Photo: pamspix – Getty Images

Abstract

Banknotes can be used to make legitimate payments, but they can also be hoarded, lost or used to facilitate
transactions in the shadow economy. Understanding how banknotes are used can assist policymakers in
responding to changes in payment behaviour and demand for cash. This article examines the value of banknotes
used for each component of cash demand and how it has changed since the COVID-19
pandemic. The share of banknotes used for transactional purposes is estimated to have fallen by
5 percentage points since early 2020, while cash use in the shadow economy has increased slightly and
the proportion of banknotes that are lost has remained unchanged. Overall, the majority of banknotes on issue
are currently used for non-transactional purposes, consistent with pre-pandemic trends.

Introduction

Demand for banknotes grew substantially over the COVID-19 pandemic period in
many advanced economies. In Australia, the value of banknotes in circulation increased by
22 per cent, or $19 billion, between March 2020 and its peak in December 2022. This
followed a period of already-strong banknote demand; in the decade prior to the pandemic, banknote demand
was growing faster than GDP (Flannigan and Parsons 2018; Flannigan and Staib 2017). Banknote demand has
since declined but remains close to its historical high. This strength, relative to growth in prices and
the economy, is despite the ongoing decline in the use of cash for day-to-day transactions over many
years (Mulqueeney and Livermore 2023). The dichotomy of strong banknote demand alongside falling
transactional use suggests banknotes are being hoarded, likely for store-of wealth or precautionary
savings purposes.

Changes in the denominational mix of banknotes that are on issue can provide insight into the diverging
trends of lower transactional versus higher hoarding demand for banknotes. Growth in low-denomination
banknotes ($5, $10 and $20) has been slow, increasing at around 1 per cent annualised on a
per capita basis since 2007 (Graph 1). Low denominations are typically used for in-person
transactions and for merchants to provide change, so subdued demand for these banknotes indicates a
reduced use of cash for consumer spending. By contrast, demand for high-denomination banknotes
($50 and $100) has been strong and the key driver of growth in overall banknote demand; high
denominations have grown by almost 5 per cent on an annualised per capita basis since 2007.
This is consistent with an increased desire in the community to hold banknotes as a precaution or
store-of-wealth, especially during times of economic uncertainty (Guttmann et al 2021).

Graph 1

A line graph showing the number of banknotes per person for each denomination from 2008 to 2023. The $5, $10 and $20 banknote lines have very little growth, while the $50 and $100 banknotes have steadily risen.

This article quantifies the sources of demand for Australian banknotes to further understand the
differences in transactional and hoarding demand, particularly since the pandemic. In doing so, we update
estimates from Finlay, Staib and Wakefield (2018)
to June 2023. Understanding the relative importance of each source
of demand can assist the Reserve Bank in forming expectations about likely future developments in
banknote demand, which is a key part of determining the Bank’s annual banknote print orders. It can
also assist in policymaking regarding cash access and distribution and contributes to our understanding
of the amount of cash used in Australia’s shadow economy.

At a single point of time, banknotes in circulation will fall into one of the following categories:

  • Banknotes that are used to facilitate legitimate day-to-day transactions in
    Australia.
  • Banknotes that have been lost or destroyed.
  • Banknotes that are used in the shadow economy (either to conceal legal transactions
    to avoid tax, to pay for illegal goods and services or to store wealth generated by the sale of
    illegal goods and services).
  • Banknotes that are hoarded – that is, held, either domestically or overseas,
    as a store-of-value, for emergency liquidity or for other such purposes.

While individual banknotes on issue move between these categories every day, the share of banknotes
attributed to each category is likely to be relatively stable over the short run. However, it is
important to acknowledge that cash is an anonymous bearer instrument and so difficult to trace and
analyse. To overcome this, where possible, we use a variety of techniques to estimate the share of
banknotes attributed to each category, thereby producing a range of estimates rather than a single
point estimate. Each estimation technique naturally relies on a range of assumptions, some of which
are more realistic than others. While we cannot be definitive or exact about how banknotes are used
in the economy, our confidence in broad trends is strengthened if we see similar results emerging
across the various techniques.

