Introduction

Cash is demanded for a number of reasons: as a medium of exchange for payments, to hold as a precautionary reserve, and to facilitate illegal activities. Using a standard indicator of cash use, the ratio of all coin and currency in circulation (CIC) to GDP, Ashworth and Goodhart (2020) show that the demand for cash has fallen from its highest level of use in the late 1940s, flattened out during the 1980s, but more recently reversed course and rose steadily after the early 2000s for a selected group of countries: the Euro Area, Japan, the UK and US. When the increase in CIC is analyzed for the UK alone, the rise in household consumption, lower interest rates, and several indicators of tax evasion are all statistically significant. As household consumption includes both the use of coin and low value currency along with cards and perhaps some checks in the UK, the issue of cash/card substitution for payments was not addressed. Knowledge of how and why cards have substituted for cash and other payment instruments provides information on the competitive environment that retail CBDC may face, if issued by central banks. It also suggests what attributes or incentives retail CBDC may provide for users and merchants for them to demand and use CBDC.

Additional information on how currency in circulation to GDP has varied across a broader selection of advanced and emerging market countries is provided by Bech, et al. (2018). They find that this ratio rose for 20 out of the 25 countries they covered. The ratio fell in 5 countries: China, Norway, Russia, South Africa, and Sweden. With the exception of Sweden, these countries were not covered by Ashworth and Goodhart (2020). Both studies showed that the value of cash and cards rose absolutely over earlier and current time periods. Bech, et al., additionally showed that the number of ATM and EFTPOS terminals also expanded, providing greater access to both payment instruments. Based on finding that some 75 percent of the countries covered experienced an increase in their CIC/GDP ratio, they concluded that the overall substitution of cards for cash has been limited. Had the substitution of cards and other payment substitutes for cash been strong enough, more countries would have experienced a falling CIC/GDP ratio.

However, focusing on the use of cash for payments in competition with cards and e-money (stored value money), a different picture emerges. This is because a rise in the ratio of currency in circulation to GDP is really an indicator that the stock of cash rose relative to the value of all types of payments used to produce GDP, such as business to business payments for intermediate inputs, employee payroll payments, international transactions, and purchases of output by the various categories of final demand. The vast majority of these payments use instruments that do not compete with cash, especially where it is normally used (at the point of sale). Although wire transfers and ACH-type payments rarely compete with cash, from the 1950s to the 1970s, checks and giro payments in many countries substituted for cash in payroll, bill payments, and at the point of sale. More recently, debit and credit card payments at the point of sale have eventually become the main cash substitute for purchases, although checks and e-money in some countries may play a minor role.

Assessing how the use of cash is changing in the market for cash requires a different measure. The one we use—a market share measure of cash use—is simply the value of cash withdrawn from ATMs (CASH) as the share of cash in the market for cash or CASH/(CASH + CARDS + E-MONEY). This is a value share measure of the flow of cash to the flow of cash and its substitutes, rather than a stock measure of CIC to the flow of GDP. This measure is not only more focused on the use of cash where it is normally used—which retail CBDC would also be designed to do—but shows which cash substitutes have had the greatest effect on the use of cash for payments over time. One study has used this indicator as a proxy (Taylor, 2006).

Our interest in the cash substitution issue is three-fold. First, we want a direct indicator of the use of cash compared to its substitutes. Such a measure should reflect how competitive cash substitutes are in the current market for cash in a country. Second, since cash use is falling, we want to know what seems to be driving this trend across countries. Third, while retail CBDC may have some effect on the future trend in cash use, it is more likely to have a greater effect on the share of existing cash substitutes (primarily cards) in our cash use measure. This depends on the demand for CBDC and the strength of the incentives that may exist to promote its acceptance and use, if issued.

To this end, in Section II we borrow from competition analysis and adopt the market share measure of cash use defined above. The effect of including over the counter (OTC) cash withdrawals at financial institutions, missing for all but 6 countries of the 25 we cover, is noted in the text and can affect the market share of cash if OTC cash acquisition is very large. Unfortunately, these data include something we don’t want and can’t be used. Our cash share measure is contrasted with two other measures of cash use: the standard measure of the ratio of CIC to GDP and the ratio of small value CIC to household consumption (HC), the latter being more representative of what cash is used to purchase. Cash use is falling for 8 of our countries using the standard measure, 13 countries using the second measure, and 24 countries using our cash share measure. Countries that have had the most substitution away from cash by 2019, and currently have the lowest share of cash use, generally are not those that have had the greatest absolute increase in this substitution over 2012–2019. Korea, Sweden, and the U.S. have the lowest cash shares in 2019 and experienced a relatively small reduction in their cash share over 2012–2019. China, Spain, and Argentina have higher cash shares and experienced a greater reduction in their shares over the same period. Thus, the downward trend in cash use is not linear but for many countries is falling at a decreasing rate. This suggests that fewer additional individuals are choosing cash substitutes once most have already done so.

