As a sidebar to the recent monetary policy and money creation debate, I have been debating JKH, over at MR, on the nitty gritty details of what the central bank can do to hit new target interbank rates. (We largely agree on the broader issues.) I'm using this post as a venue to lay out my position more clearly, with the aid of graphs. Apologies to those who haven't been following, if this seems like it comes out of nowhere. EDIT: For those who relate to the U.S., we're specifically debating this issue outside of a "floor" system, which is the situation in the U.S. currently. So think something like pre-2008 U.S., as one example. I discuss the issue more generally than the Fed in the U.S., though.
Now suppose that the central bank announces that their new target interbank rate is 2.25%. They announce that they will not be changing the ceiling or floor rates. Holding everything else constant, if the market believes that the central bank will enforce their new target rate, then it must anticipate a movement of the supply curve to the point where a 2.25% market rate would be a profit-maximizing (or cost-minimizing) outcome for the market.
If these are the expectations, the market rate should move to the new target rate prior to the central bank actually shifting the supply curve. However, the market will still expect the central bank to follow through on its implied promise of a shift in the supply curve. If and when the central bank shifts the supply curve, this will not cause the interbank rate to overshoot the target (contrary to my interpretation of JKH's and MMT'ers past assertions).
Suppose instead that the central bank announces that their new target interbank rate is 2.25%, but that they will be changing the ceiling and floor rates to 2.75% and 1.75%, respectively. Upon this news, the demand curve should shift downward. In this case, I agree with JKH comments on the automatic shifting of the demand curve. Continuing to assume a 50/50 probability that banks end the reserve maintenance period short or long at the current quantity of reserves S*, then the shifting of the demand curve alone will yield the new target rate, without requiring any change in the quantity of reserves.
In this case, if the central bank were to also inject reserves, then the interbank rate would fall lower than target. The degree to which it would overshoot would depend on how many reserves the central bank injects. But contrary to what I’ve seen in the MMT/MR blogosphere, it would not necessarily fall to the new floor of 1.75%.
As I interpret JKH (and Scott Fullwiler, associated with MMT), altering reserve quantities can only be used to defend a new target rate, not to offensively set a new one.
From their viewpoint, rates move due to an "announcement effect,"
whereby the central bank's announcement of a new target rate is enough to move
the rate without the need for any other action, all else equal. Already, JKH has clarified that he believes the announcement effect must incorporate a change in corridor rates for it to move rates, all else equal. This I agree with, but I don't consider this to be a "pure announcement effect," in which it's only the central bank's words, rather than its actions, that matter. (I'd also add that the movement of the corridor may still not be enough to hit the new target. Depending on the corridor rates and probability of banks ending short or long reserves, quantity changes might be required as well, or more theoretically, a reserve requirement change.)
More generally, my aim is to show them that altering reserve quantities can most certainly be used to offensively achieve a new target rate. Admittedly, I am employing the assumptions that Bindseil and Woodford use in their most basic monetary policy implementation models - but I view these as plenty rigorous for our purposes.
Suppose we start with the following scenario. The central bank’s target rate is 2.5%. The ceiling and floor rates are 3% and 2%, respectively. There is a 50/50 probability that the market will end short or long reserves assuming the quantity of reserves available in the market is exactly equal to the quantity of required reserves. Thus, the central bank sets the quantity of reserves equal to required reserves to achieve their target rate.
FIGURE 1
FIGURE 2
FIGURE 3
In this case, if the central bank were to also inject reserves, then the interbank rate would fall lower than target. The degree to which it would overshoot would depend on how many reserves the central bank injects. But contrary to what I’ve seen in the MMT/MR blogosphere, it would not necessarily fall to the new floor of 1.75%.
In conclusion, if all along, JKH, Fullwiler, et al. had this last case in mind, we would be in agreement. But that was never made clear, in my opinion. Furthermore, the other case, involving a shift in quantity of reserves, is just as fundamentally sound. Understanding both cases allows one to see that monetary policy implementation can be a matter of rates and/or quantity.
APPENDIX A: Changing Reserve Requirements
From a previous interbank rate of 2.5%, suppose that the central bank announces that their new target interbank rate is 2.25%. They do not change the ceiling or floors nor the quantity of reserves. Instead, they lower the reserve requirement. This should cause a leftward shift in the demand curve. In other words, given a fixed quantity of reserves, a lower reserve requirement means that banks will be less likely to end the reserve maintenance period short. As such, the interbank rate should drop below 2.5%.
Continuing to assume a symmetric probability distribution (50/50 short/long) around the quantity of required reserves, the central bank could have maintained the 2.5% interest rate if they lowered the actual quantity of reserves to the new quantity of required reserves by conducting an OMO. In our example, the central bank does not do that since they want a lower interbank rate.
APPENDIX B: Interbank Rate Targeting with Fixed Target-Ceiling Spread
Added in from comments: Assume a system similar to the U.S. pre-2008, where there is a fixed spread between the target rate and the ceiling – say it’s 2%. This means that as the target interbank rate changes, its distance from the floor will change, whereas its distance from the ceiling will remain fixed. We might call this an 'asymmetric' channel, in contrast to the 'symmetric' channel of Figure 3 above.
APPENDIX A: Changing Reserve Requirements
From a previous interbank rate of 2.5%, suppose that the central bank announces that their new target interbank rate is 2.25%. They do not change the ceiling or floors nor the quantity of reserves. Instead, they lower the reserve requirement. This should cause a leftward shift in the demand curve. In other words, given a fixed quantity of reserves, a lower reserve requirement means that banks will be less likely to end the reserve maintenance period short. As such, the interbank rate should drop below 2.5%.
FIGURE 4
Continuing to assume a symmetric probability distribution (50/50 short/long) around the quantity of required reserves, the central bank could have maintained the 2.5% interest rate if they lowered the actual quantity of reserves to the new quantity of required reserves by conducting an OMO. In our example, the central bank does not do that since they want a lower interbank rate.
APPENDIX B: Interbank Rate Targeting with Fixed Target-Ceiling Spread
Added in from comments: Assume a system similar to the U.S. pre-2008, where there is a fixed spread between the target rate and the ceiling – say it’s 2%. This means that as the target interbank rate changes, its distance from the floor will change, whereas its distance from the ceiling will remain fixed. We might call this an 'asymmetric' channel, in contrast to the 'symmetric' channel of Figure 3 above.
Assume a symmetric probability distribution of anticipated liquidity shocks and, just as a starting point, that the current quantity of reserves in the market is the quantity of required reserves. That means if the target rate is 3%, the ceiling is 5%, and the floor 0%, the interbank rate should be 2.5% (5%*.5 +0%*.5). Therefore the central bank is missing its target. To move the interbank rate up to 3%, the CB is going to have to remove a quantity of reserves so that it becomes more likely the banking system will end the reserve maintenance period short. In particular, there needs to be a 60% chance of this occurring (3% = 5%*.6 + 0%*.4). See Figure 5.
