Optimising retail deposit pricing

Banks are holding more retail deposits and also paying more for them – a recipe for reduced funding risk but also reduced profitability. Wei Ke, Ben Snowman, Ada Pham and Jens Baumgarten argue banks can mitigate the latter effect by analysing customer segments more closely

In September 2007, Northern Rock became one of the first banking industry victims of the credit crisis – forced to obtain liquidity support from the Bank of England after the wholesale markets it had been relying on seized up. Other banks fell into the same trap, and funding pressures have continued to be an issue for the banking industry throughout the past five years. The lessons are being learned by both banks and regulators.

The Basel III reforms, finalised in December 2010, include two new liquidity standards – the liquidity coverage ratio (LCR) and the net stable funding ratio (NSFR). The LCR requires banks to hold safe assets to cover estimated liquidity needs during a 30-day period of severe stress, while the NSFR requires banks to more closely match long-term assets with long-term funding. In both cases, regulators have provided incentives for banks to grow their retail deposits instead of wholesale funding – the proportion of stable retail deposits that is assumed to run off during the 30-day term of the LCR is set at a 5% minimum, for example, in contrast to the 75% or 100% assumed for unsecured wholesale funding.

The LCR does not come into force until 2015, and is currently being revisited by the Basel Committee on Banking Supervision, but even without this regulatory push, banks have been fighting for retail deposits far more fiercely in recent years. The problem is that retail deposits are expensive to acquire and maintain. At the same time, deposit growth is slowing, triggering a price war as banks battle for business. Banks need to somehow win a big market share of retail deposits while also turning a profit.

That’s not an easy task, but it can be done by closely analysing pricing data and segment-level elasticity – in essence, trying to work out how high a rate different customer segments demand. In fact, from our experience, banks that apply this kind of approach to their deposit book can see an increase in profitability – or reduction in self-funding costs – of 15 basis points for every pound sterling of deposit inflow, while still meeting their liquidity targets.

This kind of analysis can also be used to tackle specific strategic scenarios. For instance, a bank may have acquired market share of a fixed-term product with aggressive rates in the previous year and now faces losing the maturing portion of the book. It may want to revise the pricing structure without losing customers, after having centred the acquisition strategy on aggressive interest rates that attracted a lot of hot money. Or perhaps it is losing share in certain competitive markets and wants to catch up without being dragged into an all-out price war.

Banks need to somehow win a big market share of retail deposits while also turning a profit. That’s not an easy task

Several leading UK banks are already seeking to optimise their deposit business in this way – more banks should and will follow.

The context
Retail deposits, made by both consumers and small to medium-sized enterprises (SMEs), are the primary tool of self-funding for most banks today. The average term length for retail deposits is typically much longer than that of wholesale funds, and thus self-funding typically leads to much lower term volatility. This is why retail deposits tend to be more expensive than wholesale funds.

Initially, the higher cost of self-funding was not a big problem for the banks for two reasons: the Bank of England maintained a low base rate environment in a bid to jumpstart the economy; and consumers and SMEs’ urge to fly to safety led to a decline in overall price sensitivity, which eased the inherent high cost of self-funding.

The situation is different today. Though the economic indicators have not changed much for the past two years, the sustained nature of the economic crisis has blunted the urgency of flight to safety, slowing deposit inflows and forcing banks to fight harder for customers. In turn, that has increased price awareness among customers and heightened demand for favourable deposit rates. It is no wonder net interest margins have suffered – that’s partly because banks are paying more for deposits, but it’s also because loan-to-deposit ratios continue to fall (see figures 1 and 2).

The good news is that it is possible to optimise deposit interest margin – but it requires some detailed analysis of the behaviour of different types of customers.

Step one: Build a pricing database
The first and most important data source is customer transaction history over the preceding two to three years. That should include account openings by new or existing customers, account closing and adjustments to an existing account, withdrawals from an existing account or renewals to a new fixed term. These activities are most likely influenced by deposit pricing – that is, the interest rate paid – as well as other non-price factors such as seasonality, marketing spend, the macro environment and brand strength.

Account-level activities should be time-stamped and eventually aggregated into time series for each activity at the segment level. We call these aggregated time series ‘flow of funds’ or ‘flows’ (see figure 3).

The activity-specific flows represent different aspects of the customer demand for retail deposits, and must strictly satisfy the balance equation that defines the net change in total deposit balance between a pair of consecutive time periods for each deposit product. Data infrastructure limitations may prevent some banks from capturing a very detailed customer transaction history; in this case, we would rely on monthly snapshots of account balances to deduce activity-specific flows.

The second data source is historical interest rates over the past two to three years for the bank in question and its competitors. Preferably, weekly snapshots should be captured with full details of the pricing sheets. Third-party rate vendors can often provide good details on competitor rate sheets. Behavioural research suggests customers make purchasing decisions by comparing the price of one product with a reference price derived primarily from rates offered by competitors.

A good proxy for this reference price is what we call the weighted average reference rate (WARR) – a single reference rate calculated by taking the weighted average of all competitor rates, excluding the subject bank’s rate. The relative position of the subject bank’s rate against the WARR, calculated as the difference between the two, is the pricing lever that influences demand.

Consumer psychology gives at least three possible candidates for the weights in calculating WARR: deposit market share; equal weights or simple averages; and assigning 100% to best market rates. Each of these has its merit and plays a part in purchase psychology. Ultimately, we fit the demand elasticity models with all three possible WARR candidates and pick the winner based on model fit.

