wholly a time-preference theory.*102 It is of some interest to note that when Böhm-Bawerk considered the alternative roles for productivity in a time-conscious theory, he came out squarely for an interpretation that placed productivity and "impatience" on the same level.*103 Böhm-Bawerk made it very clear that he was not willing to identify his position with that of Fetter, who espoused a time-preference theory of interest without any mention of productivity considerations. Böhm-Bawerk remarked that "Fetter himself espouses a [theory which] places him on the outer-most wing of the purely 'psychological' interest theorists—'psychological' as opposed to 'technical.' He moves into a position far more extreme than the one I occupy...."*104


Notice this is actually Riswik's claim--that from time preference you can infer the marginal productivity of capital because capital can have no other yield than that consistent with time preference. I cannot say I see this, but on the other hand it has the form of a solution similar to your idea that targeting the inflation obviates the need to define Y.

If Mises could make Hayek believe it...


I just realize that I have been snookered!

In your formula, because you are discounting the first payment by the short rate, which is a different rate from the discount rate applied to the second payment, you are implicitly assuming that the coffee shop is able to fund itself by rolling over short term debt or otherwise re-financing. My core assumption is that the coffee shop is not able to do this, and must fund itself by selling long term debt.

It is no different than saying that in the lean years, the coffee shop can borrow at the short rate and repay during the fat years. It is to avoid this type of re-financing risk that firms sell longer term bonds to finance long term investments, and they sell commercial paper to finance inventories. Admittedly there is no risk in my model, but grant me this assumption for purposes of argument.

Yes, you can have negative real rates ex-post, but arbitrage occurs against nominal rates, not real rates, borrowing occurs against nominal rates, not real rates, and the CB sets nominal rates, so let's stick to nominal rates when looking at what the CB can and cannot do.


Since you're missing the Neo-chartalists so much, let me suggest two ways in which one can rescue LP even while accepting the validity of your argument :

1) The central bank sets the short rate. 'The rate of interest' is the cost of capital, the long rate. CB tries to make the long rate clear its loanable funds value, but if the long rate does not respond much to changes in the short rate, LP of investors may still be determining 'the interest rate'. This is Keynes not from the GT, but from the Treatise, so I'm not sure a Robinson-Kaldor fan will invoke it. But I like this explanation, Leijonhufvud likes it, and it is how I think about the LM curve - with or without a central bank that controls the short rate.

2) The Fisher diagram (or Bohm-Bawerk's theory of interest) or indeed any theory of a savings-investment determination of interest necessarily has trouble explaining the existing capital stock and the trade in the outstanding stock of bonds. Fisher did this by assuming 'circulating capital', i.e. all capital goes down to 0 in all periods - hence flows are stocks, investment is capital. This is an absurd assumption. Much better is to observe, as Kalecki and Keynes did, that investment is just what capitalists do. Interest rates are an input into this but not determined by this. Investment determines income, and that determines saving, since saving is just an income residual. Central bank changes the short rate to control inflation, and it only does this through controlling the profitability of banks (ok, let's not open THAT can of worms).

I think that there's a way to rescue circulating capital (or as I prefer to call it, the equivalence of marginal capital and average capital) in a manner that's consistent with Bohm Bawerk, Keynes, Hicks AND Fischer Black. But I don't want to sound like a broken record, so I will leave it at that.

It's just confirmed. I talked to the owner of a coffee shop and asked him whether he could re-negotiate his rent. He verified that the rent was a fixed payment for a long term lease, and that he could not re-negotiate it.

I asked him why he didn't lower his prices, borrow short term, and use the proceeds to pay his rent. He looked at me like I was crazy.


That is a fair way to characterise my position on time preference/ MEC, though I'd just like to add that it's not just 'the interest rate' which changes to ensure the equivalence, *sometimes* it's income. That is Keynes's main insight, and it's one that stands up even in a world where central banks try to hit the loanable funds rate.

I was actually looking forward to reading about interest rates are determined, according to those who think loanable funds is bunk. My guess was that they'd argue something like there is nothing like a supply curve when it comes to credit. Then I was hoping Nick would explain how over the long-run, monetary policy decisions create something like a supply curve, and we could all shake hands and go home contented.

rsj: OK. I can't do the math to solve for your problem, but I can set up the equations. Your coffee shop owners like a 2 year lease with the same rents on both odd and even periods. That's like an annuity.

