Professor Sunderam gave a talk at Booth about Dynamic Competition for Sleepy Deposits at the banking workshop.
It reminded me of how local small banks in China get people to open up account and do deposits. Local small banks’ major customers are grandmas — so banks would give rice, oil and/or eggs as gifts for opening up account or putting deposit.
Then, grandmas would periodically “move” their deposits from bank to bank to get free gifts. They would even compare all banks and communicate to get the best deals.
Somewhat isomorphic to PhD students searching for seminar food. Anyway…
Professor Sunderam and the author team collaborated with Fiserv, a software solution provider for a lot of small-scale banks and credit unions. They got a total of deposit data from 58 million accounts at 920 banks/credit unions.
Model model of dynamic competition between banks for deposits:
- At each time period, fraction $\phi$ of depositors “asleep” each year, rest rolling over automatically.
- Bank set deposit rate.
- Awake depositors deposit at new banks with logit demand, contingent on deposit rate and bank’s idiosyncratic quality.
Banks play Markov Perfect dynamic game; symmetric equilibria only; $\beta=0.9$ assumed. The paper solves a linearized policy function of banks that sets price, where coefficients are estimated (Table 4).
The main message seems to be: among small banks on the Fiserv platform, deposit stickiness is large, and if you take the dynamic strategic model seriously, it accounts for a big share of estimated franchise value.
Two additional thoughts:
A quick Google tells: JPMorgan serves more than 85 million retail (consumer) customers, while 68 million for Bank of America. Big banks differ from community banks vastly: consumers are younger, more urban, and more digitally engaged. Big bank’s “quality” is valued differently than small community banks, and their cost-revenue structure makes deposit interest rate enter their utility function too. The differences create a non-negligible gap in terms of structural parameters estimated and what really matters for the deposit market.
Moreover, banks aren’t actually Markov perfect strategist. Likely, community bank rate-setting is benchmarking + rules of thumb + monthly ALCO meetings. And the “as if” defense (competitive pressure produces optimization-like behavior) is not really working because FDIC insurance dampens competitive pressure, and sleepiness itself actually might protect bad pricers — local market power means survival doesn’t require optimal pricing.