Bank of Canada Staff Working Paper 2021-65: 'Predicting the Demand for Central Bank Digital Currency: A Structural Analysis with Survey Data'

'Predicting the Demand for Central Bank Digital Currency: A Structural Analysis with Survey Data'

Jiaqi Li

With 14% of central banks at the pilot stage of exploring a possible future Central Bank Digital Currency (CBDC), this paper seeks to address three related questions;

  • What would be the demand for a CBDC?

  • How would different design attributes affect demand?

  • To what extent would a CBDC impact the demand for cash and deposits?

The attributes identified which changed the demand for a CBDC were budgeting usefulness, anonymity, bundling of bank services and the rate of return. These attributes were compared with a baseline CBDC and the variance ranged from 4-52% depending on the extent to which a CBDC was seen as cash-or deposit-like.

Model approach

The approach taken differs from previous studies which have used estimated demand parameters to predict the adoption and use of CBDCs as a new payment instrument. It looks at potential holdings of CBDCs for use both as a store of value and for transactions. The approach also regarded CBDC as an imperfect substitute for deposits where CBDC can differ from deposits in a variety of attributes.

With virtually no real-life data to use, rather than use surveys of the public, which rely on their understanding of CBDCs, the author used existing Canadian data about household cash and deposit holding proportions of their liquid assets and their view of the product attributes of cash and deposits. This data allows an estimate of households’ preferences towards each product attribute.

The starting point was that households obtain a utility from holding different products (cash, deposits or a CBDC) based on their attributes, the household characteristics and the products fixed effect that captures the average impact of unobserved households’ idiosyncratic preferences.

The Canadian survey dataset was used to predict the demand for CBDC with a given CBDC design. It was applied to each chosen design attribute as well as the product attributes of cash and deposits. Alongside these assumptions were made about a range of CBDC specific effects, the impacts of household characteristics reflecting the different demographic groups value of CBDC compared with cash and deposits and the utility of holding CBDC reflecting the average impact of households’ idiosyncratic preferences ranging from fully cash- to fully deposit-like.

Impact of product attributes

The study looked at the impact of each product attribute using the percentage change in CBDC demand in response to a change in the given attribute relative to the baseline design, with all other attributes remaining unchanged. Unlike the predicted level of CBDC demand, the percentage change in CBDC demand in response to a change in a given attribute relied much less on CBDC-specific effects.

The baseline CBDC design was a CBDC which was non-interest bearing and unbundled with bank services. It was found that the baseline design replicated 70% of cash’s usefulness for budgeting and its anonymity. While 4% of median households hold cash as a liquid asset, 4% – 52% of household liquid assets would be held as a CBDC. The lower bound was a totally cash-lie CBDC and the upper bound was a deposit-like CBDC. Unsurprisingly households with higher incomes, older age groups and home owners hold more cash and were, therefore, more likely to hold a CBDC if cash-like.

  • Budgeting usefulness. Reducing the budgeting usefulness of CBDC from 70% of cash usefulness to deposit usefulness lowered CBDC demand by around 7% - 14%, depending on different assumptions for CBDC specific effects.

  • Anonymity. Reducing CBDC anonymity from 70% of cash anonymity to 0% like deposits lowered CBDC demand by around 5% - 10%. If CBDC became bundled with bank services, its demand increased by around 4% - 8%.

  • Rate of return. Increasing the CBDC rate from 0% to 0.1% could raise its demand by around 10% - 23%.

The estimated model also allows for studying the impacts of changes in CBDC designs where a combination of different attributes change at the same time. For example, moving from a cryptocurrency design with a full degree of cash anonymity to the Sand Dollar design with a low level of anonymity could reduce CBDC demand by around 6% - 10%.

Impact of CBDC on cash and deposits

In this model demand for CBDC was funded by households’ liquid assets. An increase in demand for CBDC reduced cash and deposit demand by more than the increase in demand. The CBDC specific effects played a large role in this.

If the CBDC effect is deposit-like, a CBDC with a baseline design can reduce demand for cash and deposits by about 52% on average across households.

If the CBDC effect is cash-like, a CBDC with a baseline design can reduce demand for cash and deposits by about 4% on average across households.