Pages

Tuesday, November 27, 2018

A New Paper on the Fed's Floor System

As readers of this blog know, I have an interest in the Fed's operating system. This interest has culminated in a new paper where I look at the consequences of the Fed moving from a corridor system to a floor system in 2008. In particular, the paper looks at what this change has meant for bank portfolios and, as a result, financial intermediation provided by banks. The paper concludes with some policy recommendations. I would also note that George Selgin has just released a new book on this topic. My hope is that these projects will help inform the conversation over what operating system the Fed wants as it continues to normalize monetary policy. 

Paper Outline
So what does my paper have to say? It starts by laying out the standard arguments for a floor system:
The central idea behind this move was to remove the opportunity cost to banks of holding excess reserves by offering the banks a deposit rate at the Fed—the IOER rate—that was equal to or above short-term market interest rates. This favorable return was to sever banks’ incentive to rebalance their portfolios away from excess reserves toward other assets. The IOER rate was also to put a floor under short-term interest rates so as to align them with the Fed’s desired interest rate target. Together, these two facets of the floor system would allow the Fed to use its balance sheet as a tool of monetary policy while still maintaining interest rate control. 
In this new operating system, the stance of monetary policy was no longer set by a market interest rate but by an administrative interest rate: the IOER rate. The stance of monetary policy also was no longer tied to the supply of reserves. Instead, it was linked to the quantity of reserves demanded by banks, which the Fed influenced through changes to IOER. Specifically, the Fed set the IOER rate high enough that banks’ demand for reserves became perfectly elastic with respect to the federal funds rate. As a result, changes in the quantity of reserves supplied led to identical changes in the quantity demanded, other things being equal.  
The Federal Reserve, in short, went from an operating system in which monetary policy was transmitted through open market operations to one in which it is transmitted through the IOER rate. The Fed’s operating system changed from one in which money, in the form of reserves, mattered for monetary policy to one in which money has been “divorced” from monetary policy.
Okay, so why does this "divorce" matter? For advocates of a floor system the answer is simple:
This divorce from money is seen by many observers as the key advantage... because it gives the Fed the freedom to use its balance sheet independently of its desired interest rate target. The Fed, for example, can now sharply increase the supply of reserves in response to a liquidity crisis without causing a decline in its targeted interest rate.
Skeptics of the floor system, on the other hand, see this divorce as more problematic:
Others, however, see this divorce as creating an operating system that impairs the transmission mechanism of monetary policy... These observers’ understanding starts with the standard assumptions of a floor system. First, a floor system requires the IOER rate to be set at least equal to short-term interest rates. This removes the opportunity costs to banks of holding reserves and thereby keeps their demand perfectly elastic with respect to other short-term interest rates...  
[This] can lead to a rebalancing of bank portfolios that causes the supply of loans to be lower than it would have been otherwise. Banks lend as long as the marginal cost of funding is less than the risk-free marginal return on bank lending. In the Fed’s floor system, the IOER rate sets the marginal funding cost. Consequently, by setting the IOER rate higher than other short-term interest rates, the Fed has raised the marginal costs of funding and narrowed the gap between these costs and the risk-free marginal return on bank lending. All else being equal, the narrowing of this gap implies a relative reduction in the supply of loans and therefore a relative decline in the money supply.
Are these worries merited? My paper provides an empirical look at bank portfolios before and after the advent of the floor system to see if (1) there have been big structural shifts in bank balance sheets consistent with the critics claims and (2) whether such shifts can be attributed to the Fed's floor system. 

Empirical Evidence
On (1) I start with the following figure. It shows the share of bank assets allocated to loans and to safe assets (defined as the sum of cash, treasury, and agency asset holdings). Unsurprisingly, these series are almost mirror images of each other over both cyclical and structural time horizons. They tend to move in opposite directions during recessions--the grey bars--and over longer periods. At the advent of the floor system in late 2008, the loan share began declining as the safe asset share started rising. This change has been sustained and only recently has started reversing: 


If we break the safe assets apart into its subcategories, we see that that this tight link has been historically driven by movements in treasury and agency investments corresponding with changes in the loan share. Since 2008, the driving force behind the tight link became cash not treasury and agency investments:


If we zoom in and use two scales, this apparent structural change is even clearer.  Cash shares and loan shares become mirror images of each other:


Something big happened in 2008 that continues to the present that caused banks to allocate more of their portfolios to cash assets and less to loans. While the financial crisis surely was a part of the initial rebalancing, it is hard to attribute what appears to be 10-year structural change to the crisis alone. Instead, it seems more consistent with the critics view that the floor system itself has fundamentally changed bank portfolios allocation.