FIGURE 5
Then say the CB changes its target to 2%, which means the ceiling will be 4% (see Figure 6). Now, assuming the same symmetric probability distribution of liquidity shocks, the Fed needs to inject reserves back into the system until the quantity of reserves is the quantity of required reserves (2% = 4%*.5 + 0%*.5). Otherwise, the new rate will be 2.4% (4%*.6 + 0%*.4). In this case, changing the ceiling/floor rates was simply not enough to hit the target. The central bank needed to also change the quantity of reserves, even controlling for any changes in reserve distributions or any other market abnormalities. This is most certainly a liquidity effect. I am not denying that the rate will move to 2% upon announcement, but the central bank still needs to lower the ceiling and reduce the quantity of reserves to maintain the rate as well as its credibility. At least as far as this model goes.
FIGURE 6
Sorry if I'm just misunderstanding you but my impression is that things are very different if there is a massive glut of reserves versus the situation where availability of reserves provides a constraint. In the UK, the USA and Japan we now have a massive glut of reserves such that the system is really a floor system (not a corridor system) and there is little prospect that short term rates will rise above the floor set by the interest on reserves paid by the central bank. I think the MMT position is that monetary authorities have a duty to maintain that situation and always ensure that supply of reserves is always ample such that short term rates never rise above a floor (preferably in the MMT opinion a floor of zero as in Japan). If there is a sufficient glut of reserves, then isn't it true that adding further to that glut (or somewhat trimming down a glut from massive to somewhat less massive) will nevertheless not shift rates off the floor irrespective of any announcements? Under such a scenario, monetary policy is impotent. My impression is that the whole "market monetarist" enthusiasm for "quantities not rates" is because they hope that quantities of reserves can still continue to help the economy even when rates are stuck to the floor due to a massive glut of reserves. Basically the market monetarist "quantities" idea is a justification for not resorting to fiscal policy once monetary policy has grounded at the zero bound (and so become impotent from a "rates viewpoint").
ReplyDeleteYou say that reserve requirements could counteract the glut of reserves but my impression is that they have very limited scope for doing that. John Hussman's web site says that, in the USA, there are now more reserves than there are checking deposits.
I also had a go posting about this
http://directeconomicdemocracy.wordpress.com/2014/01/05/money-demand-that-lowers-interest-rates/
Hi Stone -
DeleteYup, I'm specifically addressing the situation where the central bank is *not* running a floor system. Something like pre-2008 in the U.S., as one example. Note in the graphs, the supply of reserves is not at the floor.
In the scenario where there is a floor system, you're correct that expanding the glut, or diminishing it to a lower bound, would not affect rates. I wouldn't say this means monetary policy is impotent though, from a rates perspective. They could of course still change the floor rate.
"You say that reserve requirements could counteract the glut of reserves but my impression is that they have very limited scope for doing that. John Hussman's web site says that, in the USA, there are now more reserves than there are checking deposits."
Well I wasn't addressing a floor system, but you bring up a separate, interesting point. However, I would say that, at least theoretically, the central bank could make reserve requirements over 100% of deposits to address this issue. Also, at least in theory, reserve requirements could be based off of something other than deposit ratios... Of course, draining the glut of reserves via open market operations is the more likely option that tends to get discussed.
"However, I would say that, at least theoretically, the central bank could make reserve requirements over 100% of deposits to address this issue."
DeleteI've argued the same thing in debating Vincent Cate the hyperinflationist.
Apologies for not realizing that you were specifically dealing with the pre-2008 type situation.
ReplyDeleteIts an interesting idea about making reserve requirements a multiple of deposits (eg 200% reserve requirements or whatever). To me that seems a lot more viable than the idea of having high interest paid on reserves or selling back all of the bonds sold via QE. The long duration bonds bought via QE would fall in price if the central bank conducted a big sell off so they would fail to gather back enough reserves to cause a scarcity of reserves. Also the central bank needs to find the money to pay for interest on reserves. As far as I can gather, the central bank would have a hard time paying more interest on reserves than it receives as interest on from the bond portfolio it holds.
The central bank does not need to find the money to pay for interest on reserves. They can simply create the reserves. So that aspect isn't really a problem. This is in contrast to the Treasury, which must raise funds prior to spending.
DeleteThe idea of 100%+ reserve requirements strikes me as pretty implausible in reality, although I confess to not having a detailed knowledge of the prons and cons on the matter. At the very least, I haven't seen anyone discuss it. I just brought it up as theoretical possibility, since it highlights another integral but seldom-explored dimension of monetary policy implementation.
Is it really true that say the Fed or the Bank of England can pay interest on reserves using freshly created reserves? I thought that they were bound by the rule that they had to receive assets at market prices in exchange for any newly created reserves. The whole basis of what they do is defined by the framework of rules they work within.
DeleteI'm not aware of a rule like that...
DeleteOther banks might be allowed to print money to pay interest on reserves, but I'm pretty sure the Federal Reserve can't create dollars to pay IOR. The Fed has very strict collateralization policies, creating money de novo to pay IOR would probably break those rules.
DeleteAlright - I stand corrected. Thanks for making me look stupid now rather than later :). I don't remember this issue being addressed. JKH probably has. I'll have to revisit that Sack / Gagnon reverse repo facility paper to see if they did.
DeleteI guess then I also agree that the reserve requirement idea is maybe more interesting...
DeleteThe Fed does have the authority to pay interest on reserves (it didn't always, but it does now). Fresh money or stale money, what difference does it make? The issue is that the Fed might lose money on its investments, whether it sells them at a loss or pays increasing interest on reserves. Would the Fed potentially need a bail-out to repair its balance sheet? Nah, they'll just conjure an asset by withholding future profits from the treasury. The Fed's value as a business far exceeds any potential losses.
Delete"The idea of 100%+ reserve requirements strikes me as pretty implausible in reality,"
DeleteI've brought up the idea to Sumner and Sadowski too: Sadowski, thought that IOR was the more "modern" way to go, and likened having both methods available to using both a belt and suspenders.
Sumner agreed it could work but called it "clumsy."
I've been told before that I'm wrong about this (but I still don't see how) but to my mind QE could have a real effect on asset prices by acting as a bridge burning exercise ensuring that the Fed can't raise interest rates much in the future even if it were to want to. The more QE we have had, the more remote the chance that the Fed will raise short term rates in the next few years. The market sees that we have say another trillion dollars of surplus reserves and interprets that as being a reduced risk of an interest rate hike in the next few years and so prices bonds and stocks (and maybe even takes on debt) on that basis.
ReplyDeleteThe >100% reserve requirement idea perhaps undermines this, but to me it is still an open question whether that could actually function.
Hi ATR,
ReplyDeleteGood post.
I’ve responded back at MR in more detail – independent of your post here. I’ll add a few points, but you might go back to review my review of what I’ve said previously on this topic (along with Fullwiler and Woodford), because it looks as though you’ve misinterpreted some of it, although not severely so.
I agree very ROUGHLY with your second example. The difference is that I would not construct a hypothetical in which excess reserves were zero, because I have focused almost all of my interpretation of this topic on the pre-2008 Fed rather than the case of a symmetric channel system, and that zero excess reserve assumption is generally not the case for the pre-2008 Fed. (See Woodford’s comments on that in the case of the Fed and see the empirical evidence.) Furthermore, the example that I’ve referred to as typical includes - in the case of an announced lowering of the target rate - a pre-announcement excess reserve supply that has nudged the existing funds level a bit below the existing target, in order to prepare the market for a change in the target and a subsequent b) type response (see my MR comment for a review of my definitions of a) and b) type adjustments and a more detailed discussion of them.) That does not conflict with anything I’ve written before on this. So if you strip out those differences from your example, we are in general agreement I think about the nature of the b) effect. Your example thus qualified is fine in my view, but your conclusion that it is “contrary to my interpretation of JKH's and MMT'ers past assertions” is an incorrect extrapolation at least in my case because you have used a different starting assumption about the pre-announcement excess reserve position.
