There is a plumbing explanation for the conundrum of lower nominal yields and higher inflation. Many factors affect yields, but they are in part determined by who has money and the investment constraints they face. QE mechanically increases the investible “cash” of investors who are most inclined to buy bonds, and they have been buying bonds. In our two-tiered monetary system $1 of QE creates $2 of money – $1 of reserves (money for banks) and $1 of bank deposits (money for non-banks). On the reserves side, some banks have significantly changed their behavior and begun deploying their reserves into bonds. On the bank deposit side, the wealthy ended up with the bulk of the newly created bank deposits. The wealthy tend to spend their money on assets, and they appear to be rebalancing some of the deposits into bonds. In this post we show how low rates are pressuring banks into adding bonds to their growing regulatory liquidity portfolio, how the skewed ownership distribution of new bank deposits may be leading to more bond buying, and suggest that low yields may not be a reflection of economic conditions.
Bank Portfolio Rebalancing
Banks have not historically been large buyers of Treasuries and Agency MBS, but that is changing. Recall, Basel III regulations require big banks to hold large portfolios of High Quality Liquid Assets (“HQLA”) that have historically been largely comprised of reserves. HQLA eligible assets include include Treasuries, Treasury backed repo, reserves and Agency MBS (with a slight penalty for Fannie/Freddie MBS). When repo rates rose above the interest on reserves, banks massively shifted from reserves into repo for a little extra return. The willingness of banks to substitute between reserves and repo is one of the mechanisms through which QE inevitably pushes money market rates to the policy floor. But now at least some banks are also willing to substitute reserves for longer dated HQLA securities, extending the downward pressure to longer term rates.
QE mechanically increases the size of the banking sector and thus raises the required level of HQLA, while low returns on reserves prompts a shift into higher yielding securities. Bank of America alone has increased its holdings of HQLA securities by $400b over the past year while also decreasing their reserve holdings. Those reserves may end up at on the balance sheet of another bank, forcing them to also rebalance into securities. Note that not all big banks have been employing such a strategy, with JPM noting in their recent earnings call that they are waiting for higher yields before buying securities. If they behave like BoA the purchases could be a few hundred billion.
The Rich Savings Glut
The bulk of QE created bank deposits have ended up in the bank accounts of wealthy entities. The distribution of the money matters because the less wealthy tend to spend their money on goods and services (consumption), while the wealthy tend to spend their money on assets (savings). Historically, QE created bank deposits largely end up in the accounts of the wealthy (see this post). This time around a bit made it into the general public via stimulus payments, but the wealthy still increased their bank deposit holdings by around $3t. Note that this is a conservative number because bank leverage ratio constraints have forced some depositors to move their deposits to a money market fund (see this post).
Wealthy entities tend to buy assets, and any prudent allocation will have at least some allocation into bonds. Bond fund flow data show a surge of $400b in flows year to date, far above the pace of prior years. The combination of massive fiscal spending with massive QE essentially created trillions in free money that is being spent on everything. This puts upward pressure on all prices – cars, houses, stocks, and even bonds.
Priced to Desperation
Printing and spending trillions of dollars is obviously inflationary, but linking that with higher interest rates requires assumptions that are not true. A value investor can use a dividend discount model as a framework and input forecasted earnings to arrive at a valuation for TSLA stock (very overvalued), but it is very different to take price as an input into the model and back out “earnings expectations” (TSLA will take over the world). This is because investors use different investment frameworks that can vary significantly. An investor’s framework for TSLA may even not even have forecasted earnings as an input but instead rely on estimated dealer options positioning, or price momentum. It’s very hard to know what is in the price.
Bond investors also employ a wide range of investment frameworks where sizable investors (like the Fed and some foreign reserve managers) are not even trying to make money and other investors (like banks) are optimizing under stringent regulatory constraints. This is especially true for Treasuries, which are money like dollar assets that many classes of investors must own regardless of price. To be clear, there are also smart investors carefully studying and investing according to perceived economic fundamentals. What is not clear is the extent of their influence on market prices.
Bond prices reflect all these different investment frameworks, so backing out expectations for growth and inflation from yields can be misleading. QE as a first order effect lowers yields via direct purchases, but as a second order effect increases the buying power of bond investors who are less sensitive to economic fundamentals. This is a mechanism that can keep nominal yields low regardless of what happens in the real economy.
If banks use their reserves from QE to buy treasuries, which puts yields down, how is it that when the Fed starts QE, the 10 year yields goes up, not down?
Many things determine yields. But also note that Fed usually starts QE when the world is falling apart, like during Covid or the GFC. There is an unwinding of the flight to safety bid but things calm down.
UST yields drop ahead of QE as bad things happen, banks buy bonds, and then the banks eventually sell those USTs to the Fed via QE. When the Fed is buying USTs, the UST yields are typically going up, and banks are taking profits. Did I get any of this right? Isn’t the yield on USTs more a reflection of future short rates or neutral rate +/- term premia? QE putting downward pressure on bond yields only seems to work in the sense that the Fed is guiding short rates lower for longer. Fed ends QE before raising rates. Please rip this apart.