That takes us to question (2). The following figure shows the loan share of bank assets plotted against the spread between the IOER rate and the overnight dollar Libor rate. This IOER-Libor spread is negatively and strongly correlated with the loan share. This suggests banks invested less in loans when the relative return on bank reserves rose and vice versa.



Conversely, banks appear to have allocated more to cash assets with the IOER-Libor spread rose and vice versa:


The paper goes on to more carefully test these relationships using two-stage least squares regressions that control for endogeneity issues and other confounding influences. Moreover, the paper also provides a further breakdown of this relationship among foreign banks, large domestic banks, and small domestic banks. Collectively, the regressions point to a strong causal effect running from the IOER-Libor spread to the allocation of bank assets.

The Fed's move to a floor system, then, does seem to have influenced the amount of financial intermediation provided by banks to the private sector. The end of the paper provides some counterfactual exercises, including the following figure


This counterfactual shows the supply of bank loans would have been, at its peak, as much as $2 trillion higher in 2014. By mid-2018, with the IOER-Libor spread shrinking, it was closer to a $1 trillion shortfall. Now to be fair, other factors such as increased risk aversion, fintech, and new regulations also contributed to below trend growth in bank lending since the crisis. Still the evidence strongly suggest that financial intermediation to the private sector through banks declined because of the floor system. 

Policy Implications
The Fed's floor system, in short, has caused banks to increase their investment in the Fed at the expense of investing in the private sector. The question, then, becomes whether the Fed is any better than banks in allocating this credit. Put differently, is the Fed as a financial intermediary--funding short and lending long--really providing a better financial service than would have been provided by the private sector financial firms? It is not obvious to me that the answer is yes. 

What is obvious to me is that the Fed's floor system creates a whole set of other problems. First, as a profitable financial intermediary, the Fed is setting itself up for political shenanigans. Recall Congress taping into the Fed's capital surplus in 2018 and 2015. Though these transfers were relatively small and to some extent accounting gimmicks, they show how tempting a large, profitable Fed balance sheet can be to Congress. Second, the floor system effectively destroyed unsecured interbank lending. This market provided useful interbank monitoring and price discovery that no longer exists. Bringing this market-based monitoring back would be a nice addition to the bank regulator's existing toolbox. Third, the floor system forces the Fed to take safe asset collateral off the market which impairs other parts of the financial system. Only recently has President Trump's budget deficits begun to fill this hole. Fourth, the floor system can create bad optics for the Fed via the appearance of large 'subsidies' to large and foreign banks. In this era of populist politics, the Fed should be worried about the dangers to its independence this image could create. Finally, as noted above, I worry the Fed is a less effective financial intermediary than the private sector.

In short,  I see the costs exceeding any benefits from the floor system. That is why I advocate a return to a corridor system, but this time with the IOER explicitly setting the lower boundary. That is, the IOER would still be around, but it would be set lower than overnight market interest rates. For the past decade it has been for the most part above them. More details are in the paper.

Update: I have had several commentators ask me about the difficulty of returning to a corridor system given the relatively new LCR. Let me quote Governor Randy Quarles from a recent speech:
How banks respond to the Fed's reduction in reserve balances could, in theory, take a few different forms. One could envision that as the Fed reduces its securities holdings, a large share of which consists of Treasury securities, banks would easily replace any reduction in reserve balances with Treasury holdings, thereby keeping their LCRs roughly unchanged. According to this line of thought, because central bank reserve balances and Treasury securities are treated identically by the LCR, banks should be largely indifferent to holding either asset to meet the regulation. In that case, the reduction in reserves and corresponding increase in Treasury holdings might occur with relatively little adjustment in their relative rates of return.
In other words, since reserves and treasuries are viewed as equal HQLA there could be a substitution between the two as the Fed shrinks its balance sheet down. Quarles notes that currently some banks favor reserves, but that tells us nothing about what is ideal or what would transpire in a new symmetric corridor system. Finally, it is also worth noting that  banks can use Treasuries, but not reserves, as collateral for repos. Manhoman Singh is very good on that point. So Treasuries can actually make banks more liquid that non-hypothecatable reserves.