I would define (and I think most people would) the “announcement effect” to be the immediate response of the market to the central bank’s communication that it has changed its target rate structure – where the target rate structure includes the set of administered rates that typically change in conjunction with that communication. That is certainly consistent with Woodford’s definition of it in the case of the pre-2008 Fed explicitly (see my MR comment) and it is consistent with Fullwiler’s interpretation of it. For the pre-2008 Fed that includes the target fed funds rate and periodically the discount rate (i.e. the upper bound for the asymmetric channel). For a central bank operating a symmetric channel, that naturally includes all three rates – target, upper bound, lower bound.
Given that definition, I wouldn’t include the first example in your post here as indicative of anything I may have said on this topic. It is a pathological case that effectively considers a change in the administered rate regime itself – from a normal symmetric channel to some sort of abnormal asymmetric arrangement. It doesn’t reflect what might be considered a normal announcement effect under any regime. I can probably invent weirder scenarios that that which would require reserve action by the central bank in addition to hospitalization for bank reserve managers. It’s simply not normal. I’m certainly not denying that reserve regimes can’t be discretely changed (the U.S. has gone through one in effect through QE – at least temporarily if not permanently.) So the typical announcement effect in the case of a channel system should assume normal shifting of all components that make a channel a channel. It makes no practical sense otherwise.
... cont'd
cont'd ...
ReplyDeleteAlong those lines, I would also consider a change in reserve requirements to be orthogonal to what is normally considered to be the announcement effect in the case of a rate change. Reserve requirements are a structural issue that is beyond the dimension of the market response to a change in the interest rate target in a given central bank operating regime. I’m not denying it can’t happen – just that it’s not relevant to the type of announcement effect that relates to a given administered target rate regime – be that an asymmetric corridor, symmetric corridor, standard floor, or QE. And I’m certainly not denying that regulatory reserve requirements can’t be changed or that they will have some sort of effect on the demand for reserves in aggregate.
I’ve responded more fully at the MR post. I’ve included my interpretation there of a couple of special items you noted, such as the MMT “natural rate of interest is zero” dynamic as well as my point regarding the secular trend in excess reserves.
A final point. You say in your post here:
“As I interpret JKH (and Scott Fullwiler, associated with MMT), altering reserve quantities can only be used to defend a new target rate, not to offensively set a new one. From their viewpoint, rates move due to an "announcement effect," whereby the central bank's announcement of a new target rate is enough to move the rate without the need for any other action, all else equal.”
“Not offensively set a new one” is misleading in my view. A target change with a b) type response does not at all preclude a prior a) type adjustment. The use of the word “offensive” is insufficiently precise and potentially ambiguous in that context. I would stay away from it.
The same caveat applies to “can only be used”. Again, the split between a) and b) type responses and their relative timing is precise as defined. Obviously nothing precludes a) type actions either before or after the b) type announcement effect. It is the distinction of the b) type announcement effect that I have been consistent on all along and that Fullwiler has agreed with and that Woodford effectively agrees with in his description of it.
As I said at MR, this stuff is hyper-wonky, and it’s my second go around on it over the past several years, so I’m not terribly keen to add a lot more at this stage. It’s been a useful exercise for me to be prodded to check my previous thinking on it, but I would not change my interpretation at this point. Feel free to continue to disagree and to conclude with disagreement if you must, but this topic is not quite intriguing enough to warrant eternal debate from my perspective – at least not at this time. :) Please see my closing remark at the MR comment – and again, nice work and thanks for the discussion.
http://monetaryrealism.com/money-creation-in-the-modern-economy-bank-of-england/#comment-103012
“I agree very ROUGHLY with your second example.”
DeleteWell that’s progress! I’d note that if you agree with it, then there needs to be an acknowledgement of a liquidity effect in the case of a rate change. OBVIOUSLY, under certain regimes or rate changes, it may not be used. But to deny its fundamental existence in monetary policy implementation debates regarding rates v. quantity is an error.
“The difference is that I would not construct a hypothetical in which excess reserves were zero, because I have focused almost all of my interpretation of this topic on the pre-2008 Fed rather than the case of a symmetric channel system, and that zero excess reserve assumption is generally not the case for the pre-2008 Fed.”
That’s fine, but that changes absolutely nothing regarding the fundamental underpinning of what can drive a rate change in this model. Every country will have its own idiosyncratic empirical realities. We’ll be here forever if we try to iron those out completely.
“Your example thus qualified is fine in my view, but your conclusion that it is “contrary to my interpretation of JKH's and MMT'ers past assertions” is an incorrect extrapolation at least in my case because you have used a different starting assumption about the pre-announcement excess reserve position.”
But my pre-announcement excess reserve position is not reversed and fundamentally should not be, which is contrary to what you wrote. Your example is best illustrated by my Figure 3, and what I wrote after that. That is, the rate change (via the ceiling/floor) does all the heavy-lifting to the move the rate. Any pre- or post-announcement change in quantity of reserves would cause the interbank rate to under or overshoot. It would need to be reversed, and thus net quantity of reserves doesn’t change.
“For the pre-2008 Fed that includes the target fed funds rate and periodically the discount rate (i.e. the upper bound for the asymmetric channel).”
DeleteNot central to the debate, but I was under the impression that the discount rate moved with the target rate pre-2008 by some fixed spread. Am I wrong here?
“Given that definition, I wouldn’t include the first example in your post here as indicative of anything I may have said on this topic. It is a pathological case that effectively considers a change in the administered rate regime itself – from a normal symmetric channel to some sort of abnormal asymmetric arrangement. It doesn’t reflect what might be considered a normal announcement effect under any regime. I can probably invent weirder scenarios that that which would require reserve action by the central bank in addition to hospitalization for bank reserve managers. It’s simply not normal. I’m certainly not denying that reserve regimes can’t be discretely changed (the U.S. has gone through one in effect through QE – at least temporarily if not permanently.) So the typical announcement effect in the case of a channel system should assume normal shifting of all components that make a channel a channel. It makes no practical sense otherwise.”
This is simply not fair on multiple fronts, and is perhaps further evidence you perhaps do not fully understand what’s going on here. Again, the fundamental nature of the interbank market doesn’t change depending on the type of regime run. I chose to do it the way I did because it’s the easiest way to show that quantity changes can enforce rate changes. Furthermore, the U.S. pre-2008 didn’t run a channel system in which the corridor changed as you describe it above, anyways. But to make you happy, see Figures 5 and 6, which illustrate an example I typed at MR in response to you, which is closer to the U.S. pre-2008. The idea is having a fixed spread between the target rate and the ceiling, with a floor of 0%. Quantity adjustments are still needed to enforce rate changes, in addition to corridor changes – and they are not reversed.