In your June post (referenced above), you state “QE adds reserve assets (money for banks) and deposit liabilities (money for non-banks) to the banking system. This allowed bank deposits to grow even as banks were not able to extend credit. ” I understand how QE increases reserve assets. Can you help me understand “deposit liabilities (money for non-banks) to the banking system”? Stimmies adding to deposits is easy to understand, but I’m not sure that is what you are referencing. Thanks much.
Happy to help. See this post, which has an example near the end.
Thank you for directing me to your post. So the Mr Wealthy buys treasuries and with quantitative easing, the Fed is on the hunt for treasuries. So it in turn buys treasuries from Mr Wealthy (and primary dealers) but the payment is indirect– the Fed creates reserves, uses those reserves to make payment to Mr Wealthy’s bank, which in turn creates a deposit (liability) in the name of Mr Wealthy. How the very wealthy directly benefit from QE now makes sense.
Since the GFC bond yields ceased their loose tracking of nominal GDP levels albeit leading nominal GDP changes many times. Agreed that there is plenty of demand for bond yield, especially if banks don’t want or can’t generate sufficient loan growth.
Splitting hairs a bit, but you only get deposit growth if government deficit growing during times of stress, otherwise it’s just (less impactful) reserve growth.
If yields are so mechanically linked to QE, why did long-term yields decline in 2019 when the Fed was engaged in QT?
I was under impression that “money” is created through credit/lending?
You suggest (“In our two-tiered monetary system $1 of QE creates $2 of money – $1 of reserves (money for banks) and $1 of bank deposits (money for non-banks). “) that non-circulating and arguably non-productive deposits and reserves are more meaningful than lending in terms of “real money creation” in the economy? Thank you.
Hi Joseph – a great piece, appreciate your insights. Could I ask how exactly your source the UST holdings by bank from the UBPR reports? I can only seem to find ones that are text files and are quite difficult to find those related to UST or HQLA holdings, appreciate your help on this
For UBPR you can use this website, which also has a helpful graphical interface. For details on Treasury holdings (and not just Treasury and Agency together) you need another report FFIEC 31. That is also known as a “call report.” The same site also has that, available in pdf or csv.
Hi Joseph thanks for the analysis. Do you think that, the SFR (standing Repo Facility) exacerbate the “monetization of treasuries” (in particular long dated). When compared to 2013, the announcement of the end of the QE provoqued a huge rally in yield?
I think as the very very margin it might have some impact, but definitely not a lot. The SRF essentially takes out a tail end funding risk that the market never really considered to begin with (and probably forgot soon after last March).
Thanks Joseph. Great piece.
To what extent have overseas buyers impacted treasuries as well? The C/A surpluses of many large high savings countries have exploded since Covid due to the shift towards good consumption. Curious how much of an impact you think this has had relative to the other factors you mention.
The best data on foreign Treasury holdings I am aware of is Treasury’s TIC data, and that doesn’t show much change (though it is lagged by a few months). So at the moment it does not seem overseas buyers have been adding Treasuries. I agree that I would expect some of the C/A surplus to be invested in U.S. securities, though not necessarily Treasuries. Some foreign reserve managers buy Agency MBS/IG and even equities (Switzerland).
how does $1 of QE create $2 money?
Reserves (bank money) doesn’t function as money
This post and this post might help answer your questions.
Joseph, I read your book and just wanted to say it was extremely helpful. Appreciate it ! So is it fair to say that ever since QE started, bond market, at least its price, has been hugely manipulated, not a good gauge of future economic conditions like it used to be? Whether it will pan out as expected or not, I am more in camp where inflation is real and current Fed’s policy is doing more harms than goods, way too accommodative than it needs to be. If tapering starts like market currently anticipates, will bond market be normalized or FWIW, yield can rise to better capture economic conditions?
Bonds are like any other asset – some people buy based on economic conditions but many other things matter as well. For example, if the Treasury decided to issue fewer short dated debt and more longer dated debt, and total debt remained the same, then longer dated yields would rise based on supply/demand dynamics even if nothing else changed. The Fed is a big buyer indifferent to economic conditions, so on the margins I agree that market prices would better capture economic conditions. But it may still not capture economic conditions well.
Thank you for the reply !
Thanks for the insightful post as always. Wondering if you could comment on the fed funds market of late. Specifically, we’ve seen the rate drop again from 10bps post-IOER tweak to now persistently sitting at 8bps (with the percentiles dropping).
Wondering if you had a sense of how low that rate could set in the context of the above post about too much cash in the market. And then as a follow up, would the fed be incentivized to tweak the rate once again if it starts setting at 6-7bps? Would be curious to hear your thoughts on that scenario.
The lenders in the Fed funds market use GC repo as a benchmark. Since GC repo is stuck at 5bps, I think the right price for EFFR is 7bps (2bps for “credit risk”). I don’t think a 7bps EFFR would lead to a technical adjustment. But if bills trade negative I think it’s on the table.
so what happens next? will fed taper? will assets crash or at least correct?
What happens if fed double down on QE and corporate bonds get to or below treasuries yield? Will there be a move from corporate bonds to us treasuries, resulting in a stock market flash crash?
Very interesting, Joseph. Do you have a sense about the magnitude of the total QE effects (first and second order) on long treasury yields? The 10Y currently sits at just under 1.5%. Do you think the QE distortion is first order relatively to the counterfactual (no QE)? So the true 10Y might’ve been 3.5%. Or is it second order and that the 10Y might’ve been 1.75%?