P.S. Here is a panel from the recent Cato Monetary Policy Conference on this issue. Great comments from Stephen Williamson, George Selgin, and Peter Ireland. Josh Zumbrum was the moderator.

Monday, November 26, 2018

Janet Yellen on NGDPLT

Andrew Metrick of Yale University interviewed former Fed Chair Janet Yellen today. It was an interesting discussion and one where they talked about, among other things, what changes the Fed could bring about in light of the recently announced strategies, tools, and communication review to be held in 2019.  Janet Yellen said her idea for reform "has much in common with NGDP targeting". Her response can be seen in the video below:


Glad to see her endorse a NGDPLT-like monetary regime. 

Tuesday, October 2, 2018

"Et tu, John Williams?"



Tim Duy reports that r-star, which rose to prominence over the past few years, is experiencing a Caesar-like betrayal at the Fed:
The Federal Reserve’s “r-star” has gone full supernova. New York Federal Reserve President John Williams, its key proponent, made clear in a speech late Friday that the neutral interest rate is no longer a guiding star for monetary policy. This means a federal funds rate in the range of what is considered neutral has no special significance as far as policy is concerned... 
Williams’s attachment to r-star cannot be overstated. At a professional level, it has been a key element of his research agenda. As recently as May he said that for “the moment, r-star continues to shine brightly, guiding monetary policy, but hold steady, low on the horizon.” The moment quickly passed. Last week, he tossed aside the metric, saying that it has “gotten too much attention in commentary about Fed policy.” A remarkable shift after just two 25-basis-point rate increases since his May comments... 
Williams’ speech marks the end of a transition in policy away from explicit forward guidance. It began this past August with Fed Chairman Jerome Powell’s Jackson Hole speech in which he noted the uncertainty surrounding estimates of key variables like the neutral interest rate. Fed Governor Lael Brainard pushed this point further in a subsequent speech, adding further uncertainty by differentiating between short- and long-run neutral. It continued in the September Federal Open Market Committee statement with the removal the description of policy as “accommodative.” And it ends with the primary proponent of the r-star concept — Williams — throwing it into the trash bin of crisis-era policy artifacts.
One is tempted to say "It was good knowing you r-star". However, r-star will still be around in all the models used by the FOMC and Fed staff. Just look at, for example, the policy rules on the Board of Governor's website or in its annual report. The reported change, as I see it, is more a move toward less explicit reliance on it. Implicitly, r-star will still be important to an FOMC that relies on the Phillips curve thinking in making its decisions. 

Still, these developments do indicate there is some movement towards looking at other indicators as I noted in recent post. There I suggested one useful metric the FOMC could add to its lists of monetary policy indicators is the gap between a stable benchmark growth path for nominal GDP and its actual value. I outlined in this note several ways to create this metric and note that it is in the spirit of a NGDP level target without actually adopting one. 

There are many reasons for the Fed to start following the NGDP gap. The most practical one is its ability to help the FOMC avoid falling for the inflation head fakes created by supply shocks. Here is hoping that out of the ashes of r-star's apparent demise arises an increased desire by the FOMC to pay attention to the NGDP gap.

Friday, September 21, 2018

FOMC Preview: "We Have the Nerve to Invert the Curve"

The quote in the title should be the motto for the 2018-2019 FOMC. For the FOMC is set to raise its interest rate target next week and expected to raise it several times more in 2019 despite a flattening treasury yield curve.