Reproduced from MR:
Delete“take something like the U.S. pre-2008, where there is always a fixed spread between the target rate and the ceiling – say it’s 2%. Assume a symmetric probability distribution of anticipated liquidity shocks and, just as a starting point, that the current quantity of reserves in the market is the quantity of required reserves. That means if the target rate is 3%, the ceiling is 5%, and the floor 0%, the CB is going to have to remove a quantity of reserves so that the rate moves up from 2.5% (5%*.5 +0%*.5) to 3% (5%*.6 + 0%*.4). In this world, the central bank has to keep the banking system, as a whole, in a state of reserve deficiency. Then say the CB changes its target to 2%, which means the ceiling will be 4%. Now, assuming the same symmetric probability distribution of liquidity shocks around the quantity of required reserves, the Fed needs to inject reserves back into the system until the quantity of reserves is the quantity of required reserves (2% = 4%*.5 + 0%*.5). Otherwise, the new rate will be 2.4% (4%*.6 + 0%*.4). In this case, changing the ceiling/floor rates was simply not enough to hit the target. The central bank needed to also change the quantity of reserves, and it didn’t have to do with your a)-type adjustment. This is most certainly a liquidity effect.”
“Along those lines, I would also consider a change in reserve requirements to be orthogonal to what is normally considered to be the announcement effect in the case of a rate change.”
Agree it’s not normal, but I illustrated it to show it’s an unorthodox way to bring about a rate change within a pre-existing corridor that doesn’t involve changing administered rates or reserve quantities. A comprehensive understanding of the fundamental drivers of interbank rates should include an understanding of the role of reserve requirements, even if they aren’t commonly used as a policy tool.
““Not offensively set a new one” is misleading in my view. A target change with a b) type response does not at all preclude a prior a) type adjustment. The use of the word “offensive” is insufficiently precise and potentially ambiguous in that context. I would stay away from it.”
Fair enough. My use of the term was to indicate the implementation of a rate change that involved a permanent change in quantity – and a rate change that occurs within a corridor without sending the rate to the ceiling or floor.
To be clear, when I write:
Delete"Well that’s progress! I’d note that if you agree with it, then there needs to be an acknowledgement of a liquidity effect in the case of a rate change. OBVIOUSLY, under certain regimes or rate changes, it may not be used. But to deny its fundamental existence in monetary policy implementation debates regarding rates v. quantity is an error."
This does not leave room for an intersection of your a)-type changes with a b)-change. In other words, as I've already argued, this is incomplete: "Are there “nudges” etc? Of course – some of that will be due to a bit of market imperfection I suppose and some will be due to the fact that changes in respect of a) above are occurring continuously, and can’t be stopped just because b) happens."
And more to the point, as I have argued (here and more directly at MR), this is wrong:
"E.g. the Fed can’t inject reserves in conjunction with a b) change – UNLESS there is a reserve change required that fundamentally reflects an underling a) type incident – e.g. some unusual reserve distribution pattern that appears by coincidence on the day of a target rate change."
JP, ATR:
ReplyDelete“Other banks might be allowed to print money to pay interest on reserves, but I'm pretty sure the Federal Reserve can't create dollars to pay IOR. The Fed has very strict collateralization policies, creating money de novo to pay IOR would probably break those rules.”
I think this is effectively not an issue, based on the following reasoning:
Fed revenue consists mostly of Treasury bond interest and GSE bond interest. Those items result in debits to Treasury TGA and GSE deposit balances at the Fed respectively.
Fed expenses include IOR, which results in a credit to bank reserve accounts.
From an income statement perspective, the net of all that activity (along with other Fed income statement items such as salary expenses) becomes the Fed’s profit.
From a balance sheet perspective, the Fed pays that profit to Treasury as a debit to its own equity position (which has been inflated by the profit) and a credit to the TGA balance at the Fed.
The totality of this shuffling of Fed liability accounts is a net credit to bank reserve accounts and a net debit to Treasury and the GSE’s together - other things equal. That combination in turn nets to zero – other things equal. The redistribution of Fed liability balances alone doesn’t change the size of the Fed’s balance sheet.
But the redistribution does require further action on the part of Treasury. Treasury will concurrently and on an ongoing basis rebalance its net debit (potentially) in the TGA, through taxation and/or borrowing, both of which transfer bank reserves back to the TGA. The GSE’s will do something similar in respect of their own cash positions through their own revenue activities.
The required Treasury compensating action reflects the fact that Treasury on a net basis is still ‘eating’ the cost of the IOR that the Fed must pay to the banks – even though that IOR constitutes an effective reduction in the Fed’s cost of debt servicing (as seen in an inflated level of profit rebated from the Fed to Treasury).
The creation of new reserves and the corresponding expansion of the Fed’s balance sheet is not an IOR issue. It depends on the activity of the Fed as a principal operator in markets – not on its role as agent deposit taker for Treasury, the GSE’s, and the banks. That’s really why reserve creation and collateralization as it relates to IOR directly is a non-issue.
Regarding collateralization, the issue is the degree to which Fed assets are required to collateralize Fed liabilities. And that issue is separate from the accounting dynamic that connects income to the balance sheet. One just looks at the Fed balance sheet at a point in time to see that required collateral is in place. I understand there is a fundamental collateral rule for currency (banknotes) but I’m not aware of such a thing for reserve balances. (System repos obviously require collateralization).
really referring to system reverse repos in that final remark
Delete"IOR constitutes an effective reduction in the Fed’s cost of debt servicing"
Deleteshould be "an effective reduction in Treasury's cost of debt servicing"
yech ...
JKH, if the IOR paid out by the fed exceeded the income gained from the fed's portfolio such that it started making a loss rather than a profit, then wouldn't that be a lid on the amount of IOR paid? You say that IOR causes a deduction in the payments the Treasury receives out of fed profits (that was my understanding too) BUT what happens if IOR overwhelms ALL of those profits? People say "what if interest rates went back up to 6%" - well wouldn't that entail more IOR being paid than the fed was receiving from the Fed's portfolio?
DeleteStone, when JKH wrote:
Delete"Regarding collateralization, the issue is the degree to which Fed assets are required to collateralize Fed liabilities. And that issue is separate from the accounting dynamic that connects income to the balance sheet. One just looks at the Fed balance sheet at a point in time to see that required collateral is in place. I understand there is a fundamental collateral rule for currency (banknotes) but I’m not aware of such a thing for reserve balances. (System repos obviously require collateralization)."
I think he allowed for the hypothetical that a build-up of reserve liabilities could theoretically violate collateralization requirements. He is challenging, however, that there are collateralization rules for reserves specifically. You and JP Koning are effectively asserting there are. JKH questions that (as I originally did, but I took your word for it). Anyways, that's how I read him. Can't this be resolved by some legal research? Beowulf?
My understanding is that the Fed can only create new base money by either purchasing a limited range of assets in the open market, or by lending upon suitable collateral. It can''t create new base money to pay for things not included in the above categories, say buying new pens, paying Yellen's salary, or meeting its interest rate on reserves obligation. To meet these obligations, it needs to earn already-created base money.
DeleteJP,
DeleteTechnically, the Fed can create base money at the margin with any of those transactions. The accounting entry is a debit to Fed equity and a credit to bank reserves - both on the liability side of the Fed balance sheet.
