Problem
A bank sells deposits that the client can top up or partly withdraw before maturity. This right works like an option: if market rates fall, a client with a fixed higher rate can bring more money, and the bank bears the interest rate risk. The rate must differ from that of a plain deposit, and the question is by how much. Black-Scholes does not fit, because most of its assumptions fail for retail deposits. A wrong spread means mispricing and extra losses when the interest rate risk is realized.
Example from the article: a one-year deposit at 13%, and after six months market rates are 8%. The rational move is to top up, but historical data shows no functional link between the top-up size and this incentive.
Solution in brief
Simulate interest rate scenarios and sample top-up and withdrawal sizes from kernel density estimates of historical customer behavior, per segment. Then search by bisection for the spread at which the expected NPV of the deposit with the option equals the expected NPV of the deposit without it.
Questions this case answers
- How to price an embedded early-withdrawal and top-up option when customers are only partly rational and the classical option model assumptions fail
- How to find the spread at which the expected value of a product with an option equals that of the plain product, using simulated rates and behaviour distributions
- How to price a deposit that the client can top up or partially withdraw early
- How to estimate the spread for deposits with top-up and withdrawal options using simulated rates and customer behavior
Tools and methods
- Extended multifactor Cox-Ingersoll-Ross (CIR) model — Monte Carlo generation of short-rate paths and yield curve scenarios
- Kernel density estimation (tophat kernel, window 0.01) — distribution of top-up and withdrawal volumes by segment
- NPV parity with and without the option — defines the fair spread
- Bisection — iterative search for the spread
Google or a one-time email code. No password.