As seen in the above chart, the outright inversion of the treasury yield typically leads to a recession.  Despite this robust pattern, a growing number of Fed officials have become emboldened in their dismissal of it "since this time is different."  As Caroline Baum notes, 
In April, John Williams acknowledged that an inverted yield curve is “a powerful signal of recessions,” based on a significant body of research, including that by staff economists at his former bank. 
By September, Williams was already disavowing that signal. “I don’t see the flat yield curve or inverted yield curve as being the deciding factor in terms of where we should go with policy,” Williams said following a speech in Buffalo on Sept. 6. 
Next up was Fed Gov. Lael Brainard. She broke new ground in a speech last week when she... invoked the four most dangerous words in finance — “this time is different” — and applied them to the prospect of an inverted yield curve.
To be fair, President John Williams and Governor Lael Brainard are simply expressing the natural implications of the FOMC projected path for interest rates. As I noted before, the FOMC's own summary of economic projections implies a yield curve inversion over the next year or so. The FOMC, in short, is becoming increasingly dismissive of fears about inverting the yield curve. That is why the title of this post should be their motto for 2018-2019. Here is a t-shirt you can buy to commemorate this surge in FOMC boldness:


Some observers side with Fed officials arguing that "this time is indeed different" because the term premium is so low. To that I would first remind them that former Fed chair Ben Bernanke made a similar argument in 2006. As it turned out, the term premium had declined but so had the expected path of short-term interest rates. This can be seen in the figure below, which is constructed using the New York Fed's estimates of term premiums. It shows the 10-year minus 1-year treasury spread decomposed into (1) an expected rate path spread and (2) a term premium spread. These two component add up to the overall spread. The lesson here for the FOMC is to avoid falling for the siren call of the term premium excuse.


I would also remind naysayers that even if it were the case that the flattening of the yield curve is all due to a lowering of the term premium, the inverting of the yield curve still matters for financial intermediation. An inverted yield curve means smaller net interest margins for financial firms and thus less financial intermediation. That is, once the yield curve inverts, it goes from being a predictive tool to a causal agent. So even in the best-case scenario, one should not be cavalier about inverting the yield curve. Be careful FOMC.

P.S. Here is the backside of the t-shirt:



Monday, September 17, 2018

More Non-Star Metrics for Monetary Policy

In an earlier post and Bridge article, I discussed some ways to use nominal GDP (NGDP) as a cross-check on the FOMC "navigating  by stars" of r*, u*, and y*. The  motivation for these pieces was Fed Chair Jay Powell's concerns about the challenge of using these star variables when they seem increasingly in flux. Here, again, is a key excerpt from his talk:
Navigating by the stars can sound straightforward. Guiding policy by the stars in practice, however, has been quite challenging of late because our best assessments of the location of the stars have been changing significantly… the FOMC has been navigating between the shoals of overheating and premature tightening with only a hazy view of what seem to be shifting navigational guides.
The Fed chair is implicitly challenging us to find indicators that shed light on the stance of monetary policy without having to know r*, u*, and y*. The metric I suggested in the earlier pieces was the "sticky-forecast" NGDP benchmark growth path. Here I want to suggest another two measures for the Fed chair.

First, is a Phillips curve-like measure. I am not a fan of Phillips curves, but they remain stubbornly popular among policymakers and the public. They are supposed to show a relationship between the amount of slack in the economy and wage growth. However, most Phillips curves have not done well predicting wage growth (or inflation) over the past few years. Although there are many ways to measure slack in a Phillips curve, typically it is done using the unemployment rate. Attempts to salvage the Phillips curve using other slack measures have been made, like that from Adam OzimekErnie Tedeschi, and Nick Bunker who argue we should use the prime-age employment-to-population (EPOP) rate. Jason Furman, however, pushes back against this and other attempts to better define slack.

If we must stick with Phillips curve to explain wage growth as the FOMC seems intent on doing, here is my suggested approach. Take the spread between the U6 and U3 unemployment rates. This measure, unlike the prime-age EPOP, is stationary and therefore not subject to the Furman critique. Unfortunately, the U6 only goes back to 1994 so not a lot of historical analysis is possible with it. Since that time, the U6-U3 gap seems to be a fairly reliable predictor of wage growth.  The figure below plots this relationship. For purposes of comparison, it also highlights the late 1990s (blue) and the more recent years of the wage growth mystery (red).