One of my points is that the Fed tends to keep its equity position relatively flat over time - by transferring annual profits to Treasury - and that Treasury in the same way tends to keep the TGA balance flat over time by basic cash management. (This latter part is more correct on a pre-2008 basis when it targeted TGA fairly constantly around the $ 5 billion mark).
So the effect of Treasury profits (which take into account the types of expenses you note) and Fed cash management all net out in the wash, other things equal, so that the base is mostly unchanged as a result.
I.e. I agree with your statement - but it depends on disciplined transfers of profits to Treasury and then disciplined cash management of the TGA balance by Treasury, notwithstanding the potential to affect base money at the margin with both asset transactions and revenue/expense items.
This marginal approach to base creation is a point that Nick Rowe always seems to be making - i.e. not distinguishing between asset transactions and revenue/expense transactions in the potential creation of base money.
JKH, sorry if I'm just failing to grasp what you are saying. Are you saying that the Fed could pay interest on reserves that was greater than the earnings it gained from its portfolio and simply mark up the expense as being a reduction in its net worth? Could the Fed continue that even once it had overall negative equity?
DeleteAgain sorry for being slow witted in understanding what you are saying.
John Hussman has written on his web site that the Fed can only use its income stream from its own portfolio as its source of funds (that was my source). Is he mistaken then?
Stone,
DeleteThat’s OK because I’m doing a rotten job of explaining what it is I’m trying to say.
Start over. Think of the Fed’s balance sheet. Mostly bonds on the asset side. On the other side, we have bank reserves, currency, the Treasury deposit account, GSE deposit accounts, and a few other things. One of those things is the Fed’s equity position (i.e. capital basically).
JP’s comment is generally and materially correct in that the usual process of bank reserve expansion or currency expansion includes the purchase of bonds as a nominal facilitating match on the asset side.
However, in theory and technically, the Fed can change the monetary base – i.e. it can change the system reserve position for example – by adjustments on the right hand side of the balance sheet. In practice, in a normal (say non-QE) environment, this occurs usually only to a small degree and the changes tend to be cumulatively offsetting.
Here’s a wacky example which combines a couple of points. Suppose the Fed accumulates some profit over a certain period (I’ve forgotten how frequently it remits to Treasury). Suppose it retains that profit and doesn’t remit it. That means the equity position must have increased. Then, other things equal, if the size of the balance sheet doesn’t change, some liability account must have declined. For example, that profit may have reflected a marginal increase in revenue from Treasury bonds held – in which case the TGA balance would be debited and decline. That sort of cherry picking of the source of the increased profit is a bit artificial, but it’s that sort of idea.
Another wacky example is to assume a Fed loss instead of a profit. The equity account declines. Some liability account must have increased, other things equal. Again, cherry picking artificially, suppose the source of that loss is attributed to the purchase of too many pencils by the Fed - the tipping point for the loss so to speak. That’s an expense that is paid to the private sector. So the system bank reserve account will have increased according to that association.
Summarizing, monetary base expansion is linked generally and materially to the expansion of the balance sheet though corresponding asset purchases. But it is possible usually in a small way to affect it through liability re-composition. That said, both the remittance of Fed profits to Treasury and Treasury’s disciplined management of the TGA balance (at least on a pre-2008 basis with a $ 5 billion target) tend to stabilize the nominal size of those respective accounts on the right hand side of the Fed balance sheet. That stability is useful for controlling the reserve position at the margin – allowing the Fed to use asset management more exclusively in order to affect the level of reserves and the monetary base in total.
I’m not sure of Hussman’s intended context - but that statement seems questionable on the surface in the context of normal flow of funds language.
BTW, Nick Rowe's approach to this sort of thing is that the CB's purchase of pencils can increase the monetary base just as much as the purchase of a bond. That's consistent with the type of unusual example I presented above.
DeleteThe interesting thing to me is that this is somewhat analogous to MMT paradigm thinking. The idea is to focus on the effect of everything at the margin. MMT doesn't extend that to purchasing pencils, but their thinking on monetary operations is very much "at the margin" focused. That allows them to come up with statements such as "spending always comes first". This is the result of simply ignoring normal practice cash management constraints such as practiced by the pre-2008 Fed targeting $ 5 billion in TGA. Nick extends that sort of thinking to ignoring normal capital management practices - in this case the remittance of profit by the Fed to Treasury according to regular accounting cycles.
Thanks JKH, the contex of John Hussman's point was that in his view the Fed couldn't pay unlimited interest on reserves because they wouldn't be able to fund it over and above the income they were receiving from their portfolio.
DeleteI just had a quick look and I failed to find where he wrote that. I do hope I'm not just misremembering.
Stone,
DeleteI’m guessing Hussman’s point may be consistent with the following:
If reserves are sufficiently large enough and short term interest rates move sufficiently high enough, the Fed will start to generate losses due to IOR pressure. That would be the result of paying an increasing IOR rate, while assets in place are mostly long term at a lower fixed rate. It takes a considerable move in rates to do this because the effect of a higher IOR also has to be sufficient to net wipe out the benefit of paying a rate of zero on currency outstanding.
So he’s saying that if the Fed starts to incur losses, they wouldn’t be able to “fund” the full IOR. That’s not the best use of the term “fund” IMO, because the IOR effect is about revenues and expenses on an income statement rather than funding (i.e. creating a liability on a balance sheet).
What he’s saying is also debatable and probably doubtful. There’s actually a Fed accounting facility in place, ready to go, that will allow for temporary losses by creating “negative liabilities” instead of negative capital on the Fed’s balance sheet. The idea of a negative liability is that sometime further down the road, when the Fed would start earning a profit again, it would withhold the profit that it normally remits to Treasury – until that negative liability is unwound. Think of that negative liability as an item that would normally be a future liability (the normal remittance of projected future profits sometime down the road) that is reversed out by retaining rather than remitting those same projected future profits. Some tricky accounting, but the projected economics do make sense – because negative yield curves don’t last forever and the Fed always has the marginal benefit of paying a zero rate of interest on currency. Yield curve developments would inevitably bring the Fed back on side with respect to Hussman’s concern - and that accounting modification recognizes that.
I sympathize somewhat with Hussman. Yes, the Fed can record negative income if interest on reserve payments rise above revenues. But if we move over to a hypothetical cash flow statement, it can't legally fund this operating loss by creating new base money. The Fed would have to get a line of credit or something from a commercial bank in order to fund its deficit, since this would provide it with already-existing base money. If negative profits comes from a large writedown, however, then the issue would be moot since this is a non-cash expense.
DeleteAs I said, an accounting mechanism is already in place:
Delete“There’s actually a Fed accounting facility in place, ready to go, that will allow for temporary losses by creating “negative liabilities” instead of negative capital on the Fed’s balance sheet.”
That’s not my invention – the Fed has announced it. Presumably it was subject to legal vetting as a condition of the announcement.
Such an accounting allowance constitutes the “funding” of a loss on the balance sheet (using the term "funding" in his sense). If you view marginal IOR payments as the cause of the loss, it creates new reserve balances.
The cash flow statement (sources and uses of funds) would reflect an increase in reserves as the source of funds and a decrease in a newly created liability category as the use of funds. Those reserve balances could be drained if desired by the sale or reverse repo of bonds or blended in with QE.