Let $Ao be the annual lease payment for a shop where the lease is signed at the beginning of an odd year, and $Ae the annual lease payment for a shop where the lease is signed at the beginning of an even year. Then we have two more equations to add to my two previous equations: (remember that $150 is the VMPk in an odd year and $250 is the VMPk in an even year).

The PV of a new shop at the beginning of an odd year is:

PVo= $100/(1+ro) + $150/(1+ro)(1+re) = $200 = $Ao/(1+ro) + $Ao/(1+ro)(1+re)

The PV of a new shop at the beginning of an even year is:

PVe= $150/(1+re) + $100/(1+re)(1+ro) = $200 = $Ae/(1+re) + $Ae/(1+re)(1+ro)

So we solve for ro and re as before, and use the two new equations to solve for Ao and Ae.

The Bank of Canada sets ro and re, (so we get full employment of labour in both odd and even years), and the market sets Ao and Ae so we get full employment of shops in both odd and even years (every shop finds a tenant who is just willing to sign the lease).

Luis Enrique: maybe, just maybe, nobody had ever sat down to explain to the Neo-Chartalists how to reconcile Loanable funds with Liquidity Preference, in a way they found satisfactory.

First, it is not immediately obvious that my second diagram is equivalent to my first diagram, and that the second is also a way of showing LF.

Second, and more importantly, maybe the only reconciliation they had seen was the ISLM method (like my third [I meant fourth NR] diagram) and they had assumed (wrongly) that the ISLM model assumed a fixed stock of money, which is an assumption they really don't like. (It is true that many (most?) treatments of ISLM do assume a perfectly everything-inelastic money supply function, in which the stock of money is exogenous, but it doesn't have to be taught this way.)

Notice, by the way, that I never actually talked about "funds" in this post (except as part of the same "Loanable Funds". I talked about saving and investment instead. I did talk about "bonds", but only in my digression/critique of ISLM.

Or maybe I'm totally wrong in trying to figure out where they're coming from and why they don't like Loanable Funds.

Ritwik: I'm going to take a stab in the dark to engage your second point:

Here are two very simple and very extreme models of capital and interest. (The real world lies somewhere between those two extremes, and is much more complex).

1. The only good is a plant, which never dies. You can eat the plant, or, if you don't eat it, the bits of the plant you don't eat grow at rate m. If the current size of the plant is K, then the flow of output/income is mK, and the growth rate of the economy is sK where s is the savings rate. The rate of interest will be equal to m. The PPF in the multiperiod Irving Fisher diagram is a straight line with slope 1+m. (IIRC the plant is called the Crusonia plant? Or is it schmoo?). This is actually equivalent to the AK growth model.

Without changing any of the results, we can instead assume the plant dies every year, but the seeds from one plant will grow 1+m new plants, if you don't eat any of the seeds. This then becomes a model of circulating capital, with 100% depreciation.

2. The only real asset is a fixed stock of land, and you can never create new land. Land produces wheat, which cannot be stored. You can borrow and lend wheat, and buy and sell land, but aggregate output is exogenous. The Irving Fisher diagram has a reverse L-shaped PPF. The rate of interest (and hence the price of land in terms of wheat) is determined by the rate of time-preference.

This is an extreme fixed capital model. Land doesn't depreciate at all, and you can't consume it.

Nick @7.07,

Now your model is over-determined in multiple ways:

1) If the equilibrium rental payment is $150 in even years and $100 in odd years, but if people sign a contract to have the same payment in both years, then whatever the ECB does, rents will be too high for some stores and too low for other stores. Some stores will go out of business while others will earn excess profits.

2) Looking at your equation, divide both sides by the payment. You get a system:

200/A0 = something close to 2
200/A1 = something close to 2

This means A0 is about 100 and A1 is about 100, regardless of where the CB sets rates, and regardless of what the market clearing payment is.

ignoring terms of interest rate squared, you get

200/A0 - 200/A1 = r1 - r2 = something much less than 1 in absolute value.

Which goes to the point that if A0 and A1 are too far from each other (e.g. "wiggle") then there is no solution.