One interpretation of this figure is that even though the U3 unemployment rate is currently low and below u*, the economy is not running as hot as it was it was in the late 1990s when the U6-U3 spread was even lower than it is now. This complements the message coming from prime-age EOP Phillips curves. This would call, all else equal, for fewer rate hikes rather than more from the Fed.

Okay, enough of Phillips curve. I want to move on to another related indicator, one that does not rely on a measure of slack but is instead tied to monetary conditions. And that is another application of my 'sticky forecast' NGDP gap measure, as explained in this note. NGDP is tied to monetary conditions for several reasons. First, it measures total money spending on the final goods and services produced in the economy. Second, it is equivalent to the velocity-adjusted money supply. NGDP, therefore, is an easy way to get a bird's-eye view of money activity in an economy. 

My sticky-forecast benchmark growth measure of NGDP can be subtracted from actual NGDP to get a NGDP gap measure. I replace the U6-U3 spread in the above figure with this NGDP gap measure to get the new scatterplot below. Once again, the late 1990s and more recent years are highlighted for purposes of comparison. And, once again, it suggests that the economy is running cooler now than in the late 1990s. Most importantly, it shows there is a strong relationship between monetary conditions and the nominal wage growth rate. Money still matters. 


Friday, September 14, 2018

The Fed's Floor System: Sayonara?

Are the days of the Fed's floor system numbered? Last month I claimed that they could be if President Trump's fiscal policy continues to spawn rapid increases in the issuance of treasury bills. His administration is relying heavily on treasury bills to finance its deficits as seen below:


This increased issuance of treasury bills matters because it implies, all else equal, a rise in treasury bill yields. Below is a figure showing the  DTCC overnight treasury-repo rate relative to the IOER rate. Lately, the treasury-repo rate has been bouncing above the interest on excess reserves (IOER) rate. This development could be a big deal.


If sustained, this rise of overnight interest rates above the IOER rate could spell the end of the Fed's floor system. 

To see why, recall that the Fed moved to a floor operating system in late 2008 by paying interest on excess reserves (IOER) at a rate higher than comparable short-term market interest rates. This pushed banks onto the perfectly elastic region of their reserve demand curve and allowed the Fed to add reserves to the banking system without causing its target interest rate to change. The supply of reserves were thus "divorced" from the setting of monetary policy (Keister et al. 2008). This divorce was seen by many a virtue of the floor system, for it allowed the Fed to manage bank liquidity and monetary policy independently, changing the size of its balance sheet to control liquidity, and the IOER rate to alter its monetary policy stance.

Maintaining this floor system, therefore, requires keeping the IOER rate at or above comparable market interest rates, so that banks will be willing to hold on to any reserves that come their way. The recent uptick in treasury bill issuance seems to be undermining this requirement and raises the possibility that the Fed could be inadvertently forced off of the floor system by fiscal policy. 

Further evidence that something may already be afoot with the rise in the overnight repo rates can be seen on the aggregate bank balance sheets from the Fed's H8 database. The figure below shows the asset category "total fed funds sold and reverse repos" as the black line. The blue line shows the same category, but just for banks transacting with other banks. This figure indicates there still is not much interbank lending, but there does appear to be a surge in nonbank lending in repos. This makes sense given the recent surge in repo rates. It may also be the first crack in the armor of the Fed's floor system as it suggest banks are beginning to invest reserves in the repo market rather than at the Fed.


Something else I showed in my piece last month was the tight fit between treasury bill issuance (normalized by the total amount of marketable treasury securities) and the IOER-Libor spread. One interpretation of this relationship is that the movement of treasury repo rates gets transmitted into interbank rates via arbitrage. 


This IOER-Libor spread, in turn, is closely tied to the percent of bank assets held in the "cash" category in the H8. (Since late 2008, this category has consisted mostly of bank reserves.) The specific pattern has been that as the IOER spread rises (falls) the cash share also rises (falls) as seen below:


Conversely, when the that as the IOER spread rises (falls) the loan share of bank assets falls (rises) as seen below:


Based on the three above charts, if the treasury-repo yield continues to be above the IOER, we should expect to see a growing share of bank assets invested in loans and fewer in excess reserves.1 That would bring the end of the Fed's floor system and rise in the inflation rate.