With an asset write-down, the issue is moot with respect to the creation of reserves, but not with respect to the problem of dealing with a loss. If the write-down is large enough, it would result in an loss just as much as a cash flow deficit. If the size of the resulting loss exceeds the Fed’s equity, it might be able to use that same accounting facility I just to.
The negative liability accounting is a way of swapping current equity losses for future equity gains – by projection the retention of future earnings sufficient to reverse the negative liability. If this ever happens, it will become extremely controversial. Not many people caught it when it was announced, but it raised eyebrows. The Fed would essentially be circumventing the alternative of seeking recapitalization. Without it, it’s straightforward that the Fed would need to seek recapitalization or permission to run negative equity or some other type of “rescue” – and that would be consistent with Hussman’s point.
I agree with everything you said, but I'm not convinced about this:
Delete"The cash flow statement (sources and uses of funds) would reflect an increase in reserves as the source of funds and a decrease in a newly created liability category as the use of funds."
My bet is that the cash flow statement would show a decrease in cash flow from operations, the use of funds, and a counterbalancing increase in private sector borrowing by the Fed, not reserve issuance. The source of funds boils down to a legal question. We'll probably have to wait till the actual event transpires before we'll ever settle this.
I doubt that actual event will transpire. The interest rate path should be moderate enough and the currency component large enough for the Fed to absorb an increasing IOR rate. Also, they're not planning on selling bonds that would otherwise force marked to market losses.
DeleteP.S.
ATR, I'll get back on the other later - could be a week or so, due to unrelated matters
Okay, no worries. BTW, in my last comment, I was referring to my second 12:32PM post, but I rephrased it in the Appendix above anyways.
DeleteThanks for the informative side conversation JKH, JP Koning, and stone.
Thanks for the info about the Fed's "borrowing from the future" facility.
DeleteYou do say though that it would be controversial if they used it and you also say that they are unlikely to reverse QE sharply enough for them to need to. I'm still left thinking that that threat of it being controversial might in itself be acting to convince the market that the Fed is particularly unlikely to raise interest rates a lot in the next few years and that in itself becomes priced in as higher asset prices. I'm still left thinking that this is a rational reason for QE to act as a support for asset prices (rather than cash being a supposed hot potato or QE a supposed signal for inflation intentions by the Fed).
If the whole point of QE is to dispel any lurking fear of there being any tail risk of a Volcker 1981 style interest rate hike, then we need to wonder whether IOR could ever give such a Volcker style rate hike in a post QE world.
DeleteI'm still struggling to get my head around what it would be like if the Fed tried to do Volcker 1981 style 20% interest rates in a post-QE world. Let's imagine the Fed starts with $3T as bank reserves that it has to pay IOR on. That would double to $6T in less than four years if they were paying 20% IOR. They would then be paying out $3B per day or whatever as IOR and in the process adding to the colossal mountain of bank reserves that was the source of their headache and all the while their portfolio of bond holdings would be becoming ever smaller in relation to the stock of reserves that they would have to pay IOR for. I think there could be a real danger that it would be seen as unsustainable and there would be a flight from the USD.
Are you really saying that the Fed could do a Volcker 1981 level of rate hike in our post QE world by paying IOR?
ATR,
ReplyDeleteTabula rasa question:
From Woodford:
"It is frequently observed now that announcements of changes in the Fed’s operating target for the funds rate (made through public statements immediately following the Federal Open Market Committee meeting that decides upon the change, under the procedures followed since February 1994) have an immediate effect upon the funds rate, even though the Trading Desk at the New York Fed does not conduct open market operations to alter the supply of Fed balances until the next day at the soonest. This is sometimes called an “announcement effect”.
How do you interpret this statement?
I interpret it exactly as I’ve explained multiple times. The market may move to the new target rate upon an announcement and prior to the central bank taking any necessary action, but this is predicated on expectations of the central bank eventually taking those necessary actions. I know we both agree on this in a general sense. But specifically, the difference between you and Fullwiler vs. me and Woodford is that you think any quantity-related actions are fundamentally of your a-type:
Delete“The point is that from an analytical and pragmatic operating perspective, the ongoing supply curve changes described in a) are of a completely different breed than any supply curve requirements that might be suggested by b) alone – and there are none in the case of b), once you separate out the effect of a) and any minor market noise otherwise.”
Versus me and Woodford, who disagree with you and believe there are, in theory and in reality, independent, b-type liquidity effects. I’ll supply you several quotes from the Woodford paper to support that argument (not to mention his model), including a segment where he even directly discusses your a-type operations and places it in distinction b-type operations.
First, though, to summarize what I’ve done, I gave 4 different examples above of what a central bank might have to do enforce a rate change, even assuming the market moves to where it wants it prior to any action taken. One involved only changes in rates (Figure 3). Some involved only changes in quantity (Figure 2). One involved the more obscure and theoretical reserve requirement change (Figure 4). The last involved both a rate change and a quantity change (Figures 5 and 6), intended to be somewhat similar to what the Fed may have had to do pre-2008.
Figure 3 was an illustration of the channel-type system in the Woodford paper that doesn’t require a change in reserve quantities to enact a rate change. See
“Thus under a channel system, changes in the level of overnight interest rates are brought about by simply announcing a change in the target rate, which has the implication of changing the lending and deposit rates at the central bank’s standing facilities; no quantity adjustments in the target supply of clearing balances are required.”
Note that Woodford then goes on discuss your a-type operations:
“Open-market operations (or their equivalent) are still used under such a system. But rather than being used either to signal or to enforce a change in the operating target for overnight rates, as in the U.S., these are a purely technical response to daily changes in the Bank’s forecast of external disturbances to the supply of clearing balances, and to its forecast of changes in the degree of uncertainty regarding payment flows.”
This all implies there can be and are b-type liquidity effects, independent of a-type operations which are always occurring in the background. He in fact asserts that was the case “as in the U.S.” (prior to 2008). That is to say, central banks may sometimes change reserve quantities, without the intention of reversal (all else equal), in order to enact a rate change or enforce it if the market moved there upon announcement. That is to say, this is not an a-type operation. He asserts this again here:
“And even in the case of an operating target for the overnight interest rate, the target is not likely to be most reliably attained through daily open-market operations to adjust the aggregate supply of central-bank balances, the method currently used by the Fed.”
Context makes clear he is not talking about a-type changes. And note the sentence following your citation:
“Taylor (2001) interprets this as a consequence of intertemporal substitution (at least
within a reserve maintenance period) in the demand for reserves, given the forecastability of a change in the funds rate once the Fed does have a chance to adjust the supply of Fed balances in a way consistent with the new target.”
Again, he’s not talking about an a-type change.
What I quoted from Woodford stands on its own, without qualification. He is very clear here. He defines the announcement effect in the case of the pre-2008 Fed as I have, and specifies quite unambiguously that the market moves in response to the announcement effect without any concurrent injection of reserves. That is EXACTLY what I have said and how I have defined a b) type reserve situation. That description is what Fullwiler agreed with – and what Woodford himself agrees with here – and what accords with my b) definition. The next sentence refers to reserve adjustments starting the day after an announcement – which by my own definition are necessarily a) type changes under a new target. His statement is entirely consistent with I’ve said. And any other reference elsewhere from Woodford obviously has to be consistent with this – or he is contradicting himself – and I don’t think he would want to do that. I have given an example in my MR comment of how one of those statements is quite consistent - where he uses the phrase “implementing or signaling”. The basic point of fact is that the market moves immediately in response to an announcement without requiring accompanying reserve injection – that is what I have said and that is what he says here very explicitly. The market moves in response to the announcement - and the Fed doesn't adjust reserves further until the next day - which by definition is an a) adjustment under a new target.