I'm not sure why this is so hard -- it's clear that if we hypothesize that consumption goods are bought with multi-period fixed price contracts -- I think they are called Taylor contracts -- that this will lead to consumption good price stickiness. Same for labor.

Why then isn't it also clear that if people borrow with fixed payment bond contracts that this will lead to rate stickiness. The only wrinkle here is that I am assuming the rate of the bond contracts are set via EMH to be the average of future short rates as set by the CB, and then I am asking whether the CB can select short rates in such a way so that their average can jump around a lot, and the answer is no.

Hi Nick.

I don't think that this was covered above, but one thing I struggle with from time to time is why the standard loanable funds story focuses only on flows and ignores stocks. If one assumes that some of the existing capital stock is fungible (either because it was designed to be used flexibly or because, at its current stage of production, a change in intended final use is still possible), and can thus be repurposed (or perhaps even consumed), then, in effect, it can be converted back from the "investment side" (use of real savings) of the market to the "pool of real savings" side of the market. Does the flow-only model implicitly assume that all investments are non-fungible? Have I got this wrong?

rsj: You can either:

1. rent the shop period-by-period, with a high market rent in even periods and a low market rent in odd periods.

2. Or you can sign a lease contract for two periods with the same rent in both even and odd periods.

The two options have the same Present Value. If you sign the lease, you make losses in odd periods and profits in even periods, but the PV of those profits and losses is zero.

But if you sign the lease at the beginning of an odd period you pay less per period than if you sign the lease at the beginning of an even period. That's because your losses come now and your profits come later, and so are discounted.

I think stocks versus flows is an important distinction. aq

david: you are not wrong.

Take an extreme case of what you are talking about. A one-good model where the capital and the consumption good are identical. The desired investment curve becomes horizontal. The rate of interest is determined by technology, and preferences determine the amount of saving/investment. (The 2-period PPF in the Irving Fisher is a straight line.)

In fact, the standard aggregate neoclassical production function used simple macro models has exactly this feature. Which is why in that model the rate of interest at a point in time really is determined by MPK, and not by saving preferences at all. But that model (applied rigourously) also gives you a horizontal IS curve, so Liquidity Preference would be false even in the short run. And so all the textbooks go to an awful lot of bother bringing in adjustment costs to try to derive a downward-sloping Investment curve and downward-sloping IS curve.

Put it another way. In the simple one-good model, you don't get a relation between the flow of desired investment and the rate of interest. You get a relation between the desired stock of capital and the rate of interest. Which makes the Id curve perfectly elastic. And because empirically the Id curve doesn't seem to be perfectly elastic, they stick in adjustment costs, to slow down the transition to the desired stock of capital.

Move away from the one good model, and you don't get this problem. You need to increase the price of capital goods relative to consumption goods to get a finite reallocation of supply away from producing consumption towards producing capital goods.

Nick @2:27

Think about why store owners don't want option 1. In our perfect foresight model, it makes no sense, but if you add risk, then store owners do not want to face the risk that next month's rent will be raised more than they expect. You can think of it as roll-over risk. They have a recurring payment that needs to be made, but in order to plan they need some certainty about their costs. So they sign a contract that fixes their costs for a period, and are willing to pay more to sign this contract.

That contract is option 2 -- fixed payments. But if you are saying that the store owner should borrow in one period and repay the loan at the next period, then you are re-introducing rollover risk and completely nullifying any benefit to the 2 period contract.

So when you add risk, stores will not borrow short to pay their rent, they will borrow short to finance inventory, but they will need to pay rent out of operating income.

Similarly, creditors will not be willing to lend short to a firm that doesn't have enough earnings to cover interest costs but needs to borrow to make the coupon payment.

The above post was by me, not ZP -- sorry.

Thanks Nick for your reply.

If one wanted to adjust the flow-only model for some assumed fungibility in the existing capital stock, then:

a) the Id (and presumably Sd?) would be more elastic; and

b) both curves would shift rightward (?).

Is the critical feature fungibility as between investment use and consumption - i.e., is any fungibility between various investment uses irrelevant (except perhaps post-bubble recognition of malinvestments, assuming for a moment an element of austrianism)?

And I guess the point here is that it is not enough to say "we need to take longer interest rates into account", but rather the model should understand why people borrow long term at all.