Banknotes used for legitimate transactions

Australians typically use cash to complete everyday transactions, such as at the grocery store. This
source of banknote demand – which we call ‘transactional demand’ – is difficult to
estimate as the Bank does not track banknotes once issued into circulation. So, while we cannot directly
observe or measure transactional demand, we use four different methods to approximate its size to obtain
a range of estimates, based on Finlay, Staib and Wakefield (2018):

  1. the counting method
  2. the velocity method
  3. the banknote processing method
  4. the seasonality method.

Each of these methods is discussed in turn, before a final assessment on the proportion of banknotes used
for legitimate transactions is made based on all these approaches.

Counting method

The first approach estimates the stock of cash across six physical locations that are commonly used to
exchange or store cash for transactional purposes (Table 1). We aggregate these locations to form an
economy-wide estimate of how much cash is used for legitimate transactions. While this approach is a
useful and tangible method to estimate this stock, it relies on several assumptions and does not account
for any cash not captured in these categories.

Table 1: Physical Locations of the Transactional Stock of Cash
Location Description
Wallets Cash held by consumers on their person.
Financial institution holdings Cash held by financial institutions in bank branches, ATMs or cash depots.
Tills and self-checkouts Cash held in cash registers, safes and self-serve checkouts at the start of the day. This
is the minimum stock of banknotes that is held in cash registers at all times. It does
not include cash held due to an increase in stocks from consumers’ cash expenditure.
Unbanked business takings Cash held by businesses that has not been banked.
Gaming machines Cash held in gaming machines (e.g. poker machines) and associated safes.
Tourists Cash held by tourists in Australia or about to enter Australia. This includes cash
sourced overseas prior to entering Australia, cash sourced domestically after entering
Australia and cash held by overseas foreign exchange businesses that service tourists
about to enter Australia.

We estimate the stock of cash held in each of these locations by:

  1. estimating the number of a given location (e.g. the number of tills and self-checkouts) and
    multiplying it by an average amount held per location (e.g. the amount of cash held in each till or
    self-checkout)
  2. converting flow data to a stock by making assumptions about the velocity of cash (the number of times
    a banknote is used in a given period) through a particular location.

The counting method suggests that the transactional stock of cash fell from around $15 billion in
March 2020 to just below $14 billion at the end of June 2023 (Graph 2). Over the first
18 months of the pandemic, the transactional stock fell to $12 billion; while it has picked up
slightly over the past two years, it remains around $1 billion below its pre-pandemic peak. The
estimated fall was mainly due to large declines in the cash holdings of financial institutions and
unbanked business takings. This was partly offset by modest increases in wallet holdings, tills and
self-checkouts, gaming machines and tourist cash holdings.

Graph 2

A stacked area graph showing the contributions of six locations of transactional cash demand. The six components are: wallet holdings, tills and self-checkouts, gaming machines, financial institution holdings, unbanked business takings and tourists. Financial institution holdings are the largest component, while wallets and unbanked business takings also contribute large amounts. The graph shows transactional stock estimates from 2003 to 2023, which has risen from around $10 billion in 2003 to just below $14 billion in 2023. There was a peak immediately prior to the pandemic at around $15 billion.

The decline in transactional demand has occurred alongside a significant increase in the value of
banknotes on issue. As a share of banknotes on issue, the transactional stock has fallen by around
4 percentage points since the pandemic to around 13 per cent.

Velocity method

A single banknote can be used for many transactions, so another way to approximate the stock of cash used
for transactions is to estimate the flow of cash payments and convert this flow into a stock. The two
concepts are defined in the following equation:

Value of transactional stock
=Value of cash payments
÷Velocity of transactional stock

The domestic flow of cash payments is estimated by multiplying the value of total card payments from the
Bank’s Retail Payment Statistics by the cash-to-card payment ratio recorded in the Bank’s
Consumer Payments Survey (CPS) (Graph 3, top and middle panels). Cash payments made with cash
sourced overseas is approximated by subtracting the value of card payments and ATM withdrawals made with
an international card in Australia from total tourist spending estimates from the Australian Bureau of
Statistics (ABS) and Tourism Research Australia. Through this method, we estimate that cash facilitated
around $8 billion worth of transactions in June 2023 – a decline of around
70 per cent since its peak in December 2008 (Graph 3, bottom panel). Cash has been used
much less frequently since the pandemic; our estimates suggest that monthly cash payments have fallen by
around $1 billion.