DeleteBTW, I’ve had to remind you a number of times that I have been writing almost exclusively about the pre-2008 Fed system – not a symmetric channel system. And you keep coming back with responses that inevitably refer to a symmetric channel system when I'm referring explicitly to the Fed system. You’ve done it again here with some of your commentary. You are not only conflating a) and b) type moves, but two different systems as well, in responding to a point that Fullwiler and I – and Woodford in this case – are making about the pre-2008 Fed system. You don’t seem to be able to separate the a) and b) idea as I specified them – in the case of the pre-2008 Fed system. Fullwiler and Woodford clearly have.
I doubt you're going to turn this around on me. I’m not confusing your a/b changes as you’ve clearly described them, and I’m not conflating systems. I bring the different systems up in relation to your pre-2008 comments for very specific and relevant purposes, which you’re not understanding. Furthermore, I clearly follow Woodford et al., and I’ve spent quite a bit of time studying this paper. Other than citing select qualitative passages from his paper that you think confirm your views, you haven’t yet demonstrated a comprehensive understanding of what he’s saying – and I wouldn’t expect you to, since this is probably the first time you’ve seen it. Like Fullwiler, you haven’t engaged with the model – or any model - and what it implies for the role of administered rates vs. reserves. And, as a personal decision, I will remain skeptical of your position until you do so, because until that model is understood, one cannot see the whole picture. For reasons apparent to me but not you, I disagree with you and Fullwiler but not Woodford et al. Could it be that you’re that much more sophisticated than Woodford, and I simply cannot comprehend? Or that you’re not being clear enough? Maybe, but I really don't think so.
DeleteAnyways, sure it stands on its own, but more unpacking must be done to see how well it truly aligns with your full beliefs, because both you and he have written much more to explain what's behind that statement. This debate is not merely about the clearly observable fact that rates move upon announcement - it's about what's behind that and what is to come.
“The next sentence refers to reserve adjustments starting the day after an announcement – which by my own definition are necessarily a) type changes under a new target."
Well that’s not how I read your definition. You've defined a-type changes under a new target as those in "response to changes in the distribution of reserves in the system and the effect that those shifting distributions have on the effective ease or tightness in the system and therefore on the actual trading rate for fed funds." Another definition you give is “– e.g. some unusual reserve distribution pattern [that appears by coincidence on the day of a target rate change].” That is not the type of post-announcement reserve change that Woodford is referring to there – I’m not sure how many times I can say that, and unless you learn Woodford’s model, you might not understand this. It has nothing to do with ‘changes’ in the distribution of anything. Rather, what he’s referring to is a necessary change in the aggregate quantity of reserves such that, in accordance with a probability density function of expected liquidity shocks (that we can hold constant for sake of argument), the target rate is enforced by fulfilling the market’s expectations. I brought up the following quote “Open-market operations (or their equivalent) are still used under such a system…. these are a purely technical response to daily changes in the Bank’s forecast of external disturbances to the supply of clearing balances, and to its forecast of changes in the degree of uncertainty regarding payment flows” not because it had to do with a symmetric channel – these types of OMOs are applicable in any system - , but because those are your a-type changes and Woodford clearly states they are DIFFERENT than the type of quantity changes he’s referring to pre-2008 Fed rate changes. So I assert you are wrong: he’s not talking about a-type changes as you’ve defined them.
And your description of b-type changes is clearly wrong here, according to Woodford’s view:
Delete“E.g. the Fed can’t inject reserves in conjunction with a b) change – UNLESS there is a reserve change required that fundamentally reflects an underling a) type incident – e.g. some unusual reserve distribution pattern that appears by coincidence on the day of a target rate change.”
Again, this doesn’t have to do with an underlying a-type incident as you describe it there, as I explain above. If your point is merely that the Fed will change reserves the next day, since they announce it after the day’s OMOs are done, and thus it is not “simultaneously in conjunction,” that is completely trivial. In any case, what you write clearly implies you mean something else – having to do with unusual reserve distributions, which misses the mark, and an all-powerful shifting of the demand curve, which you are also wrong about:
“Because if the Fed has hit its previous target at the time of the target change, the demand curve drops vertically and any additional reserves put into the system will cause the rate to drop below the NEW target – just follow the geometry I described earlier. That is why even with a b) type “nudge” injection, it must be quickly reversed, other things equal.”
Again – this is only correct all in the symmetric channel system I brought up. That’s why I brought it up. And you don’t realize that. Other regimes may require additional reserves THAT WILL NOT cause the rate to drop below the new target, and it doesn’t have to do with a-type changes. Figures 5/6 is an example.
OK, I'll bite:
Delete"Rather, what he’s referring to is a necessary change in the aggregate quantity of reserves such that, in accordance with a probability density function of expected liquidity shocks (that we can hold constant for sake of argument), the target rate is enforced by fulfilling the market’s expectations... Again – this is only correct all in the symmetric channel system I brought up. That’s why I brought it up. And you don’t realize that."
Can you point me to where his model indicates this and applies it explicitly in the case of the pre-2008 Fed system. Show me exactly where he demonstrates that reserves must be adjusted in the way you describe - for the pre-2008 Fed system - that the pre-2008 adjustment must include a reserve change in a way that a symmetric channel system doesn't. And I need to see something where your own interpretation doesn't contradict that first quote I provided it. Maybe I missed it or didn't look closely enough. And if it requires the math, just point to or maybe you can also summarize how the math must dictate a different outcome for the pre-2008 Fed situation.
Thank you for biting. I apologize in advance for the length of the summary to come. Complicated stuff usually requires more explanation, although I admit I am not the best at summarizing. BTW, I’ve done extended posts on this model and the simpler but related Bindseil aggregate liquidity management model, found in this TOC http://macromoneymarkets.blogspot.com/2013_11_01_archive.html. I provide explanation for the math in the post and in an appendix, if that’s needed. I also do an extended post on Woodford’s model, which I’ll try to summarize here.
Delete“Can you point me to where his model indicates this and applies it explicitly in the case of the pre-2008 Fed system.”
DeleteIt’s all contained in Section 2.2 – don’t be put off because it’s heavily focused on a symmetric channel. Stick through it with me. I’ll highlight key points and answer your questions.
First see: “A simple model of the determinants of the demand for clearing balances can be derived as follows. To simplify, we shall treat the interbank market as a perfectly competitive market, held at a certain point in time, that occurs after the central bank’s last open-market operation of the day, but before the banks are able to determine their end-of-day clearing balances with certainty. The existence of residual uncertainty at the time of trading in the interbank market is crucial; it means that even after banks trade in the interbank market, they will expect to be short of funds at the end of the day with a certain probability, and also to have excess balances with a certain probability. Trading in the interbank market then occurs to the point where the risks of these two types are just balanced for each bank.”