In Section III, we attempt to explain why cash shares have fallen across countries. Instead of trying to determine the importance of the many separate influences that can affect the choice of cash versus its substitutes across countries (such as convenience, relative cost, availability of infrastructure, trust in electronic payments, etc.), we use a summary measure reflecting the marginal revealed preference for cash substitutes by younger adults compared to the marginal preference for cash by older adults observed in a set of payment diary studies. This concerns the changing age structure of individuals in a country’s population. The end result of using either approach should be the same since the separate influences on payment choice generate the revealed preference we see in payment diary studies. Thus, we use revealed preference, not the multiple influences on payment choice that lead to revealed preference.

In Section IV, we note that some countries already have a rather low level of cash use, indicating that considerable cash substitution has already occurred. If retail CBDC is issued, its adoption and use by users and merchants may have more effect on the share of current cash substitutes than on the future share of cash use in a country. As cards are the main cash substitute in most countries, CBDC adoption would likely be higher, all else equal, if CBDC matched or exceeded the benefits of cards. We take as given that CBDC will be structured to largely match user and merchant benefits associated with card use in terms of convenience, speed of payment, fraud control, and other desirable payment attributes. We then focus on other possible benefits associated with CBDC that may promote its adoption. These incentives are related to transaction fees or costs, such as user fees, merchant interchange fees, and merchant working capital costs. A summary is in Section V and a data appendix notes the sources of our data.

Cash Use: Different Measures for Different Purposes

To gauge the use of cash for all purposes together—small value purchases, precautionary reserve, and illegal activities—it is common to use the ratio of all coin and currency in circulation to GDP. This measure is shown in Table 1 for the countries we cover, with the exception of 6 in the Euro Area. Countries in the Euro Area know the value of Euros they have issued but not how much is used inside their borders. Consequently, only the CIC/GDP ratio for the Euro Area as a whole is reported in their place.¹

^{Table 1.}

Rising, Stable, or Falling Cash Use: All Countries, 2012–2019 (in percent)

Note: Table 1 is for the purpose of making cross-country analysis. For some jurisdictions, the best indicators of cash usage are from country-level payment diary studies. See Bagnall et al., 2016; Chen et al., 2021. Source: Author’s calculation.

^{Table 1.}

Rising, Stable, or Falling Cash Use: All Countries, 2012–2019 (in percent)

	Jurisdiction	All CIC to GDP	Small CIC to HC	Cash to Cash plus Card & E-money
1	Argentina	-40	-82	-30
2	Australia	8	0	-40
3	Belgium	na	na	-43
4	Brazil	-1	-29	-45
5	Canada	14	0	-2
6	China	-23	-30	-31
	Euro Area	16	15	na
7	France	na	na	-19
8	Germany	na	na	-19
	Hong Kong SAR	25	3	na
9	India	1	16	-28
10	Indonesia	-2	-21	-6
11	Italy	na	na	-13
12	Japan	15	6	-50
13	Korea	74	-26	-35
14	Mexico	33	-6	-14
15	Netherlands	na	na	-40
16	Russia	-14	-51	-58
17	Saudi Arabia	33	-5	-13
18	Singapore	31	1	-20
19	South Africa	-5	-25	-11
20	Spain	na	na	-28
21	Sweden	-51	-30	-53
22	Switzerland	21	12	-2
23	Turkey	-7	-35	0
24	United Kingdom	-5	-12	-40
25	United States	17	-6	-28
	Legend:	Falling	Rising or Stable

Note: Table 1 is for the purpose of making cross-country analysis. For some jurisdictions, the best indicators of cash usage are from country-level payment diary studies. See Bagnall et al., 2016; Chen et al., 2021. Source: Author’s calculation.

