Operating Procedures

The Fed sets interests rates. But how does the Fed set interest rates? The Fed is undergoing a big review of this question. We had a little workshop at Hoover, in preparation for the larger May 3 Strategies for Monetary Policy conference, which provokes the following thoughts.

Issue

Here is the issue.

The graph plots the demand for reserves, as a function of the interest rate on other short-term assets such as overnight federal funds, Libor, money market rates, and so on.

(Reserves are accounts that banks have at the Fed. The Fed sets the interest rates on such accounts.)

The lower horizontal line is the rate the Fed pays on reserves.

If the interest rate on other similar assets (overnight federal funds, Libor, repo rates) is above the interest rate on reserves, then banks should want to get rid of reserves. However, reserves are useful, as money is useful, so banks are willing to hold some even when they lose interest on reserves by doing so. The greater the interest costs -- the greater the difference between the rate banks can lend at and the rate they get on reserves -- the more they work hard to avoid holding reserves. At the end, there are legal and regulatory requirements to hold reserves.

In the flat zone, banks are satiated in reserves. Reserves don't have any marginal liquidity value. But banks are happy to hold arbitrary quantities as an asset so long as the interest on reserves is above or equal to what they can get elsewhere.

If banks can borrow at less than the interest on reserves, they would do so and demand infinite amounts. Therefore, competition among banks should drive those rates up to the interest on reserves.  Similarly, if rates banks can lend at are higher than interest on reserves then banks should compete to lend, driving other rates down to the interest on reserves. Therefore, the Fed by setting the interest on reserves sets the overall level of overnight interest rates.

Questions

Here are the questions:

1) Where should the supply of reserves be? This is the biggest question the Fed is asking right now. The three vertical lines in the graph are three possibilities.

The Fed currently fixes the supply of reserves, which is referred to as the "size of the balance sheet," so the lines are vertical. The Fed raises the supply of reserves by buying assets such as treasuries or other assets, "printing money," i.e. creating reserves, in return for the assets. The balance sheet shows the assets (e.g. Treasuries) against liabilities (reserves and cash). Yes, the Fed is nothing more than an enormous money market fund, offering fixed value floating rate accounts which it backs by treasury and other securities.

The debated is couched as "floor system" vs. "corridor system."  A "floor system" refers to the two supplies on the right, where there are so many reserves that the other interest rates will equal the rate on reserves.

There are two floor-system variants: abundant reserves, with the supply well to the right, and minimalist reserves, with the supply of reserves set to the smallest possible level, where the demand curve just hits the lower bound, "satiation" in reserves. The latter seems to be where the Fed is heading -- a minimal-reserves floor system.

In a "corridor system," the Fed has an upper and lower band for the market interest rates it wants to target. Historically this was the Federal funds rate, which is the rate at which banks lend reserves to each other overnight. It tries to place that interest in the middle of the band, by artfully putting the supply of reserves in the downward sloping component. This is how the Fed operated before 2008.

The rate at which the Fed is willing to lend reserves also provides an upper bound, which I'll get to in a minute.

2) If there is going to be a corridor, which rate should the Fed care about? The (justly) moribund federal funds rate? The overnight general collateral repo rate? Libor? One advantage of the abundant floor, is that the Fed can stay quiet about all this and let the market sort out just what kind of overnight lending it prefers.

3) If there is a band, how wide should the range between the upper and lower bound be?  1%? 0.5? 0.25%? 0.01%?

3) How free should lending and borrowing be? Who gets access to interest paying reserves, and how much interest do they get? Who can borrow reserves, and on what therms -- what collateral is acceptable, is it overnight or term borrowing, does such borrowing incur formal regulatory attention or informal "stigma"?

4) What assets should the Fed buy on the other side of the balance sheet, or accept as collateral if it lends reserves?  Just short-term treasuries? (My favorite) The current mix of long term treasuries and mortgage-backed securities? Or, perhaps, follow the ECB and BOJ and buy corporate bonds and stocks, many countries debts, or lend newly created reserves to banks and count the loans as assets?