Graph 3

A three-panel line graph from 2002 to 2023. The first panel displays the cash-to-card ratios interpolated from the RBA's Consumer Payments Survey. The line starts at around 1.2 in 2002 but steadily declines to around 0.11. The second panel shows card payment data from the RBA's Retail Payment Statistics from 2002 to 2023. It has risen over time from $9 billion a month to $80 billion a month. The third panel shows estimates of cash payments by combining data from the first and second panels. This line steadily falls from $11 billion a month to $7 billion a month.

We then estimate the velocity of transactional cash, which is the number of times the transactional stock
is used to make a payment in a month. This is approximated by mapping out cash movements through the cash
cycle. Banknotes start at a cash depot and are transported to an ATM or bank branch, before eventually
ending up in a consumer’s wallet or purse. Next, consumers spend these banknotes at a business
before they are returned to a cash depot or bank branch and begin the cycle again. We calculate the
average number of days it takes for cash to pass through a point in the cash cycle. For some legs of this
journey we have accurate data, such as the number of banknotes entering and leaving cash depots each day,
and so can calculate the average time a banknote spends in a depot. For other aspects we need to use
judgement.

We estimate that the velocity of cash has steadily declined since 2008 and fell sharply following the
onset of the pandemic; lockdowns across Australia and the limited ability to spend cash were key drivers
of this fall. Our estimates suggest that it takes almost six weeks for the entire transactional stock of
cash to turn over; that is, the transactional velocity of cash is around 0.7 when measured on a
turnover-per-month basis (Graph 4, top panel). Combining these two components suggests that the
transactional stock of cash was in the range of $9–12 billion at the end of June 2023, or
9–12 per cent of all banknotes on issue (Graph 4,
bottom panel). This method suggests that since March 2020, the transactional stock has fallen by around
$1.5 billion, or 4 percentage points, which is consistent with results from the counting method
above.

Graph 4

A two-panel line graph from 2008 to 2023. The first panel shows the estimated transactional velocity of cash, which falls from 1.1 in 2008 to 0.7 in 2023. There is a noticeable drop in the velocity of cash in 2020. A grey region surrounds the central estimate, which reflects the range of velocity estimates from different scenarios. The second panel is the estimated transactional share of total banknotes in circulation. The line shows a steep decline from 45 per cent in 2008 to 20 per cent in 2013. It stabilises between 2013 and 2019 before falling to around 11 per cent in 2023. A grey region surrounds the central estimate, which reflects the range of the transactional shares from different scenarios.

Banknote processing method

Our third approach quantifies the transactional stock of cash by estimating the rate at which banknotes
pass from retailers and banks to cash depots. Since depots only process banknotes that are actively
circulating and do not handle banknotes that are hoarded or lost, it can be used to approximate the
transactional stock of cash.

The processing frequency of all denominations has declined over the past decade, reflecting a shift in
consumer preferences away from using cash as a means of payment (Graph 5). The decline was
particularly sharp over 2020 and has remained relatively steady thereafter. The processing frequency of
$5 and $10 banknotes is low as retailers tend to keep these denominations as change, such that
they cycle through cash depots less frequently. The $50 denomination processing frequency has
converged to that of the small denominations, which may suggest an increasing use for hoarding relative
to transactional use.

Graph 5

A line graph from 2013 to 2023 displaying the average number of times each denomination was processed over the previous 12 months. All denominations except for the $100 banknote show a decline over time, although the $100 denomination has remained consistently lower than other denominations. The $20 banknote has the highest processing frequency at just above two times a year in June 2023. The $5, $10 and $50 banknotes have similar processing frequencies in June 2023, at around once a year, although the $50 banknote was historically much higher than the $5 or $10 banknote.

We can then estimate the transactional stock of cash by making two assumptions. First, we assume the
non-transactional stock of cash consists only of $50 and $100 banknotes, and so the lower
denominations are only used for transactional purposes. Second, the processing frequency of the
transactional stock of $50 and $100 banknotes is equal to the processing frequency of the
$20 denomination. Finally, this approach does not account for cash demand from the shadow economy,
so we subtract estimates of the stock of cash used for that purpose…