Bear with me on this one. I'm trying to get my head straight on money and banking, by thinking weird thoughts.
Suppose, just suppose, that (for some unknown reason) the only asset that banks liked owning was bicycles. They refused to own any other sort of asset, except, of course, central bank currency and reserves at the central bank. Banks buy bikes, and rent them out for people to ride. Banks are bike intermediaries.
On the asset side of a bank's balance sheet you see the bikes it owns, plus a small amount of currency and reserves at the central bank. On the liability side you see demand deposits, just like regular banks in the real world. And the demand deposits serve as a medium of exchange. The demand deposits are redeemable on demand at par in central bank currency, just like regular banks in the real world. A bank's profits come from the fees it earns from renting out the bikes, minus any interest it pays on demand deposits, minus administrative costs, depreciation on the bikes, losses due to stolen or damaged bikes, etc.
Whenever a bank buys a new bike, it creates a deposit, and the money supply expands. Whenever a bike rider writes a cheque to the bank to pay his monthly bike rental fee, the money supply contracts.
This is a very strange monetary system. It's probably not at all optimal.
Suppose more cyclists decided that it was more convenient to rent a bike than to own their own bike. Banks would respond to the increased demand for bike rentals by buying more bikes and expanding the money supply. A fad for owning your own personalised bike would reduce the demand for bike rentals, reduce the size of the banking industry, and contract the money supply. Changes in the popularity of cycling vs driving would have similar effects on the money supply.
Changes in the supply of newly-produced bikes would affect the money supply too. An increase in the supply of new bikes would reduce the price of bikes. Depending on the elasticity of demand for bike rentals, this could either increase or decrease banks' balance sheets and the money supply.
Mass theft of rental bikes, a rash of mechanical faults in bikes, or a fall in the value of bikes, could cause banks to fail and destroy the monetary system.
If the supply of money is a by-product of the bike rental business, anything that affects the bike rental business will affect the supply of money. The reverse is true as well. Anything that affects the money business will affect the supply of bike rentals, and the demand for bikes by banks.
We don't need to assume that banks are the only firms in the bike rental business. Non banks could also buy bikes and rent them out. But the fact that banks produce money, and also really like owning bikes, would make banks really important players in the bike rental market. Being able to pay for bikes by creating deposits would tend to give them an advantage over other bike rental firms.
A central bank that aimed for monetary stability would need to keep a very close eye on the bike rental market. If there's an increased demand for bike rentals, banks would meet that demand by buying more bikes, and increasing the money supply. But this doesn't mean that people want to hold the increased stock of money that banks created when they bought more bikes. The demand for rental bikes and the demand to hold money are presumably unrelated. The seller of the bike wants to spend the money on something else. That excess supply of money would hot potato around the economy, causing an inflationary boom. Unless the central bank took offsetting action. And a fall in demand for bike rentals would reduce the money supply, as banks stopped buying bikes. And this excess demand for money would cause a deflationary recession, unless the central bank took offsetting action.
The reverse is true as well. Any actions taken by the central bank, whether or not they were designed to offset the effects on the money supply of changes in the bike rental market, would themselves affect the supply of rental bikes. If the central bank decided to increase the supply of money, perhaps to match an increase in the demand for money, this would cause a boom in the bike rental business. By increasing the supply of reserves, making it easier or cheaper for banks to buy a bike and create a deposit, the central bank is indirectly increasing the supply of rental bikes.
If the demand for bike rentals collapsed, or if no trustworthy bike renters could be found, central banks would find it hard to increase the money supply. Central banks could create more reserves, but banks would just leave them on deposit at the central bank. They wouldn't go out and buy bikes and create deposits. They can't find anyone trustworthy willing to pay to borrow those bikes.
The bike rental business would be a macroeconomic phenomenon, not a microeconomic phenomenon. And people would tend to confuse the demand for rental bikes with the demand for money. And this confusion would be perfectly understandable, because an increased demand for rental bikes would cause both an increased quantity of rental bikes supplied and an increased quantity of money supplied.
What I have just imagined is a very weird world. But the real world looks just as weird. Banks don't like having bikes on their balance sheets. Instead, they like having certain types of IOUs on their balance sheets. We can understand why banks prefer certain types of IOUs to bikes. It makes sense for the individual bank. But does it make sense for the monetary system as a whole? Bikes, the yellow metal, IOUs; these are all very weird ways to run a monetary system. What has the optimum quantity of the medium of exchange and medium of account got to do with optimum amounts of bikes, yellow metal, or those certain types of IOUs?