Why do multi-period fixed coupon bonds even exist in the first place?
Why doesn't the non-financial sector engage in maturity transformation?
Why does the financial sector (which does engage in MT) have to be protected with an armada of credit subsidies, emergency lending facilities, and federal guarantees in order to keep from exploding every few years?

I don't think any aspect of loanable funds make sense at all unless you introduce a reasonable model of risk into the picture.

david: if the stock of capital goods were more fungible into current consumption (and vice versa) that would make the Id curve more elastic. I don't think it would have any effect on the elasticity of the Sd curve, which depends on preferences.

Maybe in a world of uncertainty people would save more and invest more if they knew that any investment projects were easily reversible so they could get their savings back even if they regretted the investment?

I think fungibility between different types of capital goods will matter in a world of uncertainty. You would invest more and take bigger risks if you knew you could always covert your capital goods into a different use if relative demands and supplies changed.

rsj: "I don't think any aspect of loanable funds make sense at all unless you introduce a reasonable model of risk into the picture."

I would put it this way: the simple LF model I have sketched above is a model of "the" rate of interest. You can think about how those curves shift over time (recognising that the current position of those curves depends on expected future interest rates) and get a theory of the time-path for "the" (one period) rate of interest over time.

But if you want to talk about things like the term structure, and the whole slew of different rates of interest on different financial and real assets, that differ not just by term but by risk, liquidity, etc., then sure. You need something more than that simple LF model. That's why God invented Finance people!

Now, you can see Arrow-Debreu as just an extension of the Irving Fisher diagram, which in turn is just a different way of looking at Loanable Funds. And *in principle* Arrow-Debreu handles risk and time, and can talk about the returns on any instrument you care to think of, because any such instrument is a bundle of Arrow-Debreu contracts. Trouble is, it assumes zero transactions costs (so all assets are perfectly liquid, and there are no enforcement problems and moral hazard problems etc. because you can observe all states of the world) and complete markets, etc. So it sort of throws away all the fun parts of Finance.


"the simple LF model I have sketched above is a model of "the" rate of interest."

You mean the risk-free rate of interest?

Lets get back to your plant. Assume the return from owning the plant is either m0 or m1  (m0 < m1) with equal probability. You and I both are the only two agents in the economy and we both hold all our wealth in plant (no other choice). Lets say you are younger than me and would like to hold some more plant exposure. I'm old and prefer to have guaranteed consumption next period. So I give you a one period loan which you use to buy some plant from me. You collateralize the loan with plenty of plant, so I'm guaranteed to get my money back with interest. It's a risk-free loan.

Since neither of us is stupid, I know for sure that I wont lend you the money at a rate lower than m0 and you wont borrow at a rate that's higher than m1.  If we are both risk averse (concave utility of consumption), then I'm also quite certain that you'll want to pay less, and I'll be willing to lend for less than the expected return (m0+m1)/2 of owning plant. Apart from that, I have no idea what rate we'll agree on. It strikes me that determining the risk-free rate is a function of our relative risk aversions (investment horizons, patience, whatever), in other words, very much a risk/finance problem as rsj says. How can Loanable Funds save us here?

K: "How can Loanable Funds save us here?"

It can't. You need a model of risk-sharing.

But it's easy to generalise it so you can.

Take the Irving Fisher diagram, and make it 3 dimensional, so there are two future states of the world. But you also have two agents, so you need to add both their preferences in.

Easy enough to do with math (if unlike me you can do math).

We each have an endowment today, and we can consume it, plus or minus what we borrow or lend, minus what I invest in the plant to increase consumption tomorrow. Let the price of consumption today be 1 (it's the numeraire. Let the price of a unit of comsumption next year, conditional on state 0 be Po. Ditto P1 for state 1.

We each max U(Ctoday)+ EB(Ctommorrow) subject to the budget constraint to solve for our consumption and saving and investment plans as a function of the two prices. Then solve for the prices that give us the equilibrium where everything adds up right. Those two prices implicitly define the interest rates.

But the equilibrium won't be exactly like you suggest. Unless you have weird preferences, even if you are more risk-averse than me, you will generally hold a mixture of debts and equity, but a smaller proportion of equity than me. In other words, your consumption in the good state will be higher than in the bad state.

