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:

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 Ozimek, Ernie 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. 

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.