The third data source is non-price factors that would also affect the demand for retail deposits. For example, marketing or promotional activities would increase product or brand awareness, leading to jumps in demand for one or more deposit products in the portfolio. Macroeconomic events such as bank failures would often trigger a flight to safety, leading to abnormal changes in demand for retail deposits. Other non-price factors should be explored and considered on a case-by-case basis. Non-price factors should be used to avoid mis-attributing these effects.

Step two: Segment-level elasticity
Understanding the price elasticity of demand with sufficient granularity is the key to successful price differentiation. In an ideal world, we would first observe price elasticity for each individual customer, then group the customers into segments of similar elasticity characteristics, and finally offer a different interest rate to each segment through rate optimisation.

However there are two roadblocks. First, building an individual-level elasticity model would require each customer to make a statistically sufficient number of historical transactions – a usually untenable constraint even with the latest advances in academic research on individual-level models. Second, segmentation based purely on elasticity characteristics is not trivial to implement and may have adverse legal and reputational implications in the UK – put simply, if one customer is offered a better rate than another simply because she or he is more price-sensitive, it may not go down very well with other customers.

The best approach we have found is a methodology called discovered segmentation. The first step is to identify any existing product segmentation dimensions, some of which reflect inherent natures of the deposit book. Typical dimensions include definitions of footprint geographies or markets, distribution channel, deposit product type, term, balance tiers and indicators of customers’ relationship with the bank. The key here is to identify segment dimensions by which we can legally and feasibly conduct and implement price differentiation.

The second step involves creating additional granularity within each segment beyond the level in use today, for example by using additional balance tiers. It is important, however, to ensure each segment continues to have statistically sufficient data.

Third, a subject bank would need to fit elasticity models – described in the following section – to each of the granular segments obtained in the previous step. Following this, it would look to identify opportunities to merge neighbouring segments based on the likeness of elasticity to reduce the complexity of the final segmentation and make it actionable. The fifth and final step involves refitting an elasticity curve for each of the merged segments.

These elasticity models are called causal models (figure 4). Their primary function is to link price inputs – that is, the cause – to demand effects at the segment level. The typical price inputs for each segment-level elasticity model are our bank’s rate and WARR – often we would calculate the difference between the two and use this relative rate as the sole price input.

The demand effects we are trying to explain statistically are based on the flow of funds from account opening, closing, and adjustments. It is important to remember to normalise the demand effects with the non-price variables we have collected previously – for example marketing activities, macroeconomic events and seasonality – so an accurate statistical measure of price elasticity can be obtained. Elasticity models at the segment level for each flow of funds should then be assembled together in accordance with the balance equation to predict the aggregate balance level of the deposit products that are in scope for optimisation.

Step three: Optimise deposit rates.
Based on the segment-level elasticity insights described above, it would make intuitive sense to harvest or maintain inelastic segments by offering relatively unfavourable deposit interest rates compared to WARR, while protecting or growing elastic segments by offering favourable ones. This intuitive strategy would eventually allow us to maximise overall profitability, which is our main objective in interest rate optimisation.

However, it does assume that a sophisticated funds transfer pricing (FTP) mechanism is in place at the subject bank, as the profit margin we calculate for each pound sterling of deposit inflow is equal to the difference between the FTP for the corresponding term length as provided by the bank treasury and the interest rate offered to the customers. An alternative but more straightforward objective applicable to banks with less sophisticated FTP mechanisms is the minimisation of self-funding cost – that is, the average deposit interest offered per pound of deposit inflow.

As we optimise the deposit rate sheet, it is important to apply rate constraints that actively manage inter-product cannibalisation. In other words, rates across the deposit product portfolio should follow a logical order – for example, longer-term deposits should command a higher rate – while rate differences between a pair of products considered as close substitutes should be regulated.

In addition to rate constraints inherent to the nature of the deposit book, there may be other business rules to keep in mind. For example, the overarching pricing strategy for the bank may dictate that we maintain a certain price position relative to WARR or the best rate in the market. We may decide to promote a particular product or introduce a new one, while retiring an old one – additional rate constraints should be added to reflect these strategic directives.

Last but not least, liquidity constraints should be applied. After all, the most basic goal of deposit management is to obtain the deposits as cheaply as possible while still reaching the liquidity target. The simplest liquidity target is the total balance of the deposit book, though we can certainly get more sophisticated or creative by assigning targets for specific promotional campaigns or specific deposit products in the portfolio.

The intuitive logic of price differentiation by elasticity characteristics is unfortunately difficult to execute manually, particularly because testing all feasible rate combinations in a brute force way is computationally inefficient. The good news is that several off-the-shelf optimisation engines can be configured to solve this problem, including the Solver add-in for Microsoft Excel in some cases.

We do caution, however, that optimisation is not all created equal. Generic optimisation engines are typically designed to solve a particular type of problem and cannot be used for other problem types. Home-grown optimisation engines offered by some deposit-pricing solution providers are based on the application of certain pre-determined pricing rules. Thus, they do not actually offer truly optimal results.

Wei Ke is a director at Simon-Kucher & Partners in New York; Ben Snowman is the firm’s London-based head of UK banking; Ada Pham is a London-based consultant with the firm; Jens Baumgarten is managing partner and head of the financial services practice in the UK and North America