This post describes the nitty gritty of what happens when the Fed purchases Treasuries. I will go into detail on the balance sheet implications for each participant, which will vary depending on whether the market participant is a bank or a non-bank. The bank/non-bank distinction matters because non-banks do not have Fed accounts and thus cannot hold reserves.
The Fed only does QE trades through Primary Dealers, who generally are not banks (they are broker-dealers) and do not have Fed accounts. (The exception is few U.S. branches of foreign banks who house their broker-dealer business in the bank entity, which do have reserve accounts). In practice, Primary Dealers tend to bank with custodian banks like Bank of New York Mellon, who specialize in collateral management services.
But the focus should not be the primary dealers as they are merely conduits. The newly created reserves ultimately end up in the account of whoever sold the Primary Dealer the Treasuries. If the seller is a Bank, then it will end up in the Bank’s Fed account. If the seller is a non-bank, it will end up in the Fed account of the bank that the Non-Bank banks with. The bank’s new reserve asset will be balanced against new bank deposit liabilities owed to the Non-Bank.
Below I walk through four scenarios of QE sales: Non-Bank Investor to Non-Bank Primary Dealer, Bank Investors to Bank Primary Dealer, Bank Investor to Non-Bank Primary Dealer, and Non-Bank Investor to Bank Primary Dealer. This should offer insight into the plumbing of QE.
Bank reserves can never leave the balance sheet of the Fed, but that does not limit how they can be spent. Reserves are a form of money and can be spent on anything. However, banks transact with other banks in a different way than how banks transacts with non-banks. This is due to our two-tiered monetary system, where not everyone is eligible to hold reserves. For the cryptocurrency fans: this is the equivalent of a bitcoin holder able to pay another bitcoin address, but unable to send bitcoin to someone with only a Ethereum address. In this note I sketch out a few illustrations that should be helpful in understanding how this works in a bank-to-bank and bank-to-non-bank scenario.
Note: Although banks can spend reserves on anything, bank are heavily regulated in what they can buy. They cannot go out and load up on equities and high yield or other risky assets without violating risk and regulatory limits. Regulation and profitability, not reserves, is what constrains a bank’s balance sheet.
Bank to Bank Transactions
In this example a bank will purchase an asset – be it a Treasury, office building, car etc. from another bank. The transaction is essentially an asset swap, where Bank A swaps $100 in reserves for an $100 asset that Bank B holds. The aggregate balance sheet of the banking system does not change.
Bank to Non-Bank Transaction
In this example Bank A will purchase an asset from a non-bank, who banks with Bank B. Note that this results in the creation of bank deposits. The Non-Bank essentially converts his asset into bank deposits through the sale. Bank deposits are created when a bank creates a loan asset or buys something from a non-bank. Whereas the aggregate level of reserves in the banking system is unchanged, the aggregate balance sheet size of the banking sector is $100 larger. The spending of reserves by any individual bank does not change the aggregate level of reserves in the banking sector, but it does shift the distribution.
In practice, banks must hold a certain level of reserves to meet regulatory liquidity thresholds. For large banks, this is usually to meet the Liquidity Coverage Ratio. The LCR mandates banks to hold a level of high quality liquid assets like reserves, Treasuries, or reverse repo backed by Treasuries in proportion to their expected 30 day outflows. Banks that spend their reserve levels will thus only buy other HQLA assets like Treasuries or Treasury reverse repo. For example, in late 2018 JPM decided to invest a large chunk of its reserves into reverse repo. This was a time when repo rates (proxied by the Secured Overnight Funding Rate) rose comfortably above the Interest on Reserves paid by the Fed on reserves.
The level of commercial bank reserves is determined by the size of the Fed’s balance sheet, and the proportion of reserves that end up in the Fed accounts of banks. When the Fed purchases securities or makes loans, it creates reserves out of thin air to fund them. A $100 purchase of Treasury securities results in the creation of $100 in reserves. A $100 FX Swap loan also creates $100 in reserves. Reserves can only be created or destroyed by the Fed, but banks are not the only entities eligible to hold reserves.
Reserves can never leave the Fed’s balance sheet, but they can be shifted around the Fed’s balance sheet. Think of it as Bitcoin ledger, where Bitcoins are paid to other wallets but always remain on the ledger. Only entities with an account at the Fed can hold reserves, so the created reserves are shuffled amongst the different Fed account holders as payments are made. For example, when commercial bank A makes a payment to commercial bank B, then reserves are wired from bank A’s Fed account to bank B’s Fed account. The total level of reserves stays the same. Most reserves are held by depository institutions such as commercial banks or credit unions, but there are other notable entities that have Fed accounts. These other entities, such as the Treasury, can at times have large holdings of reserves. The level of reserves held by the banking sector decreases when reserves move into these other Fed accounts.
We have a two tiered monetary system, where one type of money is used when transacting with the Fed and between commercial banks (reserves), and another type of money is use when transacting with everyone else (bank deposits). This note explains the two types of money, and how they interact with each other.
Reserves are an unsecured liability of the Fed that can only be held by entities with an account at the Fed. Think of it as a checking account at the Fed, except that deposits in the account can only be used to pay entities who also have a checking account at the Fed. Broadly speaking, only depository institutions like commercial banks or credit unions are eligible to have accounts at the Fed. But there are also other notable entities such as the U.S. Treasury, GSEs like Fannie Mae, and clearing houses like the CME. When these entities make payments to each other, they pay in reserves.
Since reserves can only be sent to entities who also have a Fed account, the total level of reserves in the financial system cannot be changed by account holders. Reserves can never leave the Fed’s balance sheet and are simply shifted from one Fed account to another on the Fed’s balance sheet. It is a closed system. The total level of reserves is determined by Fed actions, which create or destroy reserves. Reserves are created when the Fed expands its balance sheet by buying assets, and extinguished when those assets are repaid. One exception to this is that reserves can be converted to currency at the request of commercial banks. If a commercial bank needs $1 million in currency, it calls the Fed, who then sends an armored truck carrying $1 million in currency to the commercial bank. The Fed then deducts $1 million in reserves from the commercial bank’s Fed account.
In recent months M2 has exploded higher by almost 3 trillion, generating enormous market chatter. This note briefly describes the mechanics of how Fed actions has led to a spike in bank deposits, which in turn has led to a large increase in M2. Note that M2 is largely comprised of different types of bank deposits, including demand deposits, savings deposits and time deposits. I’ll first go over the basic principles of central bank and commercial bank money creation, then apply the principles to recent events.