The motivations here are, I think, as much political as economic, and it's better to acknowledge that. (We should understand the Fed can't do that in writing, but we can!) Having touted QE as extraordinary accommodation the Fed is under big pressure to stop stimulating. It's too late to say that QE was mostly symbolic. Having seen the Fed buy all sorts of securities, congresspeople are coming up with dandy ideas for new things the Fed can "invest" in by printing money. Having paid banks about a quarter point more than they can get anywhere else, and indeed allowed a pleasant little arbitrage to go on, the Fed is under pressure to pay other investors the same. Congress is even more full of ideas for who the Fed should lend to, and how the Fed should use its expanded regulatory powers to channel credit here and deny credit there.

"Normalization" is a pretty meaningless economic term to me -- why is whatever the Fed was doing in 2007 "normal," why is it good? But "normalization" is a tremendously useful marketing banner. We're going back to "normal," so leave us alone with your bright ideas. Well, fine, but let us quietly  go to a new normal that incorporates all the interesting things we've learned in the last 10 years.

My answer

My (radical as usual) answers:

I like the "floor" system, with abundant reserves. The great lesson of the last 10 years is, we can live the Friedman rule. We can have money that pays full interest, so that holding money has no opportunity cost, and this will not cause inflation. This is genuinely new knowledge. Liquidity is free! There is no need for people to waste time and effort on cash management. Liquidity is good for financial stability too: Banks holding huge reserves don't fail.

I go beyond the abundant floor: The Fed should not target the supply of reserves at all. The supply curve of reserves should be horizontal.  The Fed should just say, "bring us your treasuries, and we'll give you reserves and pay the IOER rate." Or, "Bring us your reserves and you can have treasuries."

Why? Well, if you want to target a price, you offer to buy and sell freely at that price. If you want to target an interest rate, target an interest rate. We have seen limited arbitrage between reserves and other assets due to lack of competition in banking and Fed restrictions, who can hold reserves, and the fixed supply.

I see no economic or financial harm whatever from arbitrary expansion of the Fed's balance sheet, if the assets are all short-term Treasuries. Reserves are just overnight, electronically transferable government debt. If the banking system wants more overnight debt and less three week to six month debt, let them have what they want. I see no reason to artificially starve the economy of overnight debt.  The Fed offers free exchange between cash and reserves; the government as a whole should offer free exchange between short term treasuries and overnight treasuries, i.e., reserves.

To accommodate the economy's desire for ample reserves, and the Fed's desire not to provide them, the Treasury should offer the same asset, and the Fed should encourage this and work with the Treasury to make it happen. 

Specifically, the treasury should issue overnight, fixed-value ($1), floating-rate, electronically-transferable debt. Let's call it treasury electronic money. Legally, this is treasury debt that any individual or financial institution can hold, just as they can hold treasury bills or treasury coins. Functionally, these are interest-paying reserves. Like reserves, but not even like T bills, these can be bought or sold immediately: Owners can transfer their ownership of $1 worth of treasury money to someone else on the treasury website, and owners can sell $1 worth of treasury money and have the money wired (i.e. the treasury sends $1 of reserves to the owner's bank) instantly. (Details here.)

Given the Fed's resistance to narrow banking, and the potential of treasury electronic money to undercut bank's (subsidized) deposit financing, I suspect the Fed's first instinct would be to fight such an innovation. The Fed should overcome that instinct and welcome a solution to the problem of providing lots of liquid assets without the (genuine, below) downsides the Fed feels about a large balance sheet.

I agree with critics that the composition of the Fed's assets should return quickly to short-term treasuries only, and in my ideal world to just this treasury electronic money. That is mostly for political economy reasons outlined below. Other assets should be on the balance sheet in emergencies only.

If the Fed feels the need to buy long-term treasurys or take them as collateral, issuing reserves in return, because of a shortage of safe assets, that means the Treasury has not issued enough short-term liquid treasurys. There are simpler ways to fix that problem. 

Other answers



At our conference, Taylor and Selgin argued for a corridor, i.e. the leftmost system with the least supply of reserves, and a fairly large band for the Federal Funds rate. Taylor would like, I think, interest on reserves to go back to zero, and reserves to go back to something like the historic $10 billion level, just scraping reserve requirements.

