It's said the theoretical aim of the ECB Two-tier system (exempt a portion of the excess reserves from negative rates) was designed to:
offset the direct costs of negative interest rates on banks, thereby helping to sustain the pass-through of low policy rates to bank lending rates
push banks to trade/lend their excess liquidity in the money markets.
I get point one, but struggling with 2), can you help out?
Consider a world in which 4 banks have 3000 EUR to deposit on a daily basis with the ECB, where all of that is deposited at the going rate of -1% (minimum reserve requirements and the excess deposits)
On any given day, due to operations, banks A, B, C and D may have any allocation of that 3000 EUR to be deposited, say 100, 1000, 400, 1500 respectively. Provided everyone is above their minimum reserve requirements none of those banks has any interest in lending/borrowing money from each other, since there is no money to be made from the activity (the non-arbitrage lending rate is -1%) and this would just incur operational headache. The next day the allocations may be completely different (due to cashflows/redemptions) but the status quo remains: still no interest in interbank lending.
Now suppose a new rule. Each bank is assigned an amount (6 times their minimum reserves requirement) which they can deposit at 0% instead of -1% with ECB. For the sake of argument lets say that each banks daily assignment is 300, 400, 200, 100 respectively. Now we have a system wide optimisation problem and given the original allocation of 3000 EUR one can identify that there is slack in the system. Bank A has 200 excess capacity, whilst banks B, C and D are 600, 200 and 1400 over their assignments. An artificial market has been created, bank A should trade with either B, C and D at a rate anywhere between -1% and 0% in a size of 200 and both trading counterparties will profit at the expense of the ECB.
By providing the incentive of collective optimisation solving via market forces, the ECB has stimulated this market.
Real recent data suggests it is being solved within 1% of optimality, although how much trading it stimulates I have no experience of, albeit I would expect a reasonable amount.