View Table

^{Table 1.}

Rising, Stable, or Falling Cash Use: All Countries, 2012–2019 (in percent)

	Jurisdiction	All CIC to GDP	Small CIC to HC	Cash to Cash plus Card & E-money
1	Argentina	-40	-82	-30
2	Australia	8	0	-40
3	Belgium	na	na	-43
4	Brazil	-1	-29	-45
5	Canada	14	0	-2
6	China	-23	-30	-31
	Euro Area	16	15	na
7	France	na	na	-19
8	Germany	na	na	-19
	Hong Kong SAR	25	3	na
9	India	1	16	-28
10	Indonesia	-2	-21	-6
11	Italy	na	na	-13
12	Japan	15	6	-50
13	Korea	74	-26	-35
14	Mexico	33	-6	-14
15	Netherlands	na	na	-40
16	Russia	-14	-51	-58
17	Saudi Arabia	33	-5	-13
18	Singapore	31	1	-20
19	South Africa	-5	-25	-11
20	Spain	na	na	-28
21	Sweden	-51	-30	-53
22	Switzerland	21	12	-2
23	Turkey	-7	-35	0
24	United Kingdom	-5	-12	-40
25	United States	17	-6	-28
	Legend:	Falling	Rising or Stable

Note: Table 1 is for the purpose of making cross-country analysis. For some jurisdictions, the best indicators of cash usage are from country-level payment diary studies. See Bagnall et al., 2016; Chen et al., 2021. Source: Author’s calculation.

The values shown for each country in Column 1 indicate the percent change in the CIC/GDP ratio over 2012–2019. The 8 countries where CIC/GDP is falling are shaded. Countries where this ratio rises or is stable are unshaded. For example, the ratio for Argentina fell 40 percent, from a small base level of 0.090 in 2012 to 0.054 in 2019.² In contrast, the Euro Area’s ratio rose by 16 percent, Japan’s rose by 15 percent, while India’s and Brazil’s were stable (as the change was less than or equal to plus or minus 1 percent). To the 5 countries Bech, et al. (2018) found that decreased their overall use of cash (China, Norway, Russia, South Africa, and Sweden), we add Argentina, Indonesia, Turkey, and the UK. Explaining why these country’s ratios fell is beyond the scope of the paper, although this is attempted for our cash share measure in Section III.

Another way to measure cash use, shown in detail in Amromin and Chakravorti (2009), is to recognize that most cash is used for lower value purchases. This focuses on coin and lower value currency denominations (small CIC). This separates cash used mostly for purchases from cash held as a precautionary reserve, held overseas, or used to facilitate illegal activities. In the 13 OECD countries they covered, the exchange rate adjusted average dollar value (excluding Switzerland) was $37. Small CIC was defined as currency denominations lower than the most supplied denomination in a country’s ATM network. For the US, the most supplied denomination from ATMs was $20, so the value of $10, $5, and $1 notes plus the value of all coins defined small CIC. We used $35 and under to define small CIC for all countries. Multiplying $35 by the average purchasing power exchange rate (PPP) over 2012–2019 for each country to obtain the average equivalent local currency value, the closest local currency denomination was then the cutoff for small CIC.³

Instead of relating small CIC to GDP, it makes more sense to relate it to that component of GDP where small CIC is most used—household consumption (HC). This gives the ratio small CIC/HC which fell for 13 countries in Column 2. As expected, each of the 8 countries with a falling CIC/GDP ratio, also had a falling small CIC/HC ratio. But there were four reversals: from a rising CIC/GDP to a falling small CIC/HC, one of which applied to the US. And, as shown in Column 3 for our cash share measure, all rising CIC/GDP ratios and rising small CIC/HC measures were reversed to falling shares of cash use using our measure.⁴ The greater the focus on where cash is used—in GDP, household consumption, or in the market for cash, the more countries are found to be reducing their relative use of cash.

More detail is available for some countries. A White Paper from the People’s Bank of China (2021) reported a payment survey indicating that the value of cash payments in 2019 was only 16 percent of the total, lower than the value of cards at 23 percent and far lower than mobile payments at 59 percent. Some 46 percent of survey respondents reported using no cash at all during the survey period. This occurred at a time when the value of cash in circulation (M0) rose from RMB 6.83 trillion by year-end 2016 to RMB 8.43 trillion by year-end 2020, an average annual increase of close to 6 percent. This is an example of where a narrow measure of cash use falls even as a broad measure rises.

A. Cash Share Changes Across Countries

As noted above, our cash share measure combines the value of ATM cash withdrawals, debit and credit card payments, and the value of e-money payments that substitute for cash.⁵ It reflects the share of cash in the market for cash: CASH/(CASH + CARDS + E-MONEY).⁶ A clearer picture of how cash use has changed over 2012–2019 is seen in the three parts of Figure 1.

Cash Share by Country over 2012-2019 and Projection to 2022

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Cash Share by Country over 2012-2019 and Projection to 2022

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

Cash Share by Country over 2012-2019 and Projection to 2022

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

The Y-axis of each of the three parts of Figure 1 reflects the cash share level in 2012 while the level in 2019 is reflected above this year on the X-axis. To provide some idea what the cash share might be in 2022 (sans Covid), the trend over the first eight years was extended for an additional three years. The three-year projection was based on estimating an inverse logistic curve for each country using observed data over 2012–2019. The two estimated parameters generated an estimate of cash share values for the extra three years, which were combined with the observed sample. A cubic spline was applied to the total augmented sample for each country which generated all the curves shown in Figure 1.⁷ The projection is only for illustration and using a ruler would have given much the same result. The main takeaway is that the trends in the majority of cash share reductions are the result of something relatively steady and smooth over time, rather than an influence that has significant up/down variation in a country.

In Figure 1, three countries—Brazil, India, and Russia—had by far the largest absolute reduction in cash shares. Brazil and India both experienced a large one-time drop in their cash shares. In Brazil, the central bank also created its own Fast Payment system (called Pix) to increase competition. There was also an economic crisis that reduced GDP by 3.5 percent in 2015. Most card use in Brazil is buy-now-pay-later, and card expenditures rose by 9 percent while cash use, for those most affected by the crisis, may well have fallen. In that year, the cash share fell by 22 pp, a much larger reduction than the 1.7 pp average annual cash share reduction over 2012–2019 for all countries. In 2016, India demonetized some 86 percent of their currency, slowly issuing new notes in their place. The cash share fell by 17 pp over three years. The surprise demonetization was to drive out “black” money accumulated through corruption and tax evasion (Chodorow-Reich, et al., 2020). Russia was different. To reduce the vulnerability of Russian banks and their domestic payment system to Western economic sanctions, the central bank established and promotes their own card system (Mir). It charges a lower interchange fee and offers users monetary incentives to expand its card use and replace established card systems (The Economist, 2021).

Only Russia is in the group of three that also experienced the greatest percentage reduction in their cash share in Column 3 of Table 1: Russia, -58 percent; Sweden, -53 percent, Japan, -50 percent. Both Sweden and Japan have already experienced a substantial reduction in their cash share and by 2019 had a relatively low cash share base (“inflating” their -50 percent reductions). Other than Brazil, India, and Russia, most other countries experienced a rather mild annual average absolute reduction in cash use over the period. In particular, Canada, France, Indonesia, Italy, Korea, South Africa, Switzerland, Turkey, and the US experienced less than a 1 pp average annual reduction in cash use.⁸

Our cash share measure is similar to a single firm market share measure in competition analysis. We believe it to be close to reflecting a one-to-one substitution of cards and e-money for cash where cash transactions are most used. However, it is not perfect. Some consumers may withdraw cash over the counter (OTC) at their bank rather than use an ATM and businesses may provide “cash-back” at the point of sale. Few countries collect this information and thus it is not included in our analysis. However, some limited information is available for 6 European countries (ECB, 2020). Based on the survey questions being asked, the available OTC data appear to be the sum of two things: OTC cash withdrawals by businesses at their financial institutions to make change plus OTC withdrawals by a bank’s consumer depositors rather than using an ATM. Unfortunately, only the latter represents information we would want to use. Business use of cash to make change does not show up in the value of a transaction and thus is not comparable to the value of card transactions in our cash share measure.

If we neglect the fact that the available OTC cash data are larger than what we want (it includes business cash withdrawals OTC for making change), including it in the numerator and denominator of the cash share measure for 6 European countries results in a smaller reduction in the cash share measures than we report in Table 1 and Figure 1. For Germany and Spain, our reported reduction would be too low by, respectively, 7 and 6 pp— two countries where banking offices are quite numerous relative to the population. For Italy and France, the reduction would be too low by 2 and 1 pp, while for Belgium and Netherlands it would be too low by less than 1 pp. In sum, the reported reduction only importantly affects 2 of the 6 countries.⁹ One final data issue is that in some countries, a portion of check, giro, or credit transfer payments may actually be replacing what could otherwise be a cash transaction. However, only total values of these payment instruments are available—not the value of each instrument that may, in some countries, be classified as a possible cash substitute. Only very well-informed persons in a country would have the experience needed to make this determination as well as have the contacts needed to collect information on OTC cash withdrawals by consumers alone and cash-back at firms.

In Table 2, eight countries have a cash share ranging from 6 percent to 20 percent in 2019. They are Korea, Sweden, US, Japan, Australia, UK, France, and Netherlands. Their average reduction in cash share level is 1.2 pp a year. At this rate, it may take 5 to 17 years for the cash share measure to be close to zero.¹⁰ The more important point is that for these countries, cash substitutes have a market share of from 80 percent to 94 percent. Going forward, the share of cash substitutes will likely only rise, even as habit formation of individuals that continue to use cash may be stronger. Next, we investigate why the cash shares in Figure 1 have fallen.

^{Table 2.}

Countries Ranked by Cash Share Level in 2019, Showing Annual Percentage Point and Percent Reductions Over 2012–2019

Source: Author’s calculations

^{Table 2.}

Countries Ranked by Cash Share Level in 2019, Showing Annual Percentage Point and Percent Reductions Over 2012–2019

Jurisdiction	Cash Share Level 2019 (%)	Annual Absolute Reduction (pp)	Annual Percent Reduction (%)
Korea	6	-0.5	-5.0
Sweden	8	-1.3	-7.5
United States	10	-0.5	-4.0
Japan	11	-1.6	-7.2
Australia	16	-1.5	-5.7
United Kingdom	16	-1.6	-5.8
France	19	-0.6	-2.7
Netherlands	20	-1.9	-5.7
Belgium	25	-2.7	-6.1
Russia	29	-5.8	-8.3
Switzerland	31	-0.1	-0.3
China	32	-2.0	-4.4
Singapore	36	-1.3	-2.8
Spain	36	-2.0	-3.9
Italy	41	-0.9	-1.9
Brazil	44	-5.2	-6.5
South Africa	45	-0.8	-1.6
Argentina	47	-2.9	-4.3
Turkey	49	0.0	0.0
Canada	49	-0.2	-0.3
Germany	52	-1.8	-2.8
Mexico	57	-1.4	-2.0
India	64	-3.6	-4.0
Saudi Arabia	72	-1.6	-1.9
Indonesia	77	-0.7	-0.9
Euro Area	na	na	na

Source: Author’s calculations

View Table

^{Table 2.}

Countries Ranked by Cash Share Level in 2019, Showing Annual Percentage Point and Percent Reductions Over 2012–2019

Jurisdiction	Cash Share Level 2019 (%)	Annual Absolute Reduction (pp)	Annual Percent Reduction (%)
Korea	6	-0.5	-5.0
Sweden	8	-1.3	-7.5
United States	10	-0.5	-4.0
Japan	11	-1.6	-7.2
Australia	16	-1.5	-5.7
United Kingdom	16	-1.6	-5.8
France	19	-0.6	-2.7
Netherlands	20	-1.9	-5.7
Belgium	25	-2.7	-6.1
Russia	29	-5.8	-8.3
Switzerland	31	-0.1	-0.3
China	32	-2.0	-4.4
Singapore	36	-1.3	-2.8
Spain	36	-2.0	-3.9
Italy	41	-0.9	-1.9
Brazil	44	-5.2	-6.5
South Africa	45	-0.8	-1.6
Argentina	47	-2.9	-4.3
Turkey	49	0.0	0.0
Canada	49	-0.2	-0.3
Germany	52	-1.8	-2.8
Mexico	57	-1.4	-2.0
India	64	-3.6	-4.0
Saudi Arabia	72	-1.6	-1.9
Indonesia	77	-0.7	-0.9
Euro Area	na	na	na

Source: Author’s calculations

Source of the Demand for Cash Versus Its Substitutes

Following earlier money demand studies, we attempt to identify the major sources of the demand for cash relative to the value of cash substitutes across countries. Specifically, what is driving the reduction in cash shares and why is the rate of change for all but a few countries relatively slow? An answer here will tell us why the cash shares in Figure 1 have fallen and how these shares may continue to change in the future if retail CBDC is not issued.

Money demand studies usually find that higher interest rates tend to reduce the demand for cash (e.g., Ashworth and Goodhart, 2020). Countries with relatively high interest rates may experience a lower demand for cash, lowering its share of cash relative to cards and e-money. These studies also include an indicator of economic activity in a country. While greater economic growth logically leads to more payments being made, by itself it is unclear why this would favor cash substitutes over cash—unless there already is a preference for one over the other and greater economic activity just reveals this a priori preference. Here economic expansion is not the driving force: it would be the a priori preference.

Such a preference would be a function of relative user prices, rewards for use, convenience in making payments, acceptance by retailers, availability of ATM and POS terminals, marketing programs, and so on—all the things that can be difficult to measure well in a regression equation. Consequently, this is not the approach taken. Instead, we specify a composite indicator of user preference for cash substitutes versus cash that is inferred from payment diary studies presented for the US and seven other countries. Payment diary studies suggest a weaker preference for cash by younger adults than for older age groups in a population, due to all those things that are difficult to measure in a single equation and to habit formation as adults age.

Two studies shows a lower share of cash transactions by younger adult age groups (Yoshizawa et al., 2021; Kim, et al., 2020). Another shows that the value share of cash used by younger adults is lower than that for older adults across six additional countries (Bagnall, et. al., 2016). A third study shows fewer cash transactions and a lower value of cash spent per transaction for younger adults (Swiss National Bank, 2020).¹¹ Transaction shares or value shares by age are not levels of use. The important measure for our purpose is the actual number of cash transactions or the total value of cash spent by age group. This is the denominator of the transaction or value shares but is not reported in published payment diary studies.¹²

A more detailed analysis of three US payment diary studies over 2017–2019 illustrates the difference in cash use by age more clearly. Looking only at two age groups—younger adults aged 18 to 44 and older adults aged 45 and older, the total number of cash transactions for the younger group is 3,119 compared to 6,684 for the older group. The total value spent using cash was $94,541 for the younger group but $153,988 for the older group (63 percent higher). In part, this reflects the fact that the older group represents 41 percent of the US population while the younger group represents 33 percent.¹³ This suggests that one reason for the general fall in cash use observed for individual countries in Figure 1 is associated with inter-generational differences in the revealed preference for cash relative to its substitutes.¹⁴

Cash Share Values and Change in Population Age Structure by Country

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Note: The numbers in Figure 2 represent the alphabetical order of the numbered countries in Table 1.Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Cash Share Values and Change in Population Age Structure by Country

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Note: The numbers in Figure 2 represent the alphabetical order of the numbered countries in Table 1.Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

Cash Share Values and Change in Population Age Structure by Country

Citation: IMF Working Papers 2022, 027; 10.5089/9798400200861.001.A001

Note: The numbers in Figure 2 represent the alphabetical order of the numbered countries in Table 1.Sources: BIS and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

Due to a lack of detailed data on the preference for cash across multiple age groups for our 25 countries, we form a simple indicator of this preference: the sum of a country’s birth rate plus its death rate, which approximates the expected annual change in the intergenerational structure of the population and thus its preference for cash or a cash substitute. The birth rate of new younger adults born in 1995 and turning 18 in 2012 approximates the addition of new, younger adults in 2012 with a general marginal revealed preference for cash substitutes. The current death rate in 2012 approximates those for older adults aged 44 and above leaving the population who have a higher preference for cash. Those in between our indicators of annual entry and exit move up the line and contribute to the change in the revealed preference for cash substitutes over cash in a population.

As individuals in each age category are changing each year, the (earlier) birth rate + the (current) death rate in year 1 (BRDR1) is the change in the age structure in year 1. By year 2, the cumulative change in age structure is the sum of the change from year 1 (BRDR1) plus the change in year 2 (BRDR2), and so on for eight years of time-series observations across 25 countries in our pooled sample of 200. This variable (sumBRDR) represents the cumulative sum of the individual annual changes in the age structure of the population. As the cash share level falls for each country over time, it is consistent with a correspondingly higher proportion of younger adults in the population—both from adding new younger adults and having fewer older adults. Here the expected relationship would be negative for each country separately over time.

The bottom half of Figure 2 illustrates how this changing age structure (sumBRDR), and its associated preference for cash substitutes, has varied over time in the pooled sample of 25 countries over 2012–2019. The numbers in the figure represent the alphabetical order of the numbered countries in Table 1. As the population becomes increasingly less cash centric, we see that the share of cash—our cash share measure in the top of Figure 2—typically falls, in some countries at a faster rate than others. The large reductions in the cash shares of Brazil (number 4), India (9), and Russia (16) shown in Figure 1 are seen here as well. The time-series component of our pooled countries shows a generally negative relationship separately for each country, as expected.¹⁵ More information is obtained from the following money demand equation using pooled data displayed in Figure 2:

where:

• CASHSHARE = the cash share values for each of 25 countries observed over 2012–2019 (N = 200) shown in Figure 1 and the top of Figure 2;

• sumBRDR = the cumulative sum of the birth rate plus the death rate for each country over 8 years, clearly an exogeneous influence on the dependent variable;¹⁶

• HC = the local currency PPP adjusted US dollar value of household consumption in each country to reflect economic activity; and

• iRATE = the monetary policy related percent interest rate, some of which are zero or negative during this period (so logging variables in the estimating equation to obtain elasticities is not feasible).

Equation (1) is estimated using a Fixed Effects model where a dummy variable is used to control for crosscountry “intercept” shifts in the panel data so that time-series effects across countries may be better locally identified. In Table 3, the Within results show that the intergenerational change in population age structure (sumBRDR) is negatively and significantly related to the variation in each country’s cash share and explains a good deal of the variation of that measure. This just reflects the country-by-country times-series relationship across countries illustrated in Figure 2. The effect of time-series changes in household consumption (HC) and the interest rate (iRATE) are both statistically insignificant.¹⁷ The Between result is also shown in Table 3 and the sumBRDR variable experiences statistically significant shifts between countries.¹⁸

^{Table 3.}

Pooled Regression Results, 2012–2019

Source: Author’s calculations

^{Table 3.}

Pooled Regression Results, 2012–2019

Variable:	Fixed Effects (Within)		OLS on Means (Between)
	Parameter	t-Statistic	Parameter	t-Statistic
Intercept			-.225	-1.38
Sum of Birth Rate + Death Rate	-.773	-11.94	6.28	3.92
Household Consumption	.305E-14	.34	-.155E-13	-1.26
Interest Rate	.259E-03	.33	.234E-02	.41
Adjusted R2	.97		.46

Source: Author’s calculations

View Table

^{Table 3.}

Pooled Regression Results, 2012–2019

Variable:	Fixed Effects (Within)		OLS on Means (Between)
	Parameter	t-Statistic	Parameter	t-Statistic
Intercept			-.225	-1.38
Sum of Birth Rate + Death Rate	-.773	-11.94	6.28	3.92
Household Consumption	.305E-14	.34	-.155E-13	-1.26
Interest Rate	.259E-03	.33	.234E-02	.41
Adjusted R2	.97		.46

Source: Author’s calculations

The interpretation is as follows: cash shares are falling (Figure 1; Table 1, Column 3) at a slow pace (Table 2, Column 2) and countries with the lowest cash shares generally have the smallest reductions (Table 2, Columns 1 and 2). The cumulative sum of intergenerational change in the age structure of a population (a higher value of sumBRDR) is negatively associated with a lower cash share (Figure 2 and Table 3). As noted in the Introduction discussing Section III, we do not try to determine the importance of the many separate influences that can affect the choice of cash versus its substitutes across countries. That involves ranking subjective answers to surveys of what payment instruments people use and why (c.f., Koulayev, et al, 2016; Stavins, 2002 and 2017, and others). Instead we use a summary measure reflected in payment diary studies indicating a marginal revealed preference of younger adults for cash substitutes compared to the marginal preference for cash by older adults. The end result should be the same since the unspecified separate influences on payment choice generate the revealed preference seen in payment diary studies.

Cost Incentives to Adopt Retail CBDC

From the above analysis, it appears that further reductions in cash use across countries are likely to be slow and relatively predictable. This could change with the introduction of CBDC.¹⁹ While CBDC is expected to have attributes that are quite similar to cash as well as cards, for most countries (20 out of 25, Table 2, Column 1) the share of cards dominates that for cash and thus is likely to be more affected by CBDC. In addition, countries where cash use is already very low should find that the introduction of CBDC would likely also have a greater effect on cards than cash. We see this already in Table 2 and it was noted in the Introduction. Countries with a very low cash share in 2019 also had a relatively small percentage point reduction in their cash share over 2012–2019. Many countries with a higher cash share had a relatively larger reduction in their shares over the same period. That is, the substitution process away from cash generally slows down as the cash share itself becomes lower and lower.²⁰

CBDC adoption would likely grow relative to cash or cards going forward if CBDC improved on the attributes of these two established instruments, and/or offered other appealing attributes. We focus on the former. Being similar to a fast debit card with P2P functionality, may not be enough. In this regard, central banks have identified many reasons for and benefits from possibly issuing retail CBDC. This would include providing countervailing market power to current card schemes, possibly lowering payment costs, making payments faster, expanding financial inclusion, providing a way to make payments when other systems may be disrupted, and limiting systemic risk to financial markets and the economy from non-bank privately-issued money. These considerations and others drive decisions on why and how central banks may decide to supply retail CBDC.

This addresses the supply side of the retail CBDC issue and are nicely outlined in Bank for International Settlements (2021). The list is comprehensive and would largely duplicate the benefits from card use except for three transaction cost/fee considerations that can importantly affect the benefits obtained from CBDC compared to cards. These include user fees, merchant interchange fees, and merchant working capital costs. In our view, these will be the main determining incentives for individuals and merchants to adopt and use CBDC and thus the main influence on the future course of the cash share measures in Figure 1.

Summary

We document the falling use of cash for payments across 25 countries over 2012–2019, showing its total decline over this period as well as its annual average percentage point and percent reduction. Our measure concerns the share of the value of cash used (cash withdrawn from ATMs) relative to the market for cash, consisting of the value of cash, cards, and e-money (a stored value alternative). Data limitations prevent the inclusion of possible additional cash substitutes in the analysis. Our measure is compared to a standard measure of cash use (the ratio of all currency and coin in domestic circulation to the level of GDP) and a more focused measure of small value currency and coin as a ratio to household consumption.

Using our indicator of cash use, all but one of 25 countries covered (Turkey) experienced a decline in relative cash use, with some reductions being quite large due to special circumstances (for Brazil, India, and Russia). Reduced use of cash in a country indicates that cash substitutes are favored over cash and this may have implications for the demand for retail CBDC (if issued). Specifically, there are countries with an already low level of cash use. In eight countries, cash substitutes make up 80 to 94 percent of the payment market where cash is primarily used, and habit formation may hinder a shift to a new payment arrangement.

In trying to explain why the share of cash is generally falling across countries, a money demand equation is estimated. Instead of looking at the demand for cash or cash substitutes separately, we attempt to determine why cash as a share of cash plus cash substitutes is falling. Revealed preference for one instrument over the other, implied in payment diary studies for eight countries, and in a reworking of US payment diary information for three years, is tied to changes in the age structure of individuals in a country’s population. It could also work through differences in economic activity across countries, but this turned out to be insignificant, as did interest rates across countries. The apparent revealed preference of younger adults for cash substitutes combined with the apparent preference of older adults for cash appears to be an important determinant of cash share variation across our pooled sample of 25 countries. This is consistent with the recent generally slow and relatively predictable annual average reduction of 1.7 pp in the share of cash use across countries.

The observed slow reduction in cash use across countries may speed up somewhat if retail CBDC is offered by central banks. However, the main effect is likely to be on the share of current cash substitutes (primarily cards) as the share of these substitutes dominates the share of cash in our measure for 20 of our 25 countries. As well, for countries that already have a low level of cash use, further reductions tend to be less than where cash use is still high, indicating a slower substitution process. While central banks have largely focused on reasons for supplying retail CBDC and how this may be achieved, our interest concerns the demand side of this issue and the incentives for the adoption of CBDC. We expect that central banks will likely design their CBDC services to provide most or all the benefits currently provided by the main cash substitutes, and potentially additional benefits. In our view, some additional incentives dealing with transactions cost will importantly influence the adoption and use CBDC and largely determine of the future use of cash and card use across countries.

Treating CBDC like cash, without a user fee, would be one incentive as this would mirror the current situation for cards. The possibility for merchants to save card fees by accepting CBDC would be another incentive. And, as CBDC would be final payment and available in a receiver’s deposit account for immediate reuse through a bank, one or multiple days of working capital cost could be saved. In a more dynamic setting, card schemes may seek to mitigate these CBDC incentives by modifying their interchange fees and providing immediate access to merchant revenues from card sales and make up the revenue loss elsewhere. Alternatively, those card firms that already own and supply Fast Payment services may choose instead to rely on these newer payment services going forward. In some countries this can involve P2P mobile payments at zero cost, as well as P2B POS transactions and B2B invoice payments, both for a fee paid for by the receiver.

We take as given that CBDC will be structured to largely match user and merchant benefits associated with card use in terms of convenience, speed of payment, fraud control, and other desirable payment attributes. We then focus on other possible benefits associated with CBDC that may promote its adoption. These incentives are related to transaction fees or costs, such as user fees, merchant interchange fees, and merchant working capital costs.