He then goes on to introduce a random variable for the liquidity shock and its cumulative distribution function (I chose to refer to that random variable’s probability density function in my previous comment, same difference). Note – this is all you need to develop the model, and it applies to any type of policy regime, whether we’re talking a floor system, asymmetric channel (Fed pre-2008), symmetric channel, ceiling system, etc. So even though he focuses on its properties as they relate to the operation of a symmetric channel, we can easily use it for any of the others.
Woodford then does a little bit of modeling to arrive at an equation that tells you what the overnight rate should be given the settings of the ceiling/floor rates, the quantity of reserves, and the probability distribution of the random variable for liquidity shocks. I’m not sure what version of the paper you have, but in this one (http://www.nber.org/papers/w8674.pdf), it is equation 2.4. If you algebraically rearrange the equation, there’s an easier interpretation of this formula: i = P(“short”)*ceiling rate + P(“long”)*floor rate (Bindseil shows this and explains how it’s the same result given by his simpler aggregate liquidity model I mentioned above http://www.nhh.no/Files/Filer/institutter/for/dp/2007/1007.pdf).
The P(“short”) and P(“long”) mean the probability that banks expect to end short or long reserves, which is determined by the intersection of the probability distribution of liquidity shocks and the quantity of reserves that the central bank ultimately provides to the banking system prior to the end of the maintenance period. In other words, the more reserves the central bank provides the banking system, the less likely they’ll end short; vice versa. The exact probabilities are determined by the precise intersection of the quantity of reserves with the probability distribution (you could imagine modeling the liquidity shock random variable using a normal distribution for example, with the center perhaps being 0). So random example: if ceiling rate = 5%, floor = 0%, and the quantity of reserves is such that the P’s are both 50%, then the interbank rate i should equal 2.5%.
**So there are 3 key determinants of rates given a fixed probability distribution of liquidity shocks: the floor rate, the ceiling rate, and the aggregate quantity of reserves.** I am holding the random variable fixed so as to put aside your a)-type reserve changes and any associated technical market irregularities that come along with it.
Okay, pause.
DeleteSo we now have an equation that shows us the levers the central bank can pull to achieve any interbank rate it desires. Because banks understand this, the interbank rate will jump to any new target rate upon announcement, as banks will anticipate that the central bank will eventually pull those levers. *But the central bank still needs to pull those levers, subsequent to any announcement-induced jump,* and doing so will not cause the interbank rate to under or over-shoot target (holding constant the probability distribution). Quite the opposite, the interbank rate will only be profit-maximizing for banks if those levers are eventually pulled. This does not contradict your tabula rasa quotation from Woodford. All that quote says is that we observe rates to move upon an announcement. I’m explaining why that occurs and what the central bank must do afterwards. Okay, now I’ll explain when the central bank needs to pull the rates lever, quantity lever, or both.
In your definition of announcement effect, you’re including an actual change in the ceiling/floor rates. That’s fine – I can run with that. As we can see in the equation above, changing those indeed should drive a change in i. But note there are other components to the equation: P(short) and P(long). These are functions of the aggregate quantity of reserves, holding the probability distribution of liquidity shocks constant. *If* changing the floor/ceiling is enough to spit out the target i according to that equation, then the announcement effect, as defined by you, is enough to maintain the new target rate (again, assuming away the need for background a)-type operations). Because after all, your definition of the announcement effect includes changing the floor/ceiling. In this case, subsequent injections/removals of reserves will cause the interbank rate to overshoot target, as you describe in your case b). However, if the probabilities that banks end short/long also need to be changed to hit the target I according to that equation – again, assuming a constant probability distribution - , then this requires changes in the quantity of reserves subsequent to the announcement. A failure to do so will mean that the rate will deviate and the central bank will lose its credibility. We’re not talking about a)-type operations here. We’re holding all of that constant.
Now that we have this model, we can consider your following request, and parse when ceiling/floor rate vs. quantity adjustments are necessary:
“Show me exactly where he demonstrates that reserves must be adjusted in the way you describe - for the pre-2008 Fed system - that the pre-2008 adjustment must include a reserve change in a way that a symmetric channel system doesn't.
DeleteOkay, first we need to agree that pre-2008, the system was not symmetric. Pre-2008, the discount rate was at some fixed spread to the target rate. After 2003, the primary credit discount window was usually 1% above the target rate. Therefore, the distance of the lending and deposit rates relative to the target overnight rate was constantly changing (e.g., floor of 0%, target 4%, discount 5% vs. floor of 0%, target 3%, discount 4%).
Why is this important? Well let’s assume a constant, normal distribution of liquidity shocks, centered at 0, meaning there is a 50/50 chance banks get hit with a positive or negative liquidity shock. Let’s also assume as a starting point that the central bank has supplied the market a quantity of reserves equal to required reserves. What does our equation tell us about where the market rate should be? i = 50%*floor + 50%*ceiling. In other words, the market rate should be halfway in between the floor and ceiling.
So putting pre-2008 Fed on hold for just one second, consider a symmetric channel system, where the target rate is always halfway in between the floor and ceiling by design. In that case, there’s no need to toggle with the quantity of reserves, putting side a-type operations. The central bank can set i = target simply by announcing changes in its floor/ceiling rates. Thus the following quote from Woodford:
“This means that an adjustment of the level of overnight rates by the central bank need not require any change in the supply of clearing balances, as long as the location of the lending and deposit rates relative to the target overnight rate do not change.”
Clearly, Woodford is not ruling out the need for quantity adjustments at all in this case, as he goes on to say. There is still the need for a-type operations, as you call them. But by inference, this implies in non-symmetric channels, quantity adjustments might ALSO be needed, in addition to rate changes, which are SEPARATE from your a-type operations.
So now consider applying this logic to the Fed pre-2008. Consider my Figures 5/6 and the math I show using Woodford’s model in my 12:32PM comment above. I won’t repeat it here. Because the target rate for the Fed pre-2008 was not located the same distance from the ceiling and floor (0%), quantity adjustments were necessary, even outside of a-type operations.
So if you want to reject Woodford's model - fine. But then you need to contend with the model, articulate the flaws you perceive, and then ideally propose something better.
DeleteJKH, Nick Edmonds' recent post (http://monetaryreflections.blogspot.co.uk/2014/04/asymmetric-redeemability-in-woodfords.html) on that same Woodford paper we were discussing led me to find this other statement in his paper, in which he very clearly asserts that he thinks quantity changes drove rate change for the pre-2008 Fed:
Delete"Thus it would be essential to move from a system like that of the U.S. a
present... to one in which instead variations in overnight rates are achieved purely through variations in the rate paid on Fed balances, and not at all through supply variations."
That clearly says he believes that the pre-2008 Fed changed interest rates through supply variations. The full quote is this
"Thus it would be essential to move from a system like that of the U.S. a
present — in which variations in the supply of Fed balances is the only tool used to affect the overnight rate, while the interest rate paid on these balances is never varied at all — to one in which instead variations in overnight rates are achieved purely through variations in the rate paid on Fed balances, and not at all through supply variations."
Actually, I'd slightly disagree with him there, because my understanding is that changes in the FFR pre-2008 went hand-in-hand- with some asymmetric change in the discount rate, which by itself should cause a new equilibrium FFR, independent of quantity changes. But if we assume changes in the discount rate away, I'd be fully on-board with him.