The Fed, as above, seems to be converging to a vertical supply curve artfully placed as far to the left as possible in the satiation region. And I want it far to the right and flat.

Why this desire for minimizing reserves and a vertical supply curve? There are interesting economic and political arguments.

Limiting money supply?

I think the deepest economic argument for a corridor, the left-most supply curve, is the feeling that it is important for the central bank to limit the supply of money, and not purely to target interest rates.

Price level determinacy is a tricky subject. The recent (post Milton Friedman, mid 1960s) view of inflation starts with MV=PY and controlling M. Friedman argued against a pure interest rate peg, saying inflation would spiral out of control. The Fed has an interest rate target, which seems to say that's irrelevant. However,  one can look at the above graph, and say that the Fed just moves the money supply to hit its desired interest rate, so even with an interest rate target, MV=PY determines the price level.

But money has disappeared from more recent economic thinking. My preferred model of the world (fiscal theory of monetary policy)  has an interest rate target, which sets expected inflation, fiscal theory which sets unexpected inflation, and money is not needed.  Conventional academic wisdom uses new-Keynesian models with `active' interest rate rules to produce determinacy. Older-school ISLM style models, which are still used by the Fed and capture  completely the verbal explanations Fed officials offer for monetary policy, also are based entirely on interest rate targets.

You can recognize the  presence of money in these models, but it doesn't do anything or contribute to price level determination or any other effect of interest rate policy.  (Maybe we should say that, as it's no longer "monetary" policy.) The Fed supplies any required money passively.

Now, one should always distrust the latest fashions in economic. models, and innovate on historically successful policies slowly. I understand how one can be nervous about casting off any pretense of monetary control, and counting entirely on an interest rate target to control inflation, no matter how much talk and modeling has gone that way since about 1982 when the Fed abandoned money targets in favor of explicit interest rate targets.

On the other hand, I don't know of any. written down model in which an interest. rate target and something like a vertical. reserve supply interact to. produce. a determinate price level. So it seems a bit of a historical memory rather than an old and coherent theory preserved.

"Normal?" the ancien regime. 

Let's review the pre-2008 system, which the corridor envisages going back to in whole (Taylor, zero interest on reserves and an inactive discount window) or in part (Selgin, I think, interest on reserves above zero but substantially below interest on other assets, a wide band between interest on reserves and an upper bound) or just touching (Fed, minimal floor).

This is the same graph, with zero interest on reserves. In the old regime, the trading desk at the New York Fed would wake up every day, try to figure out where the demand for reserves was, and set the supply so that the interest rate came out to be the Fed's target. Then it closed up shop, and left the supply of reserves alone for the rest of the day.

In this way, you might think that the Fed really is still setting M in MV=PY, just using interest rates as a guide to where to put the M, and you might be attracted to the corridor on those grounds.

But I don't think it worked that way. The first problem is, that money control only lasts a day. There was a lot of variation in money demand, leading to variation in interest rates. As one indication, here is a plot of the daily federal funds rate from Jim Hamilton in the JPE who explains all this:

Yes, this is how interest rate targets actually worked out. It turns out that even from 10 AM to the rest of the day, the money demand curve in my little drawing can shift left or right, leading to one-day spikes in the funds rate. This pattern persisted right up to the end of "normal" in 2008. (An interactive graph here, Fred's daily federal funds rate. A different nutty pattern has shown up recently.) We'll get to the obvious point that this sort of volatility and more could reemerge under a corridor.

As as a result, the supposedly fixed money supply only lasted a day. In the dashed line here, I show a shift in the reserve demand curve leading to a one-day interest rate spike. What happens then? Well, the next morning the New York Fed wakes up, says "darn," and moves the supply curve. For clarity, I moved the demand curve again, to the right, and you see the new daily supply curve.

Money demand does vary around a lot. The V(i) in MV(i)=PY is not a number but a very random variable.

The trouble is, at anything more than a day horizon, this shifting target is exactly the same as a horizontal money supply curve, as drawn in the graph. How is this different from the Fed just announcing, "We will lend you reserves at the target rate," and you get any amount of reserves at that rate? OK, you have to wait a day to get the loan, but that's not going to calm down Milton Friedman.