Some worry the liquidity coverage ratio (LCR) may complicate this story. As of 2015, the LCR started requiring banks to hold enough high-quality liquid assets to withstand 30 days of cash outflow. The LCR, consequently, has increased demand for such assets of which bank reserves and treasury securities are considered the top tier. When the IOER rate > treasury repo rates, the LCR created additional demand for reservers relative to treasuries. If, however, IOER rate <  treasury repo rates the additional reserve demand will disappear and further hasten the demise of the floor system. 

The question, then, is will the treasury bill yields continue to stay above the IOER rate? It certainly seems possible with President Trumps budget deficits. Only time will tell. 

P.S. I have a working paper on the Fed's floor system. I am beginning to wonder if it will be outdated before it gets published.

P.P.S. I see that George Selgin makes the same point near the end of this post.

1To be clear, banks cannot control total reserves, but they can control the form of the reserves. That is, banks can choose to hold excess reserves or invest in other assets like treasuries and loans that, in the aggregate, would lead to a rise in required reserves and a fall in excess reserves.

Update: One implicit assumption in my analysis is that banks are price takers, not price makers, in the overnight market. If one takes the H8 data series "fed funds sold and reverse repos" and divides it by the repo market as calculated by this New York Fed study (and updated to a more current expansion factor based on an exchange with one of the authors), one gets an average of 10 percent for the 2008:Q4-2018:Q2. Banks have been a small part of the overnight market during the Fed's floor system.

Update II: After more reflection, I am having doubts about my assumption of banks being price takers in the overnight market. I think my calculations of bank share of the repo market supports this contention, but it is hard to reconcile with the reality that overnight market rates followed movements in the IOER rate. On the other hand, if banks are able to move overnight markets then why did they fail to completely eliminate the IOER spread over comparable short-term interest for most of the past decade? 

Navigating by the Stars and Steadying the Ship Speed

I have a new article over at The Bridge where I piggyback off of Jay Powell's recent speech about the challenges of the Fed 'navigating' by the stars of  r*, u*, and y*. His concern is how to use them when they seem to be moving a lot:
Navigating by the stars can sound straightforward. Guiding policy by the stars in practice, however, has been quite challenging of late because our best assessments of the location of the stars have been changing significantly… the FOMC has been navigating between the shoals of overheating and premature tightening with only a hazy view of what seem to be shifting navigational guides.
Here is a chart I made for the piece that did not make the final cut. It shows the changing range of FOMC estimates for u* and the actual unemployment rate (u). I also added some red-lettered commentary on the chart to highlight the varying levels of FOMC concern about the gap between u and u*:


Okay, so what is the FOMC to do with all the uncertainty surrounding the navigating stars? I recommend in the article that the Fed cross check their navigational indicators by looking at nominal GDP (NGDP) relative to a benchmark growth path for it.  This growth path can roughly be thought of as the neutral level of NGDP and I show how to construct two versions of it in a recent note: a sticky forecast NGDP level and a full-employment NGDP level as seen below:


Taking percent deviations of these growth paths from the actual NGDP level gives us the following Nominal GDP gaps:


Going back to Fed Chair Powell's ship navigation analogy, looking at the NGDP gap is akin to monitoring the speed of the ship. If the NGDP gap is zero, then the ship is maintaining a steady and appropriate speed. According to the chart above, the ship is either a bit too slow per the sticky forecast measure or just about right with the full-employment measure. My preferred measure is the sticky forecast NGDP gap, so if I were Chair Powell I would currently err on the side of fewer rate hikes than more.

The main point of my note is to show how one can use NGDP as a practical way to cross check standard monetary policy indicators. I provide some other measures in the note such as a NGDP-based Taylor rule that does not rely on any of the navigational stars. I also show how a McCallum rule (which also is based on NGDP) can still be used if the monetary base is properly measured.