(And it only really works if there are hundreds of me and hundreds of you, so we have a competitive equilibrium. Otherwise it's bilateral monopoly/monopsony if there's just me and you, and I can solve for the range of efficient outcomes, but we have to invoke something like the Nash Bargaining solution to tell us exactly what happens.)

K: OK, I can see it now, like the promised land! A 3D PPF in the Irving Fisher diagram with a 3D Edgeworth Box inside it.


Since, we are talking about capital theory now, yes, the Knightian crusonia plant. I prefer your explanation 1. Like I said, I don't like 100% depreciation assumptions. But the crusonia plant may yet be saved.

To establish the *equivalence* of investment and capital, we don't need to force capital down to 0. Investment is the marginal capital. So we need to show that the marginal capital is equivalent to the average capital. Where might we look for intuition?

1) From the Austrians (and Hicks), we take 'capital is time'. However, we don't get lost in the 'period of production' distraction.

2) From Keynes (and Hicks) we take that investment is long-term and outlives the business cycle, so that capital (both existing and marginal) financing/discounting is essentially through long bonds/perpetuities.

3) From Fisher (and Knight, and Hicks, and Black) we take that the production process is an integrated whole, so that capital is not a thing, so much as the production process itself. Investment by a firm is buying more of the same firm, it is buying *time*. Thus we lay to rest fears about modelling heterogeneity of capital as long we can model heterogeneity of firms.

4) From Fisher(and Black) we take that in the aggregate, the labour/capital distinction is futile. All input is capital, and human capital is quantitatively more important.

5) From Keynes (and Black) we take that the fundamental feature of time is uncertainty. So that *buying time* necessarily involves buying uncertainty.

6) From Black (and Tobin) we take that idiosyncratic risk can theoretically be hedged away. Heterogeneity is co-variance, not variance.

1,2,3 together tell us why the marginal capital could be equivalent to the average capital, so that we can model smoothly growing capital rather than assuming 100% depreciation. A stock of capital moving through time throwing off a flow of goods and services in each period.

3,4,5 tell us that this stock of capital in uncertain, it is likely to change every minute. Much of the capital is unobservable, but our best estimate of the stock of capital comes from the stock of wealth.

The rate of interest is the return on the total capital. Y0 = r0*K0. K1 = (1+ro-c)K0, so that Investment = K1-K0 = (r0-c)K0 = Y0 - cK0. Consumption is cK0. The growth rate Y1-Y0 = r1K1 - r0K0 and is not a simple function of the 'savings rate' (even in the AK model, it is the the level of capital and output - not the growth rate - that depends on the savings rate).

5 and 6 tell us that r is stochastic, and subject to uncertainty. The IS curve is horizontal, but this is not a one good model. It is a multi-sector multi-good model where the differentiating heterogeneity and uncertainty of a sector is captured by its co-variance to the economy.

We close the horizontal IS curve with an upward sloping, stochastic AS curve for a 'real model'. For a 'monetary model', we close it with a upward sloping LM curve.

This is a theory of the risky long rate. The risk free rate is endogenously determined through a market for collateralized loans.


"you will generally hold a mixture of debts and equity"

Definitely. I said I'd sell you "some" plant. Not all my holdings.

Maybe we agree - need to think some more...

Since, we are talking about capital theory now, yes, the Knightian crusonia plant. I prefer your explanation 1. Like I said, I don't like 100% depreciation assumptions. But the crusonia plant may yet be saved.

Can it? I'm afraid I lost a thread here. Aren't you taking the subjectivist view? The problem with the Crusonia plant is that it requires no husbandry. Naturally then it cannot reflect time preference. The subjectivist claim was always that such things were mere fiction--that no capital exists without human action and therefore all capital that does exist yields in accord with time preference... that's Nick's causality and backwards L in one go.


Did you miss the entire monologue that followed?! I was saying that time preference/ MEC are perhaps two ways of stating the same economic fact. Anyway, I'm no Misesian. I'm not even very interested in his capital theory. I'm also not very interested in fables of 'no husbandry' etc. I might have contradicted myself in MIses-ian terms, I do't particularly care.

'Evaluation of uncertainty' or 'risk-bearing' is human action. So clearly r depends on human action. Consumption cK is also human action, so clearly the size of the plant depends on human action. When MEC/r is stochastic, the distinction between the technical fact of marginal productivity and the human fact of time preference is irrelevant.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad