Pandemic payment patterns.

COVID-19 has temporarily changed the relative costs and benefits of different payment methods: cash has become more costly in terms of health risks, ease of use and likelihood of acceptance, whereas debit card usage has become less costly. As a result, consumers have shifted away from cash. Based on unique daily payment diary survey data collected between January 2018 and December 2021 amongst a representative panel of Dutch consumers, we study the shift in payment behaviour and payment preferences during two lockdown periods in the Netherlands in 2020 and 2021. Since the start of the first lockdown the likelihood of debit card usage at the expense of cash has increased by 12 percentage points compared to its trend level. About 60 percent of this shift on top of the autonomous trend persisted several months after the end of the first lockdown and part of it has persisted several months after the end of the second lockdown. The results indicate that the pandemic accelerated the increased usage of debit card at the POS, especially during the first pandemic year. Also, the pandemic has resulted in a shift in payment preferences towards more contactless payments. Both effects are largest for elderly people.

Introduction

The COVID-19 pandemic changed the daily lives of people all around the globe; it not only made our lives more contactless but also the way we pay. Electronic payment instruments at physical retail locations – point of sales (POS) – became more attractive relative to cash, because the latter involves more physical contact. Retailers promoted the usage of contactless payments at the expense of cash, and banks made it easier for consumers to pay contactless. As a result, electronic payment instruments gained further ground.

We examine the impact of the COVID-19 pandemic on consumer payment behaviour at physical POS and payment preferences using unique daily payment diary data for Dutch consumers. A key advantage of our data set is that it tracks payment behaviour at the POS and payment preferences before and during two pandemic years on a daily basis. Our data also provides a wide range of background information on respondents. The data used in this paper ranges from January 1 2018 until December 18 2021. The nature of the data allows us to examine the extent to which the outbreak of COVID-19 has led to a shift in payment behaviour and payment preferences.

As a further value added – in contrast to most other studies about the effect of COVID-19 on payment behaviour – our payment diary data includes information on cash payments, in addition to information on electronic POS payments (e.g. Bounie et al., 2020; Golec et al., 2020; Kraenzlin et al., 2020). This allows us to disentangle (1) the drop in the usage of all payment instruments due to a decrease in the overall number of transactions at the POS as a result of the lockdown, and (2) the substitution between payment methods in the short run at the start of the pandemic, but also on the long run.

Besides the richness of our payment diary data, the Netherlands provides an interesting setting to examine the impact of the COVID-19 pandemic on payment behaviour as almost all POS transactions are done either by cash or debit card. Nearly all consumers in the Netherlands have adopted both cash and the debit card which are also widely accepted by merchants. The debit card can be used with PIN and contactless. In the pre-pandemic year 2019, 32% of POS payments were in cash, 24% by debit card with PIN and 43% contactless; this sums up to 99% (DNB, 2020a). People may pay contactless with their smartphone or credit card as well. Paying contactless by debit card is done much more often than paying contactless by smartphone, as 90% of all debit cards are contactless-enabled (DNB, 2020a). This also shows that if people want to switch from one payment method to another, they can easily do so. It allows us to examine the impact of the pandemic on consumers’ payment behaviour at the POS as the results are not affected by shifts in the adoption rate.

Additionally, the Dutch government, banks and retailers took several measures to combat the pandemic. During the first lockdown in the Netherlands, that was announced by the Dutch government on March 15 2020 and started on March 16 2020, people were still allowed to leave their home and visit a store as often as they wanted, except for POS in particular sectors such as restaurants and bars, recreation and culture and the services sector. Kindergartens, schools and universities were closed and people were stimulated to work from home and to avoid public transport as much as possible.

Dutch banks took measures to simplify and encourage contactless payments in order to prevent infection through manual contact. Pre-COVID-19, consumers were required to enter their PIN code when they made a payment of more than EUR 25 by inserting their payment card into the payment terminal. If payments of EUR 25 and below reached a cumulative limit of EUR 50 the PIN code was also required. The cumulative limit was increased to EUR 100 on March 19 2020, while the transaction limit was raised to EUR 50 on March 24 2020. Storekeepers stimulated people to pay electronically as it lowered the likelihood of hand contact. For example, they had door plates outside and put notices next to the counter asking people to pay electronically. Moreover, the Centraal Bureau Levensmiddelenhandel (CBL) – the Dutch organisation that looks after the interests of supermarkets – appealed to consumers to pay contactless.1 During the first months of the pandemic, consumers mentioned more frequently than before the pandemic that cash was declined by merchants. However, as time passed by this seemed no longer the case (DNB, 2020d). Overall, cash remained well-accepted in the Netherlands in 2020, whereas debit card acceptance increased (Panteia, 2020). On average 96.4% of the POS accepted cash in 2020. This does not significantly differ from 2019 when the cash acceptance rate was 97%. The acceptance of the debit card increased from 87% in 2019 to 92% in 2020. Regarding 2021, DNB (2021) shows that in 3% of the stores consumers have no real choice because the retailer pursues a card-only policy. This is in line with cash acceptance as reported by Panteia (2020).

From May 11 2020 the Dutch government gradually started to relax COVID-19 measures. This easing of the measures started with the re-opening of kindergartens and primary schools, a part of the services sector and re-opening of libraries. From June 1 2020 high schools re-opened, and venues in the recreation and cultural sector as well as cafes and restaurants were allowed to receive at most 30 guests. On July 1 2020 the pandemic appeared to be broadly under control, illustrated by the low number of new COVID-19 infections paired with a small number of people who needed hospital care to recover from a serious COVID-19 infection. As a result, from July 1 2020 most of the COVID-19 measures were relaxed by the government: the maximum of people that could visit a pub, restaurant or recreational/cultural venue was increased to 100, people were allowed to participate in sport competitions, and those working from home, were allowed to go to the office.

The second lockdown started on December 15 2020. POS in particular sectors such as restaurants and bars, recreation and culture were closed. Furthermore, the provision of non-medical close-contact services was not allowed. For example, hairdressing salons were closed. Moreover, shops that sell non-essential products, such as clothing stores and department stores were closed. People were still allowed to leave their home and visit non-closed POS (such as supermarkets and chemists) as often as they wanted. Day-care centres, out-of-school care centres, schools and universities were closed and people were encouraged to work from home and to avoid public transport. This lockdown ended on June 26 2021.2

The third lockdown started on December 19 2021. All publicly accessible locations had to close with the exception of some essential stores and services. Again, out-of-school care centres, schools and universities closed their doors. The impact of this most recent lockdown falls outside the scope of this paper due to the lack of available data.

Foreshadowing our main results, we find a large decline in cash usage at the POS. The share of POS transactions paid in cash declined from 31% at the beginning of 2020 to 13% in the first two weeks of April, reverting back to 21% of the transactions by October 13 2020. During the second lockdown that started in December 2020 and lasted to June 2021, the share of cash hovered around 23–25%. The results indicate that the pandemic accelerated the shift from cash to debit cards at the POS, especially during the first pandemic year. After nearly two years payment behaviour was back in line again with what one would have expected based on the autonomous linear trend. The pandemic has shifted payment preferences as well. The share of people who prefer to pay contactless by debit card has increased at the expense of the share of people who prefer to use their debit card with a PIN code. Surprisingly, although the usage of cash remained much lower than before the lockdowns, these lockdowns only slightly lowered the share of people preferring cash.

The remainder of this paper is structured as follows, Section 2 reviews the related literature on payment behaviour. Section 3 presents the context, data and method. Section 4 describes the regression results. We end with a discussion and conclusion in Section 5.

Related literature

Drivers of payment behaviour

In the past decades numerous studies were conducted on the drivers of payment patterns and how to influence them. A wide range of factors emerges. Various studies find that cash usage increases with age and decreases with education and income (e.g. Jonker 2007; Arango-Arango et al., 2018). In addition, people are more likely to opt for electronic payments when they need to pay a large amount than when the transaction size is low (Wang and Wolman 2016; Arango-Arango et al., 2018). Klee (2008) is the first to bridge the gap between theoretical models on money demand and empirical work on consumers’ payment choice. Using grocery store scanner data combined with census-tract level demographic information, she finds that transaction costs and opportunity costs are significant determinants of payment choice, and ultimately of cash demand. Moreover, prior research shows that payment choice depends on the ability to monitor liquidity (von Kalckreuth et al. 2014), keep control of one's budget (Hernandez et al., 2017) and the perceived speed of payment, user-friendliness, and safety (Jonker 2007; Schuh and Stavins 2010; van der Cruijsen and Plooij 2018). Financial incentives matter too (Arango-Arango et al., 2018; Bolt et al., 2010; Stavins 2018; Simon et al., 2010). In addition, payment behaviour depends on how well a payment instrument is accepted at the POS (Bagnall et al., 2016). Finally, there is a limited number of studies showing the importance of socio-psychological factors for payment behaviour (van der Horst and Matthijsen 2013; Khan et al., 2015; van der Cruijsen and Knoben 2021; van der Cruijsen and van der Horst 2019). For example, they show the importance of social norms (perceptions of how others pay and perceptions of how one should pay), attitudes, and feelings.

In spite of this large literature on the drivers of payment behaviour, relatively little is known about the effect of external shocks and measures taken by governments, banks and retailers on consumers’ payment behaviour. Using 2005–2008 data on the Netherlands, Kosse (2013) finds that newspaper articles on skimming fraud coincide with a somewhat lower debit card usage on the same day, but this effect does not survive in the long run. Choi and Loh (2019) show that the downsizing of ATMs in Singapore – a densely populated city – has increased customers’ travel distances to ATMs and increased their usage of the bank's digital platform.

The impact of the COVID-19 pandemic on payment behaviour

The COVID-19 pandemic offers a unique opportunity to study to what extent an external shock and accompanying measures by the government, banks and retailers can result in a change in payment behaviour and payment preferences. There are some first studies that point at mixed effects. Based on a yearly payment diary carried out in May 2020 in the United States, Kim et al. (2020) find that, in general, participants hold more cash in their wallet and as a store of value in their homes, compared to trends reported in the 2019 diary. Moreover, approximately 20% of the participants have switched from making in-person payments to paying online or over the phone. In a follow-up study Foster and Greene (2021) show that far less US citizens reported making in-person payments in spring 2020 (34%) than in fall 2019 (96%). Furthermore, those who made in-person payments, were about as likely to use cash in spring 2020 as they had been in the fall of 2019. Of people who paid in person at least once in the prior 30 days, 57% used cash at least once in October 2019, 59% in spring 2020, and 72% in August 2020.

Studies covering other countries report a (sharp) increase of card usage at the expense of cash and there are indications that part of these changes may be persistent. Chen et al. (2020) show some early survey evidence from spring 2020 that cash usage at the POS by Canadian citizens has decreased at the expense of debit and credit card payments, but that the role of cash as a store of value has somewhat increased. In particular, a third of the survey respondents reported that they had decreased their use of cash in response to the pandemic. Results from follow-up studies (Chen et al., 2021a, 2021b) indicate that part of these effects were temporary. Canadian citizens stated they made more usage of nearly all POS payment instruments, but especially of cash in July 2020 – just after the easing of containment measures limiting in-person payments in Canada – compared to April 2020. Furthermore, consumer cash holdings returned to pre-pandemic levels. According to Gutmann et al. (2021) the COVID-19 pandemic accelerated the trend decline in cash usage in Australia. Survey data suggest that the shift away from daily cash use may become permanent for many people; two-third of the people using less cash said they expected to continue to use less cash also after the pandemic is over. The survey results are supported by the reduction in the volume and the value of cash withdrawals at ATMs and a shift to online shopping, which have endured even after the easing of physical restrictions. For example, the share of online sales to total retail sales had risen sharply from 6.5% in the second half of 2019 to 10% since March 2020.

There are also studies showing that the pandemic has accelerated the use of electronic payment instruments in Europe. Four out of ten respondents of an ECB study carried out in July 2020 say they use less cash since the beginning of the COVID-19 pandemic and a majority of these people expect to continue this behaviour after the ending of the pandemic (ECB, 2020). The fact that electronic payment instruments have been made more convenient is the most often mentioned reason for the change in behaviour. Wisniewski et al. (2021) examine the influence of the COVID-19 pandemic on payment behaviour in the EU, using survey results collected in July and August 2020 of 5504 citizens from 22 European countries. The usage of cashless payment instruments increased at the expense of cash, because of the fear of getting infected by the virus by using cash. Moreover, their results suggest that this change in behaviour may be persistent.

In addition, there are studies focusing on one particular European country. The first group of studies uses survey data. Danmarks Nationalbank (2020) shows that contactless and online payments quickly gained ground in Denmark while cash payments fell during the lockdown. More specifically, 30% of the Danish respondents reported increased payment card use relative to before the lockdown, and 41% reported less cash usage. The Danish study also indicates that the use of cash gradually increased during the reopening of the economy by the end of August 2020. In addition, online payments have returned to pre-lockdown levels in Denmark. Schweizerische Nationalbank (SNB) (2021) uses survey and payment diary data collected in fall 2020. Compared with the previous study held by SNB in 2017 the share of cash payments in the number of non-recurring payments has dropped considerably from 70% in 2017 to 43% in 2020. Both improved appreciation of cashless payment instruments and the pandemic triggered the increased usage of cashless payment instruments. For instance, 36% of respondents state that they have made lasting changes to their payment behaviour as a result of the pandemic. Within this group, most people state that they intend to pay more frequently (contactless) by card, or to use less cash. This self-assessment is in line with the observed partial recovery in cash withdrawals at ATMs in the summer of 2020, while card usage at the POS remained above pre-crisis level. Using survey and payment diary data collected between August 18 and October 19 2020 Deutsche Bundesbank (2021) reports a strong growth in the usage of cashless payment instruments by German consumers during the pandemic. Compared to its previous payment survey covering 2017, the share of card payments at the POS rose by 9 percentage points to 30%, while the share of cash payments decreased from 74% to 60%. It seems likely that part of the substitution of cash by (contactless) card payments was fuelled by the pandemic. About half of the people who used contactless card payments for the first time during the pandemic said this was because of better hygiene or signs in shops. For Norway, Norges Bank (2021) finds that the share of cash in the total number of POS payments dropped from 7% in fall 2019 to 3% in spring 2020. Although the cash share partly recovered to 4% in autumn 2020, it went down again to 3% in spring 2021.

Studies on European countries that use card transaction data show a drop of card transactions during the pandemic. In Mínguez et al. (2020) information on the usage of cards as a means of payment is used to estimate the drop in consumptive expenditures during the lockdown in Spain. The study reports that immediately after the start of the full lockdown and the state of alert was declared in Spain, payment card spending and ATM withdrawals saw a drastic drop of around 50% (year-on-year). Payment card spending returned back to normal levels by the end of June 2020, while ATM withdrawals remained well below 2019 levels. Online purchases in Spain have shown a large increase. Bounie et al. (2020) investigate the impact of the COVID-19 pandemic using detailed French consumer card transaction data covering the period before and during the first year of the pandemic in France. They find a strong decline in both the value (−54%) and the volume (−61%) of card payments at the POS during the lockdown period in spring 2020, and a rise of 19% in the average transaction value. These findings suggest that French consumers made fewer shopping trips but larger purchases during the lockdown. The decline in POS expenditures using cards was approximately twice as large as the decline in online expenditures. However, online expenditures increased in some of the economic sectors for which home delivery of goods was feasible. Using transaction data from Dutch customers of ABN AMRO bank, Golec et al. (2020) attempt to separate the economic effects of voluntary responses to COVID-19 from those attributable to government lockdown measures. Their findings suggest that in municipalities with higher levels of infections, the impact on consumption is larger.3 For Switzerland, Kraenzlin et al. (2020) investigate the impact of the pandemic on card usage in the Swiss retail sector using card transaction data covering the year before the pandemic (January - May 2019) and the pandemic year 2020 (January – May 2020). Apart from aggregate effects on retail spending, they provide evidence for pronounced regional shifts – which persist post-lockdown – based on retail card payment spending across areas with different levels of urbanization and across the Swiss cantons. E-commerce and cash substitution are identified as main drivers.

To summarise, there is compelling evidence that the pandemic has affected payment patterns in various countries around the globe. However, the extent and duration varies a lot across countries. Moreover, in some countries the shifts in payment patterns were temporary, whereas in other countries, like Australia, Norway and Switzerland, they are expected to persist longer. A key advantage of our payment diary data set compared to survey data used in other studies is that we use a continuous daily series covering both the pre-COVID-19 period and the COVID-19 period. Moreover, we have detailed information on the transactions and respondents. Our main value added, in contrast to studies that use card transaction data, derives from the fact that we use information on both electronic and cash payments, joint with information on payment preferences.

Disease transmission via payment instruments

There are several studies on the time the coronavirus survives on payment instruments, which show very mixed results. Chin et al. (2020) find that a banknote could remain infectious for a period of 4 days after a droplet of the virus was put on it. The virus can remain stable on plastic surfaces for a week. Harbourt et al. (2020) shows that the temperature matters; the higher the temperature, the shorter the virus was detectable on US banknotes. The virus maintained its stability on banknotes for only one hour in the research by Caswell et al. (2020). After five more hours, only 5% of the virus was still present. In contrast, Riddell et al. (2020) find that even after 28 days after the coronavirus was put on the banknote it is still present. Van Doremalen et al. (2020) show that for coins the type of metal matters. The virus survives much longer on stainless steel than on copper.

Several studies show that the likelihood of getting COVID-19 via payment instruments is low. For example, Tamele et al. (2021) report a very low risk of getting the virus via the use of cash. Schijven et al. (2021) also find a very low risk. Even in the worst case scenario the risk that a person gets COVID-19 via a person-in-person cash transaction is less than once every 107 years. In line with this, Todt et al. (2021) show that the chance of transmission through cash and payment cards is unlikely.

The signals that the public received regarding the transmission of the coronavirus via payment instruments changed throughout the pandemic. At the start of the pandemic there were fears of getting the virus via person-in-person payment transactions. As a result, limits on contactless payments were increased in many countries to stimulate contactless payments (Mastercard 2020). There were also central banks that quarantined and disinfected cash (Auer et al., 2020; King and Shen 2020). However, soon it became clear that the risk of infection via payment instruments is very low. Various central banks, such as the European Central Bank (ECB) and De Nederlandsche Bank (DNB), communicated this to the public at large (e.g. Panetta 2020, DNB 2020c). Note also that DNB is making agreements with banks, retailers and civil society organisations within the National Forum on the Payment System to keep cash accessible and available within the Netherlands (DNB 2022). This is key to ensure that everyone can continue to participate in the payment system.

Data and method

Payment diary data

To analyse the influence of the pandemic, and the two lockdown periods on consumers’ payment behaviour at physical POS we use unique payment diary data collected from Dutch consumers. DNB and the Dutch Payments Association (DPA) commissioned the data collection. The main goal of the DNB/DPA Survey on Consumers’ Payments (SCP) is to measure payment behaviour of Dutch consumers (Jonker et al., 2018). Members of the GfK market research-panel, aged 12 years or older fill in the questionnaire. The results give a representative picture of payment behaviour at the POS of the Dutch. All days are adequately covered and provide a representative picture (Jonker et al., 2018, p.12). The resulting dataset consist of different respondents for different points in time. Respondents can participate more than once, but there is at least three months in between. Note that the reported payment behaviour does not depend on being a new or trained participant (Hernandez et al., 2017). Data from the SCP were used before to research consumer payment behaviour (Jonker and Kosse, 2013; Jonker et al., 2017; van der Cruijsen et al., 2017; Arango-Arango et al., 2018; van der Cruijsen and Knoben, 2021).

Survey participants register their payment behaviour on the registration day.4 They give detailed information on the transactions they made during the day, such as the payment instrument used, how much they spent at each POS, and what sector the POS belongs to. In addition, participants answer an additional questionnaire. We use this latter part to get insight in payment preferences and to construct variables capturing personal characteristics, such as gender and age.

For our analysis of payment behaviour we use data from January 1 2018 until December 18 2021, so our data ends before the introduction of new COVID-19 measures in response to the third wave of COVID-19 infections, which came into effect on December 19 2021. We use those payment dairies where the respondent made at least one payment at a POS on the registration day. Overall, respondents made about 131 thousand POS payments. We exclude payments that were not made with cash or the debit card, leaving us with 125,651 POS payments. We focus on cash and debit card usage, as these are by far the most frequently used means of payment at the POS in the Netherlands. For our analysis of payment preferences, we use data collected from January 1 2019 until December 18 2021; 2018 is not included because payment preferences were not part of the 2018 SCP. This results in 72,478 payment diaries of 50,479 different people. So, on average respondents in our payment preference sample participated 1.4 times and we have 67 diaries per day. In particular, 98% of the payment diaries were filled in online and 2% of the respondents partook by telephone.

Fig. 1, Fig. 2 provide a first impression on the development of payment behaviour, respectively payment preferences over time. Fig. 1 shows 14-days moving averages of the share of POS payments made by cash or by debit card by age category from January 14 2018 until December 18 2021. The figure highlights four key moments in time: the start of the first lockdown on March 16 2020, the end of the first lockdown on July 1 2020, the start of the second lockdown on 15 December 2020 and the end of the second lockdown on June 26 2021.

Fig. 1

Payment behaviour during 2018–2021.

Fig. 2

Payment preferences during 2019–2021.

The pandemic triggered a strong decline in cash usage (Fig. 1). At the start of 2020, the proportion of cash in the total number of cash and debit card POS payments still stood at 31%. Bottoming out at 14% on 12 April 2020, cash transactions rebounded to 23% at the end of June 2020, but gradually dropped again and amounted around 21% at the end of November 2020, still well below pre-pandemic levels. At the start of the second lockdown period on 15 December 2020 the share of cash stood at 25%, at the end of the second lockdown on June 26 2021 it had declined to 23% and on December 18 2021 it stood at 24%. While the decrease in cash use is seen across all age groups, it is more pronounced amongst consumers above 65 than amongst consumers between 12 and 64. Furthermore, since the second half of July 2020 no major changes have taken place anymore in consumers’ payment behaviour.

During the pandemic, consumers bundled some of their purchases to reduce the number of shopping trips and because of the closure of some physical POS locations. There was a sharp drop of 22% in the average number of daily POS payments from 1.5 POS payment per respondent per day before the first lockdown in 2020 to 1.2 POS payments per respondent per day during the first lockdown. The second lockdown – in which all non-essential sectors were closed – also led to a decline from on average 1.3 POS payments per day per respondent between the two lockdowns to 1.1 POS payments per respondent per day during the second lockdown. At the end of the second lockdown the average number of POS payments per respondent per day increased to 1.4.

The change of the average value of POS payments points at bundling of transactions too. The average value of POS payments increased by 13% from EUR 22.58 before the first lockdown in 2020 to EUR 25.51 during the first lockdown. The second lockdown led to a rise in the average transaction size from EUR 24.93 between the two lockdowns to EUR 28.38 during the second lockdown. The average transaction size of supermarket payments shows a similar pattern, indicating that the change in average transaction size was not caused by the change in products people bought.

Payment preferences have considerably changed as well since the start of the lockdown; contactless payments clearly gained ground. Fig. 2 shows 14-days moving averages of the share of respondents preferring different payment methods. Since March 16 2020 substantially more consumers preferred paying contactless, whereas the share of people preferring to pay with their debit card in the traditional way (so including a PIN code) decreased. Both paying contactless by debit card and mobile phone became more popular. The share of respondents preferring these payment instruments increased from, respectively 38% and 7% on March 15, 2020 to 51% and 13% on December 18, 2021, resulting in a combined increase of 19 percentage points. This occurred at the expense of the share of people preferring to use the debit card in a traditional way, which dropped from 29% to 14%. The share of people preferring to pay with cash only declined by 2 percentage points, from 21% to 19%.

Some of the observed change in payment behaviour may not be due to shifted preferences, but rather to behaviour adjusting more quickly towards actual preferences. Van der Cruijsen et al. (2017) report a gap between payment preferences and actual payment behaviour. At the time of their research half of the Dutch consumers who preferred the debit card to cash did not use the debit card to pay a majority of their POS transactions. They conclude that changing payment behaviour is a challenging task; even when consumers have changed their preferences, they find it hard to change their behaviour. The pandemic might have helped to align people's payment behaviour more quickly with their payment preferences.

The overall impact of the pandemic on retail consumption by consumers was relatively moderate. The payment diary data show a decline of 1% in the total nominal value of POS and online sales in 2020 and an increase of 5% in 2021, both relative to retail consumption in 2019. However, the lockdown periods clearly had an impact on the timing of consumption during the year and on the way people made their purchases; POS or online. Fig. 3 presents the total value of monthly payments at the POS, online payments, and the deviation of the total sum of POS and online sales relative to the total sum of these sales in the same month in 2019.

Fig. 3

Value POS and online consumption.

The first period of the first lockdown which started in March 2020 was marked by plummeting payment transactions, in terms of both numbers and value. POS payments showed the sharpest decline, while online payments remained fairly stable.5 The total nominal value of POS and online payments bottomed-out in April 2020 when it was 19% lower than in April 2019. From May onwards economic activity gradually recovered. The resurgence showed such strength that, by June the total value of POS and online consumer payments together was in fact 18% higher than in June 2019. In April and May, the rebound in consumption could be observed mainly in online sales in product categories such as food, health care, fashion, domestic appliances, media & entertainment and toys, but from June onwards also POS consumption was back on track (+5%). Between June 2020 and the first weeks of October 2020 – a period with only minor restrictions – consumption was above its 2019 level.

However, due to rising COVID-19 infections, the Dutch government gradually announced new restrictions starting from mid-October 2020, and it announced a full lockdown between December 15 2020 and June 26 2021. This lockdown led to a second drop in consumer expenditures between November 2020 and February 2021 relative to the same period in 2019. Although the second lockdown was more severe than the first, as it also included a curfew and closure of non-essential shops, Dutch consumers and the Dutch retail sector were better prepared. Online expenditures by consumers almost instantly showed double digit growth figures compared to the same period in 2019, which was not the case yet in the first lockdown. The rapid substitution of POS by online sales probably led to a less severe downfall in consumer expenditures than otherwise would have been the case.

There are indications that during the second pandemic year 2021 the substitution of POS by online sales was more widespread than during the first pandemic year. In 2020 usage of online sales grew relatively strong amongst people aged 19 – 34 years, who already made many online purchases before the pandemic, whereas in 2021 there was also a strong uptake in online sales amongst people aged 35 – 74 years.

The type of products different age groups started to buy online during the pandemic differed. In the first pandemic year people aged 34 and younger more frequently bought (electronic) household appliances, fashion, and products for media & entertainment than in 2019, whereas food and beverages were mainly more frequently bought by people aged 45 and over, and health care products by people aged 35 – 54 and 75 and over. In the second pandemic year – when also non-essential shops were closed for some time – online sales of (electronic) household appliances, fashion and on media & entertainment went up amongst almost all age groups, while people aged 25 – 34 also made more online purchases of food and beverages.

Empirical approach

To test whether consumers’ payment behaviour and payment preferences have changed due to COVID-19 and the accompanying measures, we estimate different sets of regressions.

Dependent variables

We use Debitcard as dependent variable. This dummy variable is 1 for debit payments and 0 for cash payments. Debit payments include both traditional debit card payments, contactless payments with the debit card and contactless debit payments initiated via an app on people's smartphone.6 Prior to the pandemic, cash was accepted by circa 97% of the retailers and the debit card by approximately 87% (DNB, 2020b).

For the analysis of payment preferences we construct a variable that captures consumers’ preferred payment instrument. This variable Paymentpreference is 1 for respondents who prefer to use cash, 2 for respondents who prefer to pay in a traditional way with their debit card (using a PIN code), 3 for respondents who prefer to pay contactless with their debit card and 4 for respondents who prefer to pay with their mobile phone.7

Explanatory variables

We use several variables related to the pandemic and the subsequent government measures to explain the choice between debit or cash. We use Lockdown_1_start, a dummy that takes the value 1 since the start of the first lockdown in the Netherlands, so from March 16 2020 onwards. We also include the dummy Lockdown_1_end which takes on the value 1 from July 1 2020 onwards when most of the government measures were relaxed, and which allows us to analyse whether people are returning to their old pre-pandemic payment behaviour. The variables Lockdown_2_start and Lockdown_2_end capture the start and the end of the second lockdown. In particular, Lockdown_2_start takes on the value 1 from December 15 2020 onwards, whereas Lockdown_2_end is 1 from June 26 2021 onwards. Furthermore, we include the continuous variable COVID_19 which equals the number of daily new infections with COVID-19 per 100,000 inhabitants in the respondent's province of residence at day t (source: National Institute for Public Health and the Environment). This variable reflects the incidence of COVID-19 contamination in the geographical area where the respondent lives.8 As there was a shortage in test capacity until June 1 2020, only people with serious COVID-19 symptoms and people working in health care were allowed to get themselves tested for COVID-19. After June 1, test capacity was increased and also people with mild symptoms have been encouraged by the government to get themselves tested. This resulted in a more accurate measurement of the number of new COVID-19 infections, and also in an increase in the measured number of new COVID-19 infections. In order to correct for the change in test policy we also include variable COVID_19_June_2020 that equals COVID_19 from June 1 onwards, but is zero before June 1 2020. Fig. 4 shows that there are provincial differences in the incidence of COVID-19 contamination. Furthermore, it shows that the number of new COVID-19 infections before June 1 is much smaller than after June 1 when the test capacity increased.

Fig. 4

The number of daily new COVID-19 infections per 100,000 inhabitants by province during 2020–2021.

In addition, we control for a wide range of other variables. In all regressions, we include the following individual-specific dummy variables in the set of explanatory variables: Male, Between_25_34, Between_35_44, Between_45_54, Between_55_64, 65_and_over, Education_low, Education_high, Income_low, Income_high, Income_unknown, Partner, Children, Native, and we include 11 province dummies reflecting geographical differences in payment behaviour. The reference person is a non-native, middle-educated woman who has a gross household middle income between EUR 23,400 and EUR 51,300 a year, does not live with a partner, has no children, is 24 years or younger and lives in the province of Noord-Holland. Appendix A describes all the variables in more detail and includes summary statistics.

The regressions that explain payment behaviour include additional transaction specific variables. We include the following dummies: Amount_EUR_5_10, Amount_EUR_10_20 and Amount_EUR_20_and_more as it is expected that consumers’ payment choice also depends on the transaction size. The reference category is Amount_EUR_5_and_less. We also include branch dummies to control for branch-specific payment behaviour resulting from difference in acceptance, social norms and payment patterns across branches: Retailstores_food, Retailstores_non-food, Petrolstations, Vendingmachines, Streetvending, Cafes_and_restaurants, Recreation_and_culture, Transport, and Services. The reference category is Supermarkets. Furthermore, we control for day-specific payment behaviour by including the day of the week dummies: Monday, Tuesday, Wednesday, Thursday, Friday, and Saturday. Sunday is the reference day. We also include the variables Log_average_transaction and Log_number_of_transactions that reflect the average value and the number of transactions made at the POS by the respondent at day t. During the pandemic people may have altered their purchasing behaviour, such as making fewer but larger purchases in order to reduce the risk of getting contaminated, or because of the closure of certain retail sectors. These variables capture the impact of changed purchasing behaviour at the POS during the pandemic on the choice of payment instrument. We also include the variable Trend in all regressions. This variable increases every day by 1/365 and captures the autonomous trend in debit payments and payment instrument preferences as we need to control for the fact that also in the absence of COVID-19 and the accompanying measures a shift in payment behaviour and preferences would have occurred.

Methodology

Payment behaviour

Regarding payment behaviour at the POS, we use a series of pooled binomial logit regression models to analyse consumers’ choice of payment instrument when paying for a purchase i at a given day t and the possible impact of the pandemic on it. See Cameron and Trivedi (2010) for a description of the logit model. We focus on the two dominant payment instruments: cash and the debit card who together cover more than 95% of the POS payments in the Netherlands, and we exclude transactions with other payment instruments.

We distinguish between the observed usage of the debit card for transaction i at day t Debitcard, and the underlying continuous latent variable . In our baseline model we assume that depends on respondent characteristics x,9 transaction characteristics w and calendar effects c, which include the trend and the day of the week dummies:

We do not observe the value of the latent variable, but we do see whether the debit card or cash was used:

We then have

Section 3.2 provides an overview of the covariates included in x and c . We assume that are from the logistic distribution, with zero mean and standard deviation (Cameron and Trivedi, 2010). As this analysis takes place at transaction level instead of respondent level, and many respondents have made multiple purchases, we need to cluster the standard errors by respondent to take potential correlation across transactions made by the same respondent into account.

In the next step of our analysis, to assess the influence of the lockdown in the Netherlands on debit card usage at the POS, we include several variants of time dummies. First, we include lockdown dummies so that the argument of F in (3) becomesto assess whether the lockdown has raised debit card usage at the start of the first lockdown , and whether it has a lasting impact on people's payment behaviour. Regarding persistence, we distinguish between the overall impact of the first lockdown and the overall impact of the two lockdowns Second, we add in (4) two continuous variables reflecting the daily number of newly registered COVID-19 infections in the geographical area where the respondent livesin order to assess whether the actual degree of contamination in the respondent's province influences his payment behaviour. Third, we assess whether the lockdown has a heterogeneous impact on different consumer segments and in different branches by adding in (5) interaction termswhere denotes the different age and income groups and the different branches.

Regarding consumer segments we focus on age cohorts and income segments, as both age and income are strong predictors for debit card usage (see e.g. Jonker, 2007; Bagnall et al., 2016).10 The impact of the lockdowns on card usage may differ between sectors, as there is a lot of variation in debit card usage between them (DNB, 2020a) and because the way government measures have affected business differs per retail branch.

The impact of the start of the lockdowns and the subsequent easing of the government measures on debit card usage is reflected in the corresponding marginal effects from Eqs. (4) and (5). Whether the first lockdown influences debit card usage in the short run, can be assessed by testing . Rejection implies that the lockdown did affect the likelihood of debit card usage just after the start of this lockdown. Whether the first lockdown has a long-lived impact can be assessed by testing and the overall impact of the two lockdowns can be assessed by testing .

The coefficients in Eq. (6) reflect the influence of the lokdowns on the likelihood of debit card usage for a payment by age, income and branch. The influence of the start of the first lockdown by age and income is captured by , and its impact by branch by . Whether the start of the lockdown led to a change in the likelihood of a debit card transaction for specific age and income groups can be assessed by testing = 0. Analogously, we test the influence of the beginning of the lockdown on debit card usage by branch. The long-term effect of the first lockdown on debit card usage by age and income and branches can be used to assess whether the first lockdown has had a long-lived impact on the likelihood of debit card usage by testing: , respectively =0. Acceptance indicates that the likelihood of debit card usage by people in that specific age or income group or branch has returned to its pre-lockdown trend level before the start of the second lockdown. Analogously, the overall long-term effects of the two lockdowns on debit card usage by age and income and branches can be used to assess the long-term effect of the two lockdowns on the payment behaviour at the POS of people in these specific age and income groups or branches. Acceptance of the test that the cumulative effects of the two lockdowns for specific age or income groups or branches are not significantly different from zero suggests that the likelihood of debit card usage by people in that specific age or income group or branch has returned to its pre-pandemic trend level.

Payment preferences

Regarding payment preferences, we use a similar approach. We use a series of multinomial logit regression models to analyse consumers’ payment preferences. First, we run a regression with the set of controls and Lockdown_1_start, Lockdown_1_end, Lockdown_2_start and Lockdown_2_end to test the impact of the pandemic and accompanying measures on the likelihood of preferring particular payment instruments. Second, we include COVID-19 control variables. Third, we include interaction terms with the age and income dummies to test whether the effect of the pandemic differs for different age and income groups.

Regression results

Shift in payment behaviour towards more card payments

Our main finding is that the COVID-19 pandemic has led to an increase in the likelihood of a debit card transaction at the POS in the Netherlands at the expense of cash, see Table 1 (column 2) and Fig. 5 with the realized and predicted shares of debit card payments in the total number of POS payments. Since the start of the first lockdown on March 16 2020 the likelihood that consumers use their debit card instead of cash has increased by 12 percentage points compared to debit card usage before the lockdown. The main part of the shift in payment behaviour lasted seven months after the start of the lockdown as the Wald test based on: using the estimation results in column 3 Table 1 is rejected at the 1% level of significance (p = 0.0002). Part of the change was still there after the end of the second lockdown as shown by the outcome of the Wald test: which is also rejected at the 1% level of significance (p = 0.001), see Table 2 . From July 1 2020 onwards the likelihood that consumers use a debit card for a POS transaction has fallen by 5 percentage points, followed by two further reductions of nearly 2 percentage points at the start of the second lockdown on December 15 2020 and its ending on June 26 2021. In total 3 percentage points of the initial 12 percentage points increase in debit card usage above the trend level appears to be long-lived. As times goes by more information will become available on the temporary or permanent nature of the change.11

Table 1

COVID-19 and payment behaviour: regression results, average marginal effects based on logit regressions.

	(1)	(2)	(3)	(4)
	Debit card	Debit card	Debit card	Debit card
Lockdown_1_start		0.12***	0.11***	0.05
Lockdown_1_end		−0.05***	−0.04***	0.02
Lockdown_2_start		−0.02**	−0.02*	−0.01
Lockdown_2_end		−0.02**	−0.02**	−0.02
COVID_19			0.01**	0.01**
COVID_19_June_2020			−0.01**	−0.01**
Male	0.01*	0.01**	0.01*	0.01**
Between_25_34	−0.03***	−0.03***	−0.03***	−0.02*
× Lockdown_1_start				0.01
× Lockdown_1_end				−0.03
× Lockdown_2_start				−0.03
× Lockdown_2_end				−0.03
Between_35_44	−0.05***	−0.05***	−0.05***	−0.04***
× Lockdown_1_start				0.02
× Lockdown_1_end				−0.03
× Lockdown_2_start				−0.02
× Lockdown_2_end				0.01
Between_45_54	−0.09***	−0.09***	−0.09***	−0.09***
× Lockdown_1_start				0.04
× Lockdown_1_end				−0.06*
× Lockdown_2_start				0.01
× Lockdown_2_end				−0.03
Between_55_64	−0.13***	−0.13***	−0.13***	−0.13***
× Lockdown_1_start				0.09***
× Lockdown_1_end				−0.07**
× Lockdown_2_start				−0.03
× Lockdown_2_end				−0.01
65_and_more	−0.16***	−0.16***	−0.16***	−0.18***
× Lockdown_1_start				0.10**
× Lockdown_1_end				−0.05
× Lockdown_2_start				0.01
× Lockdown_2_end				−0.02
Education_low	−0.04***	−0.04***	−0.04***	−0.04***
Education_high	0.05***	0.05***	0.05***	0.05***
Income_low	−0.05***	−0.05***	−0.05***	−0.04***
× Lockdown_1_start				−0.05**
× Lockdown_1_end				−0.00
× Lockdown_2_start				−0.02
× Lockdown_2_end				0.03
Income_high	0.06***	0.05***	0.06***	0.05***
× Lockdown_1_start				−0.01
× Lockdown_1_end				0.01
× Lockdown_2_start				0.02
× Lockdown_2_end				−0.01
Income_unknown	−0.01**	- 0.01**	- 0.01**	0.01
× Lockdown_1_start				−0.00
× Lockdown_1_end				0.00
× Lockdown_2_start				−0.03
× Lockdown_2_end				0.03
Partner	0.02***	0.02***	0.02***	0.02***
Children	−0.01*	−0.01*	−0.01*	−0.01
Native	0.01	0.01	0.01	0.01
Province_Groningen	0.02	0.02	0.01	0.01
Province_Friesland	0.02*	0.02*	0.01	0.01
Province_Drenthe	−0.00	−0.00	−0.01	−0.00
Province_Overijssel	−0.01	−0.01	−0.02**	−0.02**
Province_Gelderland	−0.00	−0.00	−0.01	−0.01
Province_Utrecht	0.03***	0.03***	0.02**	0.02***
Province_Zuid-Holland	0.01	0.01	0.00	0.00
Province_Zeeland	−0.02	−0.02	−0.02*	−0.02*
Province_Noord-Brabant	−0.02**	−0.02**	−0.02***	−0.02***
Province_Limburg	−0.08***	−0.08***	−0.08***	−0.08***
Province_Flevoland	0.04***	0.04***	0.03***	0.03***
Amount_EUR_5_10	0.08***	0.08***	0.08***	0.08***
Amount_EUR_10_20	0.13***	0.13***	0.13***	0.13***
Amount_EUR_20_and_more	0.20***	0.19***	0.20***	0.19***
Retailstores_food	−0.07***	−0.07***	−0.07***	−0.08***
× Lockdown_1_start				0.04**
× Lockdown_1_end				−0.04
× Lockdown_2_start				0.03
× Lockdown_2_end				−0.02
Retailstores_non-food	−0.02***	−0.02***	−0.02***	−0.02***
× Lockdown_1_start				0.03
× Lockdown_1_end				−0.03
× Lockdown_2_start				−0.02
× Lockdown_2_end				0.06**
Petrolstations	0.04***	0.04***	0.04***	0.05***
× Lockdown_1_start				−0.09***
× Lockdown_1_end				0.05
× Lockdown_2_start				0.04
× Lockdown_2_end				−0.03
Vendingmachines	0.06***	0.06***	0.06***	0.07***
× Lockdown_1_start				−0.08*
× Lockdown_1_end				0.04
× Lockdown_2_start				−0.01
× Lockdown_2_end				0.07*
Streetvending	−0.30***	−0.31***	−0.31***	−0.37***
× Lockdown_1_start				0.09***
× Lockdown_1_end				0.06*
× Lockdown_2_start				−0.03
× Lockdown_2_end				−0.04
Cafes_and_restaurants	−0.09***	−0.08***	−0.09***	−0.09***
× Lockdown_1_start				0.07***
× Lockdown_1_end				−0.07**
× Lockdown_2_start				0.05**
× Lockdown_2_end				−0.03
Recreation_and_culture	−0.19***	−0.18***	−0.18***	−0.20***
× Lockdown_1_start				0.17**
× Lockdown_1_end				−0.13
× Lockdown_2_start				0.02
× Lockdown_2_end				0.03
Transport	−0.08***	−0.08***	−0.08***	−0.09***
× Lockdown_1_start				−0.03
× Lockdown_1_end				0.17*
× Lockdown_2_start				−0.10
× Lockdown_2_end				−0.11
Services	−0.29***	−0.29***	−0.29***	−0.26***
× Lockdown_1_start				−0.09***
× Lockdown_1_end				0.01
× Lockdown_2_start				−0.01
× Lockdown_2_end				0.01
Monday	0.02***	0.02***	0.02***	0.02***
Tuesday	0.03***	0.03***	0.03***	0.03***
Wednesday	0.03***	0.03***	0.03***	0.03***
Thursday	0.03***	0.03***	0.03***	0.03***
Friday	0.04***	0.04***	0.04***	0.04***
Saturday	0.03***	0.03***	0.03***	0.03***
Log_average_transaction	0.03***	0.03***	0.03***	0.03***
Log_number_of_transactions	0.00	0.01*	0.01**	0.01**
Trend	0.04***	0.03***	0.03***	0.03***
Pseudo R2	0.126	0.130	0.131	0.134
Log-pseudolikelihood	−67,360.0	−67,107.1	−65,509.0	−65,271.5
Wald χ2	9059.9***	9173.3***	9140.9***	9548.4**

Note: The number of observations is 125,651 in columns 1 and 2, and 122,616 in columns 3 to 4. The dependent variable is Debitcard. This variable equals 1 for debit payments and 0 for cash payments. Debit payments include both traditional debit card payments, contactless payments with the debit card and contactless debit payments initiated via an app on people's smartphone. Standard errors are clustered at respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Fig. 5

Realized and estimated share of debit card in the total number of cash and debit card payments, January 1 2018 – December 18 2021.

Table 2

Wald tests: impact lockdowns on debit card usage by age, income and branch.

	Wald test Lockdown start1		Wald test Lockdown start1 +lockdown end1		Wald test Lockdown start1 +lockdown end1 +Lockdown start2 +lockdown end2
	Wald χ2	p-value	Wald χ2	p-value	Wald χ2	p-value
All transactions (column 3)	98.62***	0.00	77.23***	0.00	10.84***	0.00
Reference (column 4)	2.62	0.11	9.81***	0.00	2.32	0.13
Age
Between_25_34	2.91*	0.09	2.87*	0.09	1.62	0.20
Between_35_44	8.50***	0.00	7.92***	0.00	0.74	0.39
Between_45_54	15.70***	0.00	7.28***	0.01	0.00	0.99
Between_55_64	34.27***	0.00	20.74***	0.00	0.60	0.44
65_and_more	50.91***	0.00	50.41***	0.00	16.50***	0.00
Income
Income_low	0.03	0.86	1.06	0.30	0.01	0.92
Income_high	1.70	0.19	9.81***	0.01	4.66**	0.03
Income_unknown	2.78*	0.10	12.71***	0.00	3.07*	0.08
Branch
Retailstores_food	6.25**	0.01	8.04***	0.00	3.64**	0.06
Retailstores_non-food	5.03**	0.03	8.03***	0.00	6.98***	0.01
Petrolstations	0.95	0.33	0.97	0.33	0.03	0.86
Vendingmachines	0.28	0.60	0.88	0.35	2.94*	0.09
Streetvending	12.29***	0.00	49.28***	0.00	13.25**	0.00
Cafes_and_restaurants	10.02***	0.00	7.26***	0.01	5.40**	0.02
Recreation_and_culture	7.19***	0.01	7.00***	0.01	8.99***	0.00
Transport	0.05	0.83	8.86***	0.00	0.32	0.57
Services	0.89	0.35	0.25	0.62	1.89	0.17

When we include COVID_19 and COVID_19_June_2020, we observe that the number of new COVID-19 infections correlates positively with debit card usage during the first months of the pandemic (column 3). For every newly infected person per day per 100 thousand inhabitants the likelihood that a purchase is paid with a debit card increases by 1 percentage point. However, after June 1, new infections do not influence the likelihood of debit card usage anymore, as the estimated effect of the interaction term COVID_19_June_2020 is negative and of the same order of magnitude as the effect of COVID_19. We use a Wald test to test whether the sum of these parameters differs significantly from zero. The results show that their cumulative effect does not differ significantly from zero (p = 0.2433). This finding therefore suggests that before June 1 the number of new infections per day influenced people's payment behaviour at the POS, but that after June 1 this effect has become negligible. Furthermore, the estimated effects of the two dummies for the first lockdown changed only mildly and the estimated effects for the dummies for the second lockdown remain unaltered. The marginal effect of Lockdown_1_start is reduced by one percentage points to +11 percentage points and the marginal effect of Lockdown_1_end changed from −5 percentage points to −4 percentage points. These results suggest that the larger part of the change in debit card usage from March 2020 onwards stems from the measures taken by the government, banks and retailers and that only a small part appears to be attributable to the prevalence of the COVID-19 virus amongst the Dutch.

Regarding the other control variables, most outcomes are in line with expectations. All the effects as listed in columns 1–3 are average marginal effects for the entire period of our data. In column 4 we also show average marginal effects for age, income and branch interacted with the four lockdown dummies. We find that debit card usage decreases with age. For instance, consumers aged between 35 and 44 are 5 percentage points less likely to pay a POS transaction with the debit card than those aged between 12 and 24 (reference group, columns 1–3), while people of 65 and older are 16 percentage points less likely to use their debit card than the reference group. Furthermore, we find that the likelihood of using the debit card instead of cash increases with education and income. In addition, people who have a partner are 2 percentage points more likely to use their debit card for a POS payment than those without a partner. In addition, we find small effects for gender and having children. There are also regional differences. People living in the province Limburg (southern part of the Netherlands) are 8 percentage points less likely to use the debit card than those living in the province Noord-Holland (western part including the capital city Amsterdam; reference group), while people living in the province Flevoland are 4 percentage points more likely to use the debit card.

Characteristics of the transaction and people's purchasing behaviour at day t matter. Debit card usage strongly increases with the transaction amount. For example, consumers are 20 percentage points more likely to use the debit card when the transaction amount is EUR 20 or more than when the amount is EUR 5 or less (reference group). Furthermore, we find that the likelihood of using a debit card increases with the average value of the purchases made at day t, but is not affected by the number of POS payments made. Debit card usage also strongly depends on the branch. Consumers are least likely to pay by debit card in case of street vending; the likelihood of debit card usage is 31 percentage points lower than in the supermarket (reference group). They are the most likely to use the debit card at vending machines for drinks and snacks (+6 percentage points).

Regarding calendar effects, we find moderate effects for trend and day of the week. The variable trend shows that even without the pandemic, there is an upward trend of increased debit card usage at the expense of cash. Without the inclusion of the four lockdown dummies, the estimated yearly rise in the share of debit card payments is estimated at 4 percentage points (column 1). When the lockdown dummies are included (columns 2 - 4) the estimated yearly rise in the likelihood of debit card usage is 3 percentage points. These findings suggest that the pandemic appears to have triggered a rise in debit card usage within half a year time during the first lockdown that would usually have taken around 2–3 years of time, as the average Trend indicates a 3 percentage points yearly rise in debit card usage (column 2).12

When including the interaction terms (Table 1, column 4), the marginal effects of the four lockdown dummies decrease and amount to +5 percentage points for Lockdown_1_start, +2 percentage points for Lockdown_1_end, −2 percentage points for Lockdown_2_start and −1 percentage point for Lockdown_2_end (column 4).13 These outcomes can be interpreted as follows: since the start of the first lockdown the likelihood that a purchase done in the supermarket by a 12–24 year old with an intermediate household income has increased by 5 percentage points, and there is no evidence of a partial turning back to cash after the end of the second lockdown on June 26 2021 (see also the discussion of the Wald tests in Table 2). None of the overall lockdown dummies remain significant, showing that the effect is heterogenous across the various groups.

The lockdown has had the strongest effect on the payment behaviour of elderly people. The estimated interaction effects with age show no significant difference in the way the lockdown has influenced debit card usage of 25–44 year olds and 12–24 year olds (reference group). However, the start of the first lockdown has led to a 9 percentage points higher rise in the likelihood of debit card usage amongst people between 55–64 years of age than amongst younger people. For people aged 65 and over the impact on debit card usage is slightly higher, i.e. +10 percentage points. So elderly people, who pre-pandemic used more cash, have reacted stronger to the first lockdown than younger people. For the interaction effects with Lockdown_1_end, we find a significant effect for the 45–54 year and the 55 −64-year-olds. The likelihood that they pay with the debit card after July 1 2020 reduced 6, respectively 7 percentage points more than for the reference group. For people older than 65 years we also find negative interaction effects, but these are not significant. We do not find significant interaction effects with age for the dummies indicating the start and the end of the second lockdown.

The impact of the first lockdown also differs by household income. The rise in the likelihood of debit card usage at the start of the first lockdown is less steep for people in the lowest income group (−5 percentage points) than for people in the intermediate income group (reference group). There is no difference in the effect of the lockdown between people in the intermediate and the other two income groups. Again, we do not find significant interaction effects for the dummies related to the second lockdown.

The impact of the two lockdowns also differs by branch. The initial rise in debit card usage in the first lockdown is stronger in cash intensive branches, such as recreation and culture (+17 percentage points), street vending (+9 percentage points), cafes and restaurants (+7 percentage points) and specialized food stores (+4 percentage points) than in supermarkets (reference). In three of these branches debit card usage drops relatively strong after July 1 2020 compared to what happens at supermarkets, i.e. in recreation and culture by −13 percentage points, in cafes and restaurants by −7 percentage points and in street vending by −6 percentage points, indicating a (partial) reversion to cash. In specialized food stores there is no significant evidence of a relatively stronger return to cash than in supermarkets. In three branches, the likelihood of using the debit card has dropped since the start of the first lockdown relative to debit card usage in supermarkets: at petrol stations (−9 percentage points), in services (−9 percentage points) and at vending machines (−8 percentage points). For petrol stations and vending machines, this may be explained by the already high share of debit card payments prior to the lockdown, which was above 80% before the pandemic (DNB, 2020b). Inspection of the data shows that debit card usage hardly altered during the lockdowns in these branches, reducing the gap in the share of debit card payments between them and the supermarkets where debit card usage did rise. There are three significant interaction effects of the second lockdown with branches. The start of the second lockdown triggered a relatively stronger increase in debit card usage in cafes and restaurants (+5 percentage points) than in supermarkets, which is not reversed by the end of the second lockdown. Furthermore, at the end of the second lockdown, there is a relatively strong increase in debit card usage in non-food retail stores (+6 percentage points) and retail vending (+7 percentage points) compared to supermarkets.

Using the estimation results from column 4 in Table 1, we test the impact of the first lockdown and the cumulative effect of the two lockdowns for different age groups and branches. For the reference group, i.e., payments in the supermarket by people in the age group 12 – 24 years old with a medium income, the estimated effect of the start of the first lockdown on the likelihood of debit card usage is positive and is almost significantly different from zero at the 10% level (p = 0.11), suggesting that the likelihood that the reference transaction was paid with a debit card may have risen after the first lockdown. The sum of the estimated effects of the start and the end of the first lockdown is positive and differs significantly from zero. However, the estimated cumulative effect of the four lockdown dummies together does not differ significantly from zero. So, there is some evidence of a lasting effect of the pandemic on debit card usage for the reference transaction just after the first lockdown, but this effect did not persist after the end of the second lockdown.

For the five other age groups the start of the first lockdown – as measured by the cumulative effect of Lockdown_1_start and the corresponding age group interacted with Lockdown_1_start – has a significant positive impact on debit card usage. This also holds for the cumulative effect of the start and the end of the first lockdown, also including interaction terms with age for these five age groups. However, the cumulative effect of the start of the two lockdowns and the end of the two lockdowns, including interaction terms with age group, is only significantly different from zero for people aged 65 and over. These results indicate that the pandemic has a persisting impact on the usage of the debit card only for the oldest age group. People aged between 25 and 64 used the debit card more frequently after the start of the first lockdown, and also just after the end of the first lockdown, but this effect did not last until the end of the second lockdown. So for them the pandemic only triggered a temporary change in their payment behaviour, on top of the autonomous trend.

Regarding income, the estimation results show that the likelihood of debit card usage by people who did not report their income rose at the start of the first lockdown and that it has remained relatively high also after July 1 2020 and even after the end of the second lockdown. For people with a low income, the test results indicate that the pandemic has not changed their payment behaviour at the POS; their debit card usage remained low. People with a high income did not change their payment behaviour significantly at the start of the first lockdown. However, the sum of the estimated effects of Lockdown_1_start and Lockdown_1_end for this group is positive and significantly different from zero and also the cumulative effect of all four lockdown dummies and their interaction with age. This suggests that during the pandemic the likelihood of debit card usage by them has risen.

Both in debit card intensive branches (petrol stations, vending machines, transport) and in services there is no significant impact of Lockdown_1_start, although separate regressions for these sectors (available upon request) show that the lockdown dummies do have the expected sign. This indicates that debit card usage has not changed in these branches since the start of the first lockdown. For two of these branches there is no significant cumulative effect of the dummies marking the start and end of the first lockdown and the four lockdown dummies together, indicating that debit card usage has not changed during the two lockdowns, apart from the general upward trend. For transport the cumulative effect of Lockdown_1_start Lockdown_1_end is significantly different from zero. In this branch the likelihood of a debit card payment has actually risen after July 1 2020 (Table 1; column 4). A similar effect is observed at the end of the second lockdown in the vending machine branch. For the branches specialised food stores, stores selling non-food, street vending, cafes and restaurants, and the recreation and culture branch both the start of the first lockdown and the cumulative effect of Lockdown_1_start and Lockdown_1_end as well as the cumulative effect of all four lockdown dummies are significantly different from zero, indicating that debit card usage has risen since the start of the first lockdown and stayed relatively high after July 1 2020 and after June 26 2021.

Shift in payment preferences towards more contactless payments

Stated payment preferences have clearly changed due to the pandemic; the likelihood that someone prefers to pay contactless has increased. The first lockdown has increased the likelihood that someone prefers to pay contactless by debit card by 7 percentage points (Table 3 column 1c) and the likelihood that someone prefers to pay contactless by mobile phone by 1 percentage point (column 1d). The likelihood of preferring to use the debit card in a traditional way decreased with 6 percentage points (column 1b). The likelihood that people prefer to use cash decreased as well, namely by 2 percentage points (column 1a). The variable trend shows that there is a trend of increased preference for contactless payments by phone at the expense of a preference for other payment instruments. We do not observe a reversal of payment preferences after the end of the first lockdown. On the contrary, the likelihood that people prefer to pay contactless by debit card increased by an additional 3 percentage points after July 1 2020, whereas the likelihood that someone prefers to use the debit card in a traditional way declined by an extra 1 percentage point. There is a 2 percentage points decrease in the likelihood that people prefer to pay with their mobile phone. The start of the second lockdown coincided with an additional 6 percentage points increase in the likelihood of preferring to pay contactless by card, whereas the end of this second lockdown led to an extra 1 percentage point increase. The likelihood that someone prefers to use the debit card in a traditional way or prefers to use a mobile phone further decreased during the second lockdown. The end of the second lockdown resulted in an increase of the likelihood of preferring to pay with cash of 1 percentage point. All in all, these findings indicate that the pandemic has resulted in a shift in payment preferences towards contactless payments by debit card.

Table 3

COVID-19 and payment preferences: regression results,average marginal effects based on multinomial logit regressions.

	Baseline				With new infections
	(a)	(b)	(c)	(d)	(a)	(b)	(c)	(d)
	Preference cash	Preference debit card: traditional	Preference debit card: contactless	Preference mobile phone	Preference cash	Preference debit card: traditional	Preference debit card: contactless	Preference mobile phone
Lockdown_1_start	−0.02***	−0.06***	0.07***	0.01**	−0.02**	−0.06***	0.07***	0.01
Lockdown_1_end	0.00	−0.01**	0.03***	−0.02***	0.00	−0.01	0.03***	−0.01***
Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
Lockdown_2_end	0.01*	−0.01	0.01**	−0.02***	0.01*	−0.01	0.02**	−0.02***
COVID_19					0.00	0.00	0.00	0.00
COVID_19_June_2020					0.00	0.00	0.00	0.00
Male	−0.01***	0.00	−0.02***	0.04***	−0.01***	0.00	−0.02***	0.04***
Between_25_34	−0.02**	0.04***	−0.01	−0.01	−0.02**	0.04***	−0.01	−0.01
Between_35_44	0.04***	0.07***	−0.09***	−0.01**	0.04***	0.07***	−0.09***	−0.01**
Between_45_54	0.07***	0.08***	−0.13***	−0.03***	0.07***	0.08***	−0.13***	−0.03***
Between_55_64	0.10***	0.10***	−0.15***	−0.05***	0.10***	0.10***	−0.15***	−0.05***
65_and_more	0.09***	0.15***	−0.15***	−0.08***	0.09***	0.15***	−0.15***	−0.08***
Education_low	0.05***	0.02***	−0.04***	−0.03***	0.05***	0.02***	−0.04***	−0.03***
Education_high	−0.05***	−0.02***	0.06***	0.02***	−0.05***	−0.02***	0.06***	0.02***
Income_low	0.09***	0.00	−0.07***	−0.01**	0.09***	0.00	−0.07***	−0.01**
Income_high	−0.08***	−0.02***	0.07***	0.04***	−0.08***	−0.02***	0.07***	0.04***
Income_unknown	0.03***	0.01*	−0.03***	−0.01	0.03***	0.01*	−0.03***	−0.01
Partner	−0.03***	0.01**	0.02***	0.00	−0.03***	0.01**	0.02***	0.00
Children	0.01	0.00	−0.01	0.01*	0.01	0.00	−0.01	0.01
Native	−0.01	−0.01*	0.02***	0.00	−0.01	−0.01*	0.02***	0.00
Province_Groningen	0.00	0.02	−0.01	−0.01	0.01	0.02	−0.01	−0.01
Province_Friesland	0.01	0.03***	−0.02	−0.02**	0.01	0.03***	−0.02	−0.02**
Province_Drenthe	0.00	0.03**	0.00	−0.03***	0.00	0.03**	0.00	−0.03***
Province_Overijssel	0.00	0.00	0.01	−0.02***	0.00	0.00	0.01	−0.02***
Province_Gelderland	−0.01	0.01*	0.01	−0.02***	−0.01	0.01*	0.01	−0.02***
Province_Utrecht	−0.03***	0.00	0.03***	0.00	−0.03***	0.00	0.03***	0.00
Province_Z-Holland	−0.02**	0.02**	0.00	0.00	−0.02**	0.02**	0.00	0.00
Province_Zeeland	−0.02	0.04***	0.01	−0.03***	−0.02	0.04***	0.01	−0.03***
Province_N-Brabant	−0.02**	0.04***	−0.02*	0.00	−0.02**	0.04***	−0.02*	0.00
Province_Limburg	0.02*	0.06***	−0.05***	−0.03***	0.02*	0.06***	−0.05***	−0.03***
Province_Flevoland	−0.05***	0.05***	0.03*	−0.02**	−0.05***	0.05***	0.03*	−0.02**
Trend	−0.02***	−0.03***	−0.02***	0.07***	−0.02***	−0.04***	−0.02***	0.07***
Pseudo R2	0.07				0.07
Log-pseudolikelihood	−76,051.9				−76,049.2
Wald χ2	6334.2***				6328.6***

Note: The dependent variable Paymentpreference can take four values: 1 for respondents who prefer to use cash, 2 for respondents who prefer the debit card in a traditional way, 3 for respondents who prefer to pay contactless by debit card and 4 for respondents who prefer to pay contactless with their smartphone. The number of observations is 67,113 . Data period: January 1 2019 – December 18 2021. Standard errors are clustered at the respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Regarding the control variables, most findings are in line with expectations. The likelihood that someone prefers to use the debit card in the traditional way (with PIN code) increases with age, whereas the likelihood of preferring to pay contactless decreases with age. Cash usage is most popular amongst people above 55. The likelihood of preferring cash and of preferring the debit card in a traditional way decreases with education and income, whereas the likelihood of preferring contactless payment methods increases with education and income. People with a partner are less likely to prefer to use cash than people without a partner. Natives are 2 percentage points more likely to prefer to pay contactless by debit card and 1 percentage point less likely to prefer to use the debit card in a traditional manner than non-natives. There are regional differences in payment preferences. For example, people living in Noord-Brabant or Limburg are less fond of paying contactless by debit card than inhabitants of the province Noord-Holland, whereas people living in Utrecht or Flevoland are frontrunners with respect to the preference for this payment instrument.

Additional regressions indicate that the changes in payment preferences stem from the measures taken by the government, banks and retailers and are not caused directly by the fear of getting infected as the likelihood of preferring contactless payments does not relate significantly to the severity of the pandemic, which is proxied by the number of new infections by province. We run regressions with the inclusion of COVID_19 and COVID_19_June_2020. Payment preferences are not significantly related to these variables. Moreover, the results are very similar to those of our baseline regression. Again, we find that the pandemic has resulted in a shift in payment preferences (Table 3, part 2). The likelihood of preferring to pay contactless by debit card increased substantially.

Last, we find that the impact of the pandemic and the measures taken depends on the age of consumers. This is the outcome of an additional set of regressions in which we include interactions between the age dummies and lockdown dummies and between the income dummies and lockdown dummies. The results are in Table B.1 of Appendix B. For example, the shift towards a stronger preference for contactless payments is most pronounced for people aged 65 and above. The effect of the start of the first lockdown on the likelihood to prefer to pay contactless by debit card is 4 percentage points higher for people aged 65 and above than for people below 25 (the reference group). The impact of the lockdown on payment preferences does not depend on income.

Table B.1

COVID-19 and payment preferences: with the number of new infections and interaction terms,average marginal effects based on multinomial logit regressions.

	(1)	(2)	(3)	(4)
	Preference cash	Preference debit card: traditional	Preference debit card: contactless	Preference mobile phone
Lockdown_1_start	−0.03	−0.03*	0.07***	−0.01
Lockdown_1_end	0.01	−0.01	0.01	−0.01
Lockdown_2_start	0.03	−0.06***	0.06***	−0.03***
Lockdown_2_end	0.00	0.00	0.02	−0.02**
COVID_19	0.00	0.00	0.00	0.00
COVID_19_June_2020	0.00	0.00	0.00	0.00
Male	−0.01***	0.00	−0.02***	0.04***
Between_25_34	−0.04***	0.05***	0.00	−0.01
× Lockdown_1_start	0.04	−0.03	−0.01	0.01
× Lockdown_1_end	0.01	−0.02	0.02	−0.02
× Lockdown_2_start	−0.03	0.06*	−0.04	0.01
× Lockdown_2_end	0.01	0.04	−0.04	−0.01
Between_35_44	0.03***	0.07***	−0.08***	−0.02**
× Lockdown_1_start	0.03	0.00	−0.04	0.00
× Lockdown_1_end	−0.03	−0.01	0.03	0.01
× Lockdown_2_start	0.00	−0.01	0.01	0.00
× Lockdown_2_end	0.00	0.01	−0.01	0.00
Between_45_54	0.07***	0.08***	−0.11***	−0.03***
× Lockdown_1_start	0.03	0.00	−0.04*	0.02
× Lockdown_1_end	−0.01	−0.01	0.02	0.00
× Lockdown_2_start	−0.02	0.02	0.01	−0.01
× Lockdown_2_end	0.01	−0.02	0.01	0.00
Between_55_64	0.10***	0.12***	−0.17***	−0.05***
× Lockdown_1_start	0.03*	−0.05**	0.01	0.01
× Lockdown_1_end	−0.02	0.00	0.03	−0.01
× Lockdown_2_start	−0.03	0.01	0.01	0.01
× Lockdown_2_end	0.01	0.01	−0.01	−0.01
65_and_more	0.12***	0.17***	−0.19***	−0.09***
× Lockdown_1_start	−0.02	−0.05**	0.04*	0.02
× Lockdown_1_end	−0.02	0.00	0.02	0.00
× Lockdown_2_start	−0.03	0.02	0.02	−0.02
× Lockdown_2_end	0.01	−0.02	0.01	0.00
Education_low	0.05***	0.02***	−0.04***	−0.03***
Education_high	−0.05***	−0.02***	0.06***	0.02***
Income_low	0.08***	−0.01*	−0.07***	0.01
× Lockdown_1_start	0.00	0.00	0.01	−0.02
× Lockdown_1_end	0.01	0.02	−0.02	−0.01
× Lockdown_2_start	0.01	0.00	−0.01	0.00
× Lockdown_2_end	0.00	0.00	0.00	0.00
Income_high	−0.08***	−0.02***	0.08***	0.03***
× Lockdown_1_start	−0.01	0.00	0.00	0.01
× Lockdown_1_end	0.02	0.00	−0.01	−0.01
× Lockdown_2_start	0.00	0.00	−0.01	0.01
× Lockdown_2_end	0.00	0.00	0.00	0.01
Income_unknown	0.04***	0.01	−0.04***	−0.01
× Lockdown_1_start	−0.02	−0.01	0.02	0.01
× Lockdown_1_end	0.00	0.00	0.00	0.00
× Lockdown_2_start	−0.01	0.03	−0.01	−0.01
× Lockdown_2_end	0.01	−0.01	−0.01	0.02*
Partner	−0.03***	0.01**	0.02***	0.00
Children	0.01	0.00	−0.01	0.01*
Native	−0.01	−0.01**	0.02***	0.00
Province_Groningen	0.01	0.02	−0.01	−0.01
Province_Friesland	0.01	0.03***	−0.02	−0.02**
Province_Drenthe	0.00	0.03**	0.00	−0.03***
Province_Overijssel	0.00	0.00	0.01	−0.02***
Province_Gelderland	−0.01	0.01*	0.01	−0.02***
Province_Utrecht	−0.03***	0.00	0.03***	0.00
Province_Zuid-Holland	−0.02**	0.02**	0.00	0.00
Province_Zeeland	−0.02	0.04***	0.01	−0.03***
Province_Noord-Brabant	−0.02**	0.04***	−0.02*	0.00
Province_Limburg	0.02*	0.06***	−0.05***	−0.03***
Province_Flevoland	−0.05***	0.05***	0.03*	−0.02**
Trend	−0.02***	−0.04***	−0.02***	0.07***
Pseudo R2	0.07
Log-pseudolikelihood	−75,937.4
Wald χ2	6673.5***

Note: The dependent variable is Paymentpreference. The number of observations is 67,113. Data period: January 1 2019 – December 18 2021. Standard errors are clustered at the respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

The findings are robust

For both the regressions on debit card usage and on payment preferences we have conducted robustness checks.

Check 1: payments made with the desired payment instrument

As a first robustness check for debit card usage we re-estimated Eqs. (4) and (5), only using the payment transactions that were paid with the payment instrument that respondents wanted to use and compare the outcomes with those reported in Table 1, columns 2 and 3. This allows us to test whether our initial findings on the overall impact of the lockdown, the spread of the COVID-19 virus and accompanying measures were mainly influenced by altered payment acceptance policies by retailers or not. For each payment respondents were asked to indicate whether they could pay with the payment instrument they wanted to use. Before the lockdown respondents could not pay with the desired payment instrument in less than 2% of the payments at the POS. During the first weeks of the first lockdown this share increased, reaching a peak of 5% in the first half of April 2020, but after a few weeks the share declined and returned to its pre-lockdown level of around 2%. Instead of using 125,651 payment transaction for Eq. (4), we use 122,616 payment transactions made at the POS. For Eq. (5) we use 120,513 payments instead of 123,452 payments (Table C.1 in appendix C).

Table C.1

COVID-19 and payment behaviour: paid with the planned payment instrument, paid in the supermarket, average marginal effects based on logit regressions.

	(1) All transactions	(2) All transactions	(3) Paid as planned	(4) Paid as planned	(5) Supermarket	(6) Supermarket
	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card
Lockdown_1_start	0.12***	0.11***	0.13***	0.11***	0.11***	0.10***
Lockdown_1_end	−0.05***	−0.04***	−0.05***	−0.03***	−0.04***	−0.03***
Lockdown_2_start	−0.02**	−0.02*	−0.02***	−0.02**	−0.03**	−0.02*
Lockdown_2_end	−0.02**	−0.02**	−0.02**	−0.02**	−0.02**	−0.02**
COVID_19		0.01**		0.01**		0.01
COVID_19_June_2020		−0.01**		−0.01**		−0.01
Male	0.01**	0.01*	0.01**	0.01**	0.02***	0.02**
Between_25_34	−0.03***	−0.03***	−0.03***	−0.03***	−0.04***	−0.03***
Between_35_44	−0.05***	−0.05***	−0.05***	−0.05***	−0.07***	−0.07***
Between_45_54	−0.09***	−0.09***	−0.09***	−0.09***	−0.11***	−0.11***
Between_55_64	−0.13***	−0.13***	−0.13***	−0.13***	−0.14***	−0.14***
65_and_more	−0.16***	−0.16***	−0.16***	−0.16***	−0.16***	−0.16***
Education_low	−0.04***	−0.04***	−0.04***	−0.04***	−0.04***	−0.04***
Education_high	0.05***	0.05***	0.05***	0.05***	0.05***	0.05***
Income_low	−0.05***	−0.05***	−0.05***	−0.05***	−0.06***	−0.05***
Income_high	0.05***	0.06***	0.05***	0.06***	0.07***	0.07***
Income_unknown	- 0.01**	- 0.01**	- 0.01**	- 0.01**	- 0.01*	- 0.01**
Partner	0.02***	0.02***	0.02***	0.02***	0.02***	0.02***
Children	−0.01*	−0.01*	−0.01*	−0.01	−0.00	−0.01**
Native	0.01	0.01	0.01	0.01	0.01	0.01
Province_Groningen	0.02	0.01	0.02*	0.01	0.02	0.02
Province_Friesland	0.02*	0.01	0.02*	0.01	0.02*	0.02
Province_Drenthe	−0.00	−0.01	−0.00	−0.01	0.01	0.00
Province_Overijssel	−0.01	−0.02**	−0.01	−0.02**	−0.00	−0.01
Province_Gelderland	−0.00	−0.01	−0.00	−0.01	0.01	0.00
Province_Utrecht	0.03***	0.02**	0.03***	0.02**	0.03***	0.03**
Province_Zuid-Holland	0.01	0.00	0.01	0.00	0.01*	0.01
Province_Zeeland	−0.02	−0.02*	−0.01	−0.02*	−0.01	−0.01
Province_Noord-Brabant	−0.02**	−0.02***	−0.02**	−0.02***	−0.00*	−0.00
Province_Limburg	−0.08***	−0.08***	−0.08***	−0.09***	−0.06***	−0.06***
Province_Flevoland	0.04***	0.03***	0.04***	0.03***	0.03**	0.03*
Amount_EUR_5_10	0.08***	0.08***	0.08***	0.08***	0.05***	0.05***
Amount_EUR_10_20	0.13***	0.13***	0.13***	0.13***	0.08***	0.08***
Amount_EUR_20_and_more	0.19***	0.20***	0.19***	0.20***	0.13***	0.14***
Retailstores_food	−0.07***	−0.07***	−0.07***	−0.07***
Retailstores_non-food	−0.02***	−0.02***	−0.02***	−0.02***
Petrolstations	0.04***	0.04***	0.04***	0.04***
Vendingmachines	0.06***	0.06***	0.06***	0.07***
Streetvending	−0.31***	−0.31***	−0.30***	−0.30***
Cafes_and_restaurants	−0.08***	−0.09***	−0.08***	−0.08***
Recreation_and_culture	−0.18***	−0.18***	−0.18***	−0.18***
Transport	−0.08***	−0.08***	−0.07***	−0.07***
Services	−0.29***	−0.29***	−0.27***	−0.28***
Monday	0.02***	0.02***	0.02***	0.02***	0.00	0.00
Tuesday	0.03***	0.03***	0.03***	0.03***	0.00	0.00
Wednesday	0.03***	0.03***	0.03***	0.03***	0.01	0.01
Thursday	0.03***	0.03***	0.03***	0.03***	0.00	0.00
Friday	0.04***	0.04***	0.04***	0.04***	0.02*	0.02*
Saturday	0.03***	0.03***	0.03***	0.03***	0.01*	0.01*
Log_average_transaction	0.03***	0.03***	0.03***	0.03***	0.04***	0.04***
Log_number_of_transactions	0.01*	0.01**	0.01*	0.01**	−0.00	−0.00
Trend	0.03***	0.03***	0.03***	0.03***	0.02***	0.02***
Pseudo R2	0.130	0.131	0.131	0.133	0.090	0.090
Log-pseudolikelihood	−67,107.1	−65,509.0	−65,620.6	−64,088.5	−26,558.7	−25,919.0
Wald χ2	9173.3***	9140.9***	8999.9***	8978.5***	2981.23***	2966.7***
Number of observations	125,651	122,616	123,452	120,513	51,755	50,458

Note: The dependent variable is Debitcard. This variable equals 1 for a debit payment and 0 for a cash payment. Debit payments include both traditional debit card payments, contactless payments with the debit card and contactless debit payments via an app on people's smartphone. Standard errors are clustered at respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

For Eq. (4) we find that since the start of the first lockdown the likelihood to pay with the debit card has increased by 12 percentage points, that from July 1 2020 onwards the likelihood of a debit card payment has dropped by 5 percentage points, and that the start and end of the second lockdown resulted in a drop in the likelihood of debit card usage of 2 percentage points. These results show that the impact of the start of the first lockdown is 1 percentage point lower than in the baseline model and that the other three lockdown effects are equal. Also, for the re-estimation of Eq. (5) we only find small changes. The initial impact of the first lockdown and the influence of the COVID-19 variables remain nearly unchanged. We observe that only the estimated effect of Lockdown_1_end on the likelihood to pay with the debit card slightly diminishes from −4 percentage points to −3 percentage points. Grosso modo, the majority of the estimated effects of control variables are unaltered. For Eq. (4) the estimated effects of four controls changed and for Eq. (5) the estimated effects of five controls changed, mostly leading to a 1 percentage point change in estimated average marginal effects, and for one control in Eq. (5) to a change of 2 percentage points.

Check two: supermarket data

As a second robustness check for debit card usage, we re-estimated Eqs. (4) and (5), only using the transactions made at supermarkets. We compare these outcomes with those reported in Table 1, columns 2 and 3. Supermarkets have never been closed during the pandemic, as the supermarket branch was one of the essential retail branches. Furthermore, supermarkets continued to accept both cash and debit card payments during the pandemic. Focussing on supermarket data therefore allows us to test whether our baseline findings are mainly due to composition effects or not. In the pre-pandemic year 2019, supermarkets accounted for 37% of all POS transactions, with an average transaction value of EUR 20.51. Cash accounted for 30% of the total number of supermarket transactions and the debit card for 70%. For all POS payments together the share of cash was 32% in 2019 and the average transaction value was EUR 22.37, indicating that the average transaction value is slightly lower in supermarkets than in all other branches together and that the likelihood of debit card usage is slightly higher, but that overall people's payment behaviour in supermarkets reflects their payment behaviour in all POS branches together rather well. Instead of using 125,651 payment transactions for Eq. (4), we use 51,755 supermarket transactions. For Eq. (5) we use 50,458 payment transactions instead of 123,452 payment transactions (Table C.1 columns 5 and 6 in appendix C).

Overall, our results for the supermarkets are in line with those covering all POS transactions in all branches that we included in the baseline regressions. For Eq. (4) we find that since the start of the first lockdown the likelihood to pay with the debit card in the supermarket has increased by 11 percentage points, that from July 1 2020 onwards the likelihood of debit card usage has dropped by 4 percentage points, that the start of the second lockdown was marked by a drop in the likelihood of debit card usage of 3 percentage points, and that the end of the second lockdown resulted in a drop in the likelihood of debit card usage of 2 percentage points. These results show that the impact of the start and the end of the first lockdown was 1 percentage point lower than for all POS transactions together, leading to the same net impact of the first lockdown, and that the impact of the start of the second lockdown was 1 percentage point larger than for POS transactions. The effect of the end of the second lockdown is the same.

Also, for the re-estimation of Eq. (5) we find results that are in line with those for all POS transactions. The initial impact of the start and the end of the first lockdown are 1 percentage point lower, leading to the same net result. The impact of the start and end of the second lockdown are equal to the estimated effects for all POS transactions together. In addition, the estimated effect of the influence of the COVID-19 variables remains unaltered, although the significance is reduced.

Most of the estimated effects of the other control variables are in line with those for all POS transactions together, although there are a few differences worth mentioning. These differences hold for both Eqs. (4) and (5). First, the day of the week effects are somewhat less pronounced than for all POS payments. Only on Fridays and Saturdays the likelihood of debit card usage is (slightly) higher than on Sundays, whereas for all POS transactions together the likelihood of debit card usage is higher on all days of the week. Regional differences are less pronounced as well, suggesting that these mainly manifest themselves in other branches. Furthermore, the estimated impact of the transaction value on the likelihood of debit card usage is smaller, although also in supermarkets there is a clear positive relation between transaction value and debit card usage.

Check three: estimations by transaction amount

As a third robustness check for debit card usage we re-estimated Eqs. (4) and (5) separately by transaction value. We distinguish four categories: ≤EUR 5.00, EUR 5.01 – EUR 10.00, EUR 10.01 – EUR 20.00 and >EUR 20.00 (Table C.2 in Appendix C). We compare the outcomes with those reported in Table 1, columns 2 and 3, and with the test results in Table 2, row ‘all transactions’. This allows us to test whether our findings on the overall impact of the lockdown, the spread of the COVID-19 and accompanying measures on consumers’ payment behaviour at the POS differs by transaction amount.

Table C.2

COVID-19 and payment behaviour: regressions per amount,average marginal effects based on logit regressions.

	(1) All transactions	(2) All transactions	(3) ≤ EUR 5.00	(4) ≤ EUR 5.00	(5) EUR 5.01 – 10.00	(6) EUR 5.01 – 10.00	(7) EUR 10.01 – 20.00	(8) EUR 10.01 – 20.00	(9) > EUR 20.00	(10) > EUR 20.00
	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card	Debit card
Lockdown_1_start	0.12***	0.11***	0.16***	0.12***	0.16***	0.16***	0.14***	0.12***	0.07***	0.07***
Lockdown_1_end	−0.05***	−0.04***	−0.06***	−0.02***	−0.08***	−0.07***	−0.07***	−0.05***	−0.02**	−0.02
Lockdown_2_start	−0.02**	−0.02*	−0.04***	−0.03***	−0.02	−0.01	−0.02*	−0.01	−0.00	−0.00
Lockdown_2_end	−0.02**	−0.02**	−0.03**	−0.03**	−0.03	−0.03	−0.02*	−0.03*	−0.02	−0.01
COVID_19		0.01**		0.02***		0.00		0.01		0.00
COVID_19_June_2020		−0.01**		−0.02***		−0.00		−0.01		−0.00
Male	0.01**	0.01*	0.03**	0.03**	0.02**	0.02**	0.00	0.01	−0.01**	−0.01**
Between_25_34	−0.03***	−0.03***	−0.00	−0.00	−0.02	−0.02	−0.04**	−0.04**	−0.01	−0.02*
Between_35_44	−0.05***	−0.05***	−0.04***	−0.04***	−0.05***	−0.04***	−0.06***	−0.06***	−0.02	−0.02**
Between_45_54	−0.09***	−0.09***	−0.10***	−0.10***	−0.11***	−0.11***	−0.10***	−0.10***	−0.03***	−0.04***
Between_55_64	−0.13***	−0.13***	−0.17***	−0.17***	−0.15***	−0.15***	−0.13***	−0.14***	−0.04***	−0.05***
65_and_more	−0.16***	−0.16***	−0.25***	−0.25***	−0.19***	−0.19***	−0.15***	−0.15***	−0.05***	−0.05***
Education_low	−0.04***	−0.04***	−0.05***	−0.05***	−0.04***	−0.04***	−0.04***	−0.03***	−0.03***	−0.03***
Education_high	0.05***	0.05***	0.07***	0.07***	0.05***	0.05***	0.05***	0.05***	0.03***	0.03***
Income_low	−0.05***	−0.05***	−0.04**	−0.04**	−0.06**	−0.06**	−0.07***	−0.07***	−0.05***	−0.05***
Income_high	0.05***	0.06***	0.05***	0.05***	0.06***	0.06***	0.05***	0.05***	0.05***	0.05***
Income_unknown	0.01**	0.01**	−0.02**	−0.01	−0.02**	−0.02*	−0.01	−0.01*	−0.00*	−0.01
Partner	0.02***	0.02***	0.03***	0.02***	0.02**	0.02**	0.04***	0.03***	0.02***	0.02***
Children	−0.01*	−0.01*	−0.02***	−0.02***	−0.01	−0.00	−0.00	−0.00	−0.00	−0.00
Native	0.01	0.01	0.00	0.00	0.02***	0.02***	0.00	0.00	0.01	0.01
Province_Groningen	0.02	0.01	0.00	−0.00	0.01	0.01	0.05***	0.04***	0.01	0.00
Province_Friesland	0.02*	0.01	0.02	0.02	0.01	0.00	0.02	0.02	0.02	0.01
Province_Drenthe	−0.00	−0.01	0.01	0.00	−0.03	−0.04*	0.00	−0.00	0.01	0.00
Province_Overijssel	−0.01	−0.02**	−0.02	−0.02*	−0.00	−0.01	−0.01	−0.01	−0.02**	−0.03**
Province_Gelderland	−0.00	−0.01	−0.02	−0.02*	0.00	−0.00	0.01	0.00	−0.00	−0.01
Province_Utrecht	0.03***	0.02**	0.03*	0.02	0.02	0.02	0.03**	0.03**	0.02**	0.02
Province_Zuid-Holland	0.01	0.00	0.02*	0.01	0.01	0.00	0.02*	0.01	0.00	−0.00
Province_Zeeland	−0.02	−0.02*	−0.08***	−0.08***	0.00	−0.00	−0.01	−0.01	0.02	0.01
Province_Noord-Brabant	−0.02**	−0.02***	−0.02**	−0.03**	−0.02	−0.02*	−0.01*	−0.01	−0.02*	−0.02**
Province_Limburg	−0.08***	−0.08***	−0.12***	−0.12***	−0.12***	−0.13***	−0.05***	−0.05***	−0.05***	−0.05***
Province_Flevoland	0.04***	0.03***	0.05***	0.04**	0.06***	0.05**	0.07***	0.06***	−0.00	−0.00
Amount_EUR_5_10	0.08***	0.08***
Amount_EUR_10_20	0.13***	0.13***
Amount_EUR_20_and_more	0.19***	0.20***
Retailstores_food	−0.07***	−0.07***	−0.12***	−0.12***	−0.08***	−0.08***	−0.04***	−0.05***	−0.05***	−0.05***
Retailstores_non-food	−0.02***	−0.02***	−0.05**	−0.05**	−0.03**	−0.03**	−0.01	−0.01	−0.01**	−0.01**
Petrolstations	0.04***	0.04***	−0.17***	−0.17***	−0.05***	−0.04***	0.01	0.02	0.09***	0.09***
Vendingmachines	0.06***	0.06***	0.01	0.01	0.22**	0.22**	0.24***	0.25***	0.010***	0.010***
Streetvending	−0.31***	−0.31***	−0.45***	−0.45***	−0.32***	−0.32***	−0.24***	−0.25***	−022***	−022***
Cafes_and_restaurants	−0.08***	−0.09***	−0.07***	−0.07***	−0.11***	−0.11***	−0.10***	−0.10***	−0.09***	−0.09***
Recreation_and_culture	−0.18***	−0.18***	−0.28***	−0.28***	−0.17***	−0.17***	−0.15***	−0.15***	−0.13***	−0.13***
Transport	−0.08***	−0.08***	−0.07***	−0.07***	−0.08***	−0.09***	−0.12***	−0.13***	−0.11***	−0.11***
Services	−0.29***	−0.29***	−0.31***	−0.31***	−0.28***	−0.28***	−0.27***	−0.27***	−0.22***	−0.22***
Monday	0.02***	0.02***	0.06***	0.06***	0.03*	0.03*	0.01	0.01	−0.01	−0.01
Tuesday	0.03***	0.03***	0.06***	0.06***	0.03**	0.03**	0.02*	0.02**	0.01*	0.01
Wednesday	0.03***	0.03***	0.07***	0.07***	0.02	0.02	0.01	0.02	0.02*	0.02**
Thursday	0.03***	0.03***	0.05***	0.05***	0.02*	0.02*	0.02	0.02	0.01	0.01
Friday	0.04***	0.04***	0.05***	0.05***	0.04***	0.04***	0.02	0.02	0.03***	0.03***
Saturday	0.03***	0.03***	0.02	0.02	0.03**	0.03**	0.02***	0.03**	0.03***	0.03***
Log_average_transaction	0.03***	0.03***	0.04***	0.04***	0.02***	0.02***	0.03**	0.03***	0.03***	0.03***
Log_number_of_trx	0.01*	0.01**	0.00	0.00	0.01	0.01	0.01	0.01*	0.01	0.01*
Trend	0.03***	0.03***	0.05***	0.05***	0.04***	0.04***	0.02***	0.02***	0.01	0.01
Pseudo R2	0.130	0.131	0.126	0.127	0.095	0.097	0.081	0.082	0.080	0.082
Log-pseudolikelihood	−67,107.1	−65,509.0	−19,140.3	−18,760.4	−14,357.0	−14,015.5	−14,901.9	−14,535.6	−18,026.6	−17,529.5
Wald χ2	9173.3***	9140.9***	2833.9***	2842.0***	1851.2***	1833.3***	1780.4***	1769.5***	2240.2***	2246.2***
Impact lockdown: Wald χ2
(A) Lockdown_1_start =0	179.99***	98.62***	87.9***	34.7***	85.23***	56.57***	79.14***	42.21***	36.42***	23.60***
(B) Lockdown_1_start + Lockdown_1_end = 0	73.54***	77.23***	46.8***	47.6***	26.99***	27.55***	29.15**	30.47**	19.75**	21.89***
(C) Lockdown_1_start + Lockdown_1_end + Lockdown_2_start + Lockdown_2_end = 0	7.71***	10.84***	4.17**	4.85**	3.28*	4.30**	2.46	3.70*	3.67*	5.53**
Number of observations	125,651	122,616	31,729	31,132	24,601	24,006	27,688	26,988	41,614	40,475

Note: The dependent variable is Debitcard. This variable equals 1 for a debit payment and 0 for a cash payment. Debit payments include both traditional debit card payments, contactless payments with the debit card and contactless debit payments via an app on people's smartphone. Standard errors are adjusted for clusters at respondent level. Wald test (A) examines whether the likelihood of debit card usage has remained the same after the beginning of the first lockdown (apart from the general trend). Wald test (B) examines whether the likelihood of debit card usage has remained the same after the end of the first lockdown (apart from the general trend), and Wald test (C) examines whether the likelihood of debit card usage has remained the same at the end of the second lockdown (apart from the general trend). ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Our results indicate that the initial effect of both the start and the end of the first lockdown on the likelihood of debit card usage is significant four all four categories. The effect of the start of the first lockdown declines with the transaction value and ranges between +7 percentage points for >EUR 20.00 payments to +16 percentage points for ≤EUR 5.00 payments. The effect of the end of the first lockdown ranges between −8 percentage points for payments between EUR 5.01 and EUR 10.00 to −2 percentage points for payments with a minimum value of EUR 20.01. The estimated effects for the first lockdown are smaller in the specification including COVID-19 infections, but they are still significant for all transaction amounts. Wald tests show that the cumulative effect of the start and the end of the first lockdown on the likelihood of debit card usage is significantly different from zero for all transaction amounts and for both the specification including and excluding COVID-19 transactions. The impact of the start and end of the second lockdown only had a significant effect on the likelihood of debit card usage for payments up to EUR 5.00 and payments with a value between EUR 10.01 and EUR 20.00. Wald tests show that the cumulative effect of the start and end of the two lockdowns on the likelihood of debit card usage is significantly different from zero for 7 out of 8 regressions. For payments between EUR 10.01 and EUR 20.00 the test could not be rejected in the specification excluding COVID-19 infections (p = 0.12). Overall, the results suggest that the pandemic shifted the likelihood of debit card usage upwards for all transaction amounts and that this effect lasted after the end of the second lockdown at the end of June 2021. In addition, the overall impact of the pandemic was largest for the smallest transaction amounts, which prior to the pandemic were also more likely to be paid in cash.

Check 4: alternative start and end dates first lockdown

Last, we checked to what degree our findings are sensitive to the construction of Lockdown_1_start and Lockdown_1_end. In the baseline model Lockdown_1_start takes the value 1 as of March 16 2020. As alternatives we use March_10_2020 (the day of the CBL announcement to consumers to pay contactless), March_19_2020 (the day that the cumulative transaction limit of contactless payments was increased to EUR 100), and March_24_2020 (the day the transaction limit of contactless payments was increased to EUR 50). In addition to these three alternatives, we run regressions with time dummy June_15_2020 instead of Lockdown_1_end. By then travelling became easier.

Our results are robust to the use of these alternative start and ending dates. In all cases, we find that the pandemic resulted in a significant increase in the likelihood that consumers use the debit card, which are in line with the effects in the baseline model (see Table C.3 in Appendix C). The effect of using alternative dates for the start of the first lockdown on the likelihood of debit card usage ranges between +12 and +13 percentage points and the effect of using the alternative end date June 15 amounts −5 percentage points and −6 percentage points. Usage of alternative start and end dates in the first lockdown did not lead to changes in the estimated effects of Lockdown_2_start and Lockdown_2_end. This robustness test also confirms that payment preferences have shifted: the likelihood that people prefer to pay contactless has increased and the likelihood that they prefer to use the debit card with PIN code has declined (Table C.4 in Appendix C). The cumulative effect on the likelihood that people prefer to pay contactless by debit card ranges between 16 and 17 percentage points.

Table C.3

COVID-19 and payment behaviour: alternative timing of event dummies,average marginal effects based on binomial logit regressions.

		(1)	(2)
		Debit card	Debit card
Baseline model	Lockdown_1_start	0.12***	0.11***
	Lockdown_1_end	−0.05***	−0.04***
	Lockdown_2_start	−0.02**	−0.02*
	Lockdown_2_end	−0.02**	−0.02**
	COVID_19		0.01**
	COVID_19_June_2020		−0.01**
Alternative 1	March_10_2020	0.12***	0.11***
	Lockdown_1_end	−0.05***	−0.03***
	Lockdown_2_start	−0.02**	−0.02*
	Lockdown_2_end	−0.02**	−0.02**
	COVID_19		0.01**
	COVID_19_June_2020		−0.01**
Alternative 2	March_19_2020	0.13***	0.12***
	Lockdown_1_end	−0.06***	−0.04***
	Lockdown_2_start	−0.02**	−0.02*
	Lockdown_2_end	−0.02**	−0.02**
	COVID_19		0.01**
	COVID_19_June_2020		−0.01**
Alternative 3	March_24_2020	0.13***	0.11***
	Lockdown_1_end	−0.06***	−0.04***
	Lockdown_2_start	−0.02**	−0.02*
	Lockdown_2_end	−0.02**	−0.02**
	COVID_19		0.01***
	COVID_19_June_2020		−0.01***
Alternative 4	Lockdown_1_start	0.13***	0.12***
	June_15_2020	−0.06***	−0.05***
	Lockdown_2_start	−0.02**	−0.02*
	Lockdown_2_end	−0.02**	−0.02**
	COVID_19		0.00
	COVID_19_June_2020		−0.00

Note: The dependent variable is Debitcard. The same set of control variables is included as in Table 1 model 2, respectively model 3. The number of observations is 125,651, respectively 122,616. Data period: January 1 2018 - December 18 2021. Standard errors are clustered at respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Table C.4

COVID-19 and payment preferences: alternative timing of event dummies,average marginal effects based on multinomial logit regressions.

		(1)				(2)
		(a)	(b)	(c)	(d)	(a)	(b)	(c)	(d)
		Preference cash	Preference debit card: traditional	Preference debit card: contactless	Preference mobile phone	Preference cash	Preference debit card: traditional	Preference debit card: contactless	Preference mobile phone
Baseline model	Lockdown_1_start	−0.02***	−0.06***	0.07***	0.01**	−0.02**	−0.06***	0.07***	0.01
	Lockdown_1_end	0.00	−0.01**	0.03***	−0.02***	0.00	−0.01	0.03***	−0.01***
	Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
	Lockdown_2_end	0.01*	−0.01	0.01**	−0.02***	0.01*	−0.01	0.02**	−0.02***
	COVID_19					0.00	0.00	0.00	0.00
	COVID_19_June_2020					0.00	0.00	0.00	0.00
Alternative 1	March_10_2020	−0.02***	−0.05***	0.06***	0.01**	−0.02***	−0.05***	0.07***	0.01
	Lockdown_1_end	0.00	−0.02**	0.04***	−0.02***	0.00	−0.02**	0.03***	−0.02***
	Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
	Lockdown_2_end	0.01*	−0.01	0.01**	−0.02***	0.01*	−0.01	0.02**	−0.02***
	COVID_19					0.00	0.00	0.00	0.00
	COVID_19_June_2020					0.00	0.00	0.00	0.00
Alternative 2	March_19_2020	−0.02***	−0.05***	0.06***	0.01**	−0.02***	−0.06***	0.07***	0.01*
	Lockdown_1_end	0.00	−0.01**	0.03***	−0.02***	0.00	−0.01	0.03***	−0.02***
	Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
	Lockdown_2_end	0.01*	−0.01	0.01*	−0.02***	0.01*	−0.01	0.01**	−0.02***
	COVID_19					0.00	0.00	0.00	0.00
	COVID_19_June_2020					0.00	0.00	0.00	0.00
Alternative 3	March_24_