In fact, a flat supply curve is exactly what the Fed was founded to provide. Money demand does vary around a lot. The V(i) in MV(i)=PY is not a number but a very random variable. So if you keep M fixed, you get a lot of i variation in the short run, and presumably a lot of needless PY variation in the long run.  The Federal Reserve Act says "provide an elastic currency" with a fixed M provided by gold did not do. Interest rate targets do.

But a flat supply curve, at month to year horizons, is exactly what Milton Friedman warned against in his 1968 address, arguing against interest rate targets. Suppose the shift in demand curve comes from an increase in inflation, a higher P in MV=PY. Then, by having a flat demand curve, the Fed raises M every time P goes up, and the price level can spiral away. There is no "anchor." Classic doctrine: an interest rate peg leads to spiraling inflation or deflation. (As a reminder, classic doctrine was dramatically proved wrong in the zero bound era, but stick with it for now.)

Enter the Taylor rule. One can read the Taylor rule as solving this conundrum in MV(i)=PY models. (The Taylor rule seems to work well in lots of completely different models, one of its many virtues.)

There is a difference between a shift in the money demand curve coming from inflation, and a velocity or money demand shock: you can measure inflation. Suppose when the Fed sees a higher price level it raises the interest rate target, as shown. The higher price level will also induce a greater money demand, to the blue curve. Now the effective money supply curve is upward sloping, rather than flat.  (Take a derivative to understand the Taylor rule as conventionally expressed in terms of inflation.) When the Fed does not see a higher price level, it accommodates, giving the flat red supply curve.

An interest rate target that varies with the price level or inflation rate has all the advantages of a peg -- it provides elastic money when there are velocity shocks -- but it avoids the monetarists' complaint, in that it does not accommodate higher price levels. We can have our  cake and eat it too.

I don't think Friedman would be happy, as he might say inflation happens with long and variable lags, and the money demand shift happens before you can see the inflation. But Friedman is no longer with us, and doesn't get the chance to modify his views with the evidence that inflation has ended under interest rate targets, and the amazing period of the zero bound, which turns on its head the experience of 1940s and 1950s interest rate pegs that so influenced him.

I'm still not convinced for other reasons.  First, we have lost sight of the issue. None of this bears on the question whether the operating procedures should have a vertical or horizontal daily supply curve. If the Fed implemented its interest rate target with a flat daily money supply curve, but raised the level of that target with the price level or inflation, we get the same implied upward sloping supply curve in response to inflation. It does not matter if the little black lines in my picture are horizontal or vertical. They play no role in the "M" control that controls "PY" -- the shifting of the target when P goes up does everything.

Indeed the arguments at the time were nothing like this. There was a lot of vague talk about how fixed quantities would force the banks to make adjustments that prices would not. Once again with the Fed, talk went way ahead of modeling. Demand curves are curves, and prices are the same thing as quantities.

[Warning: MathJax equations in the next paragraph that don't show on all devices]

Second, I think MV(i)=PY has disappeared from monetary economics for a good reason, so the whole latter graph is no longer relevant. When V(i) responds to interest rates, as it obviously does, then MV(i)=PY does not determine P. The classic example: Take logs, so \(m=log(M)\), etc., and linearize to \(m_t + v_t = p_t + y_t\). To accommodate interest on reserves, write V as a function of the interest costs, \(v=\alpha(i_t-i^m_t)\) Add the relation between interest rates and inflation, \(i_t = p_{t+1} - p_t\), and we have
\[m_t - \alpha(p_{t+1}-p_t - i^m_t) = p_t + y_t.\] Fixed \(m_t=m\) determines a difference equation for prices, but there is no unique solution.

(An interesting problem: Money targets and passive interest rate targets each leave indeterminacy. Can the combination determine the price level? The above equations say no, as the Fed targets the interest on reserves, and can't peg \( i\) and \(m\) simultaneously. But maybe there are better equations. At any rate, I don't know of a current model that does this.)

Monetarists always talked about how velocity is interest elastic in the short run it was "stable" in the long run. But it's not. Money demand -- reserve demand especially -- has exploded by a factor of 300 at \(i-i^m\), and it's not ever coming back as long as that is the case.

More importantly, just what is M? There were always many different aggregates around, and all pay interest. Our financial system today is awash in extremely liquid assets that all pay within a few basis points of the same interest rate. The tiny $10 billion of reserves are just a tiny tax on deposits, not the last hair of a dog which, if you held it still, would cause the dog to wag.

Most generally, we already live the Friedman rule, and economy of extremely liquid interest-paying assets, which you can regard as electronic money, or barter of bonds rendering money irrelevant. The only question is whether the Fed (or treasury) will provide that money, or whether it will be provided privately through repo and similar devices, which occasionally blow up.

Moreover, any corridor system implies artificially rationing liquidity -- artificially making reserves scarce. What is the point of that? You don't slow a car down by running it deliberately low on oil. In Friedman's day, it was thought you had to -- the gas pedal was stuck, so that full market interest on reserves would lead to galloping inflation. The huge lesson of the last decade here and in Europe, and quarter century in Japan, is that this doctrine is false.* (Footote leads to a digression on cash management.)

In sum, the screaming lesson of the last 10 years in the US, and 25 in Japan, is that a "liquidity trap" with arbitrary reserves does not cause any inflation. Money control has vanished from monetary economics for about 40 years. And a daily vertical supply curve combined with 24 hour supply readjustments isn't monetary control anyway.  It seems time to eliminate the memory of money from the operating procedures. I think the Fed agrees, which is why it wants to operate a floor system -- yet a minimal floor system.

There really is a pretty clear test here: Do you believe that an interest rate target is sufficient to control the price level, or do whatever else the Fed wants to do with monetary policy? If so, as pretty much all monetary economics of the last 40 years states, then why bother with anything else. If not, well, let's talk about why not, just what are we trying to accomplish, and get some papers written on what is wrong with a pure interest rate target.

I sense this is where the Fed is as well, which is why they favor the minimal-reserves floor.  But that still doesn't answer, what is the minimum level of reserves, why minimal reserves, and why control the size of the balance sheet at all? We'll need to find other questions to which this might be the answer.

Also, where is the minimum? Before 2008, reserves were tiny, $10 billion, compared to today's $1,500 billion. Could the supply curve move that far to the left?

The return of volatility, and the upper bound 

At our little conference, Darrell Duffie pointed out a problem with the minimal floor and the corridor: The Hamilton volatility could reappear.

The demand for reserves used to be dominated by required reserves, the amounts banks must hold as a fraction of their checking account and similar deposits. That was on the order of $10 billion.

Now, however, the demand for reserves may be dominated by the regulatory demand for "high quality liquid assets." This demand could be above $1 trillion, necessitating a much larger balance sheet. For that reason, look for reserves to stay large for the foreseeable future.

Moreover this new demand it is also likely to volatile, so a corridor system with a truly fixed supply of reserves, or a floor pushed too far to the minimum could have quite large variation in overnight rates. Darrell showed some graphs suggesting this is starting to happen, with occasional rate spikes.

The regulatory liquid asset demand is fundamentally different as well. If banks have x deposits, there is little they can do but hold a fraction of x reserves. Many assets, including treasuries, can qualify as high quality liquid assets. So the ability to substitute, and the desire to do so in response to interest rate or minor market pricing differentials is larger. If we go to a floor with a minimal but actively managed system, with the New York Fed intervening every day but then fixing the quantity for the day, good luck to them keeping on top of it.

Darrel passed on an intriguing idea. (I didn't catch the source; from the NY Fed. I'll update when I find it.) First, it would help to lower the demand for reserves if the HQLA regulation treated treasuries on fully equal terms with reserves.

Second, the Fed should allow banks to borrow reserves any time they want, with treasurys as collateral, in a repurchase agreement. If a bank sells treasuries, it still takes a day (two?) to actually get the money. If the Fed stood ready to instantly transform treasurys into reserves, then treasurys would really be just as liquid as reserves. So, treasurys satisfy the regulatory demand for HQLA in place of reserves, but treasurys can be turned in to reserves at moment's notice if really needed, i.e. to pay people in a run.

It's a little bit sneaky really -- here is a "huge supply you can have anytime you need it" is not that different from "here is a huge supply you can have." We're still begging the question just why is reducing the supply of reserves so all-important. The question that seems to be answered here is that the Fed wants for some reason a fig leaf of small reserves appearances covering the reality of large reserves.  That suggests politics is behind much. of the issue here. If it is, it's a clever solution.

The flat supply curve returns

In my graphical treatment, under this arrangement the supply curve of reserves is flat, at the upper part of the range or whatever rate the Fed says banks can borrow reserves. As a result, if the demand shifts around, spikes in interest rates (and the money-market turmoil they represent) are avoided.

This is a much deeper change, however. It means the Fed really is adopting my suggestion, as you can see by the graph. It's not fixing the balance sheet at a given value, with some sort of escape hatch, as it may sound. It is a flat supply curve. That potential is also already there from the discount window, where banks can borrow reserves at a higher rate, though the Fed discourages that practice.

Let me clarify the logic of this supply curve, as it's important later. Suppose the Fed offers to lend reserves to a bank. The Fed taps a few clicks on its computer and increases the amount of reserves. It credits those to the account of the bank. Now, the Fed can do one of three things.

1) It can treat the loan itself as an asset. The bank owes it money after all. Yes, hold on to your seats, this is possible. You might object that creating money out of nowhere, giving it to a bank, and then calling the loan an "asset" is a fishy way to run a monetary system. You'd be right, and central banks know it. It's not wrong, it's just dangerous, because if the bank fails -- i.e. gives the money away and can't pay it back, then the central bank can't soak up the money later either. That's why central banks usually

2) It can require collateral for the loan, in the form of securities. The loan still counts as the asset, but the Fed is a little more protected. Better yet,

3) It can take the collateral in the form of a repurchase agreement. The Fed actually buys the collateral asset, say a treasury, in return for the new reserves. It agrees to reverse the transaction the next day or week or so, and these are typically rolled over. This is better than collateral, because in a bankruptcy you have to ask permission to grab collateral, which the bank still technically owns, while in a repurchase agreement, it's yours to keep costlessly. (Almost. Yes, I know the world is more complex than this.)

4) Or, the Fed can just buy the asset and not call it a loan after all.

Central banks almost always lend against collateral or via repurchase agreements, or buy assets.

In private accounting, a repurchase agreement is (I think) still counted as a collateralized loan, even though you do own the security.

You might object that the bank that borrowed the reserves typically quickly gets rid of them. There is no point in borrowing at a high rate to invest at a low rate. So maybe it should count as reducing reserves, as the bank owes reserves to the Fed. But then the bank buys an asset or makes a loan, and whoever is on the other side of that keeps the reserves. The bank's debt of reserves to the central bank is an asset to the central bank, not a negative liability.  The aggregate quantity of reserves rises.

I drag you through all this to emphasize how close all these operations are. They all increase the size of the balance sheet -- more assets, more reserves. And, though this is not necessarily clear in central-banker-ese, they really are functionally the same thing. Lending, repurchasing, standing ready to buy, are all pretty darn close and are functionally a flat supply curve.

While the Fed is at it, it might as well add a flat supply curve at the IOER rate, as I have shown. If banks are free to bring securities and get reserves, why not let them be free to bring reserves and get securities?

So here we are. If the Fed does not want the floor system with abundant reserves -- to the right of any conceivable demand -- and it wants to avoid rate spikes, it will likely end up with my flat supply curve anyway. The immense "size of the balance sheet" question is reduced to a very narrow and technical question -- at what point does the asset side of the balance sheet switch from assets the Fed has deliberately chosen to buy to assets that the Fed takes as repurchase agreements or in response to an offer to buy? But we're giving up on controlling the size of the balance sheet.

Lending and spreads

We have some serious questions to go. What is the spread between the interest on reserves rate and the lending rate? Who gets to borrow? On what terms?  What securities should the Fed accept as collateral or purchase?

I favor a strong Keep It Simple approach on all these questions. Knowing the Fed, I expect it all to get complicated and muddy, and full of attempts to control all sorts of ephemeral financial epicycles.

First, keep a very narrow gap between borrowing and lending rates. The point here is to control interest rates, and the narrower the gap the better controlled they are. Summing up everything, really, I would ask the Fed 1) What are you trying to achieve by controlling the size of reserves and the balance sheet? and 2) What are you trying to achieve by letting market rates float up and down inside a wide band? 

I have heard many answers to the second, but none that make sense to me. Something about market discipline and letting markets work and giving the Fed a feel for market pressures and so on.

Second, I would advocate two rates and simple rules. One borrowing rate, the interest on reserves rate, one lending rate, a repo rate, the Fed only takes short term treasuries as it only purchases short term treasuries (better yet, treasury electronic money, but that hasn't been invented yet), and the Fed will either purchase or repo. This is for normal times -- yes in panics central banks repurchase or borrow against all sorts of other things, and in QE the Fed may want to buy other things. I'd rather it didn't but let's not confuse the argument. This is about the conduct of normal everyday monetary policy, not about financial crisis fighting extremes.

The Fed already offers two rates on reserves, a lower one to money market funds and a higher one to banks. It is proposing to offer a lower one still to narrow banks. This kind of discrimination only invites political and legal challenges, and looks to outsiders much too much like subsidies to Fed-favored customers and attempts to direct the industrial organization of banking. Next, Congress will want higher IOER for community banks, or banks that fund solar projects, and so on.

If the narrow bank affair, and the incoherent discussion that preceded allowing money market funds to invest in reserves is any guide, just how the Fed opens a repurchase facility will be even more complicated and contentious. And there are some genuine concerns.

The Fed already already has a collateralized lending facility, in the discount window. It strongly discourages its use, called "stigma." The Fed would rather banks borrowed any needed reserves elsewhere, so other banks rather than the Fed are in charge of monitoring a bank's health, and needing to borrow at the Fed is a sign of trouble. Then needing to borrow at the Fed is a sign of trouble, and banks really avoid doing it. Fair enough, break glass in time of emergency.

One concern is that banks would routinely turn to the Fed to get their money rather than the tiresome business of getting private parties to lend them money. The ECB's experience is instructive here.

The Fed has a large stock of securities it owns outright, and most of the balance sheet is such. The ECB started out with few assets, and created euros against collateral for most of its balance sheet. While the Fed is doing next to no collateralized lending today (as far as I know), the ECB is actively lending against all sorts of collateral, including corporate bonds and sovereigns, down to fairly low ratings.

There are some differences between repo lending and asset purchases. You can be a lot more picky about what assets you buy, if you choose to buy them. And a bank, worried about its profits and losses, can be less picky about repo collateral than about purchases. After all, the countrparty has promised to pay you back, so the credit quality is really first and foremost that promise rather than a bet on the underlying asset. So repurchase agreements typically allow a wide range of collateral. Moreover,  the other side will generally offer the junkiest security as collateral it can.

And so it has been for the ECB in both its lending and outright purchase programs. Set up a list of what you'll take, and people give you the worst stuff on the list.

Central banks are not just there to make money, they are running monetary policy. And when you allow anything but short-term treasuries as collateral, repo becomes a means of financing. Buy a high yield risky bond, then give it as collateral to the central bank, to borrow the money that you need to buy the bond in the first place. In this way the ECB is not just providing reserves, it is financing a string of dodgy bonds, in dodgy banks, and dodgy sovereign debt. If that were to happen in the US, it would again attract political attention as well as being unwise policy. Being a central bank you have to act impartially, but then you need clear simple rules that keep people from taking advantage of you.

I think the answer is pretty clear: Keep the discount window as is, an emergency lending facility that charges a decently high rate, takes a lot of different kinds of collateral. Add the repo facility that only takes short-term treasurys, or HQLA complaint assets, and so avoids becoming a means of financing.

I expect the Fed to do something entirely different. Given the mantra of controlling the balance sheet, it will operate the repo facility with lots of rules and restrictions. Only some banks get to use it, different banks get different rates, maybe banks have to show they need HQLA, or it will add some sort of stigma. As long as we are keeping political economy in mind, all of this is dangerous to the Fed. I urge the Fed not to go in this direction.

The asset side and size of the balance sheet. 

We still haven't answered, just why does the Fed seem so hell-bent to lower the balance sheet to the minimum?

