Modeling Deposit Maturity with Bond Replication Portfolios

Summary

The document discusses how to improve a bank’s estimate of the maturity of non-maturing chequing deposits for funds transfer pricing. The starting method assigns active accounts the historical average time until full withdrawal, split by customer demographics and balance bands. That estimate may not capture how deposit balances behave as a pool over time.

It proposes separating the account-volume time series into a stable core and a less stable component, then matching the core’s liquidity outflows with cash flows from a bond portfolio. The duration of that portfolio serves as an approximation of the deposits’ maturity. The answer offers this as one approach and points to literature for a broader overview, but gives no implementation details, validation results, or treatment of changing rates, customer behavior, or model risk. The replication approach therefore needs assumptions and testing suited to the bank’s deposit base and pricing purpose.

Key ideas

• Historical average account life can be segmented by customer demographics and balance bands.
• A deposit volume series can be decomposed into core and less stable components.
• A bond portfolio can be constructed to match the core deposits’ liquidity outflows.
• The replication portfolio’s duration can approximate deposit maturity.

Tags

# How to model the maturity term of non maturing deposit accounts


# How to model the maturity term of non maturing deposit accounts












My client (bank) currently follows a naive method to model the maturity term of chequing accounts. We need to model the maturity to correctly calculate the FTP pricing of these chequing accounts.

The method is :we look historically what is the average life of the checking account before the customer withdrew all the cash and take that average as the maturity of an active chequing account. The averages are per custoemr demographic, and also account balance bins.

How can I improve on this basic model?

## Answer by Cettt (score 3)

https://quant.stackexchange.com/a/46868

I am afraid there is no short answer to that question. However there is some literature you can check. In this paper the author gives an overview over different methods and lists a lot of references.

One approach is to decompose the volume timeseriies of your checking accounts into two parts:





Then you try to replicate the core part of this time series with a bond portfolio: the idea is that the cashflows of the bonds should match the liquidity outflow of your accounts (such models are often called replication models).

The maturity term of the deposits can then be approximated by the duration of the bond portfolio.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.