A lot of the answer, I think, comes from the asset side of the balance sheet - - which assets the Fed has bought, and provided reserves in return. In the various emergency support (2008-2009) operations the Fed first bought toxic assets and commercial paper. In the QE operations, it bought long term treasurys and mortgage backed securities. Other central banks are buying corporate bonds, stocks, and sovereigns. All of these operations are designed to push the asset prices around, and have basically nothing to do with the supply of reserves. They also push asset quantities around. Buying MBS with interest-paying reserves is funding mortgages by issuing government debt or interest paying money.  The motivation turned from crisis-management to "stimulus" in face of the zero bound.

Bringing this all to an end is a good idea. Using the banner "normal" to do so is even better. The more quickly the Fed can go back to something like a rule against buying anything but short-term treasurys the better. The Fed needs a pre-commitment, "yes, you'd like us to buy green new deal bonds, but we have a firm rule, we only buy short-term treasurys."

But I still get the sense that the Fed views the "size of the balance sheet" as a separate policy tool from the level of short-term rates, apart from the composition. Bigger balance sheet, more stimulus. That just doesn't make any sense to me. In the satiation region, short term treasurys and reserves are perfect substitutes. If I take your $20 and give you 2 $5 and 1 $10, it doesn't stimulate anything. Again, old ideas stick around, and perhaps this has something to do with MV=PY, and monetary expansion somehow a separate tool from interest rates. If so, that just doesn't make any sense at all, and raising reserves from $10 billion to $3,000 billion with zero effect on inflation, despite a chorus of opeds forecasting Zimbabwean hyperinflation, ought to do the trick. Expanding and contracting reserves in the satiation region has no effect on anything. There is no reason to want a small supply of reserves.

KISS dear Fed. Floor with abundant reserves, or better yet a very narrow band with flat supply curve. An all treasurys balance sheet.  Reduce temptations to micromanagement.

Next up: The target is up for review too. Price level target? Nominal GDP target? Symmetric or asymmetric -- do we make up for past inflation as well as past disinflation? r*? Level rule or difference rule -- should the Fed set rates according to economic conditions or raise and lower rates in response to conditions? More wild and crazy Grumpy ideas?

Last word: I occasionally am a little critical of the Fed in these writings. But the fact that the Fed is having this strategy review, that it is doing so quite openly, is really admirable.  I put up my views as usual, but these are as always malleable. I look forward to the May 3 conference and then maybe I can understand other views better.

-------

*As one little indication of the advantages of an immensely liquid banking system, Adam Copeland, Linsey Molloy, and Anya Tarascina at the NY Fed show how abundant reserves have changed payments. When reserves were scarce everyone waited for the end of the day to make payments really quickly. Now, it's all calmly stretched out through the day. Lots of effort at cash management to use as few reserves as possible is just a social waste.

In the scarce-reserves era, reserves were not actually scarce. The Fed allowed intraday overdrafts. You could write checks in the morning, knowing checks were coming in to cover them in the afternoon, i.e. borrow reserves from the Fed during the day. The resulting reserves were immense. By 4 pm the entire dance had to end, leading to a lot of late in the day scrambling. These are the modern day counterparts of Friedman shoe-leather that is saved by zero interest rates or full interest on abundant reserves.

Lots of observers, including the Fed in its TNB complaint, bemoan the death of the Federal Funds market. But that's a great thing! Why should banks scurry around borrowing money overnight? The whole financial crisis was about too much overnight lending! The Fed should cheer any death of overnight funding markets it can engineer!

-------------

Update

Scott Sumner posts some interesting comments. I agree with most of them. Scott points out that we're really not sure which way interest rates are connected to inflation.  I'm not sure either. I am sure that scientific knowledge on this point is weak.  Scott talks a lot about M2, but does not advocate a return to money targeting. Neither do I. Scott ends by arguing "Instead of targeting interest rates, we should target NGDP futures prices." I actually sort of agree. My preferred wild idea is to target CPI futures, or, equivalently, the spread between indexed and non-indexed debt. Scott says I advocate an interest rate target, which I do not. In this post, I think about how the Fed should implement an interest rate target given that the Fed has decided it wishes to follow an interest rate target. That's the question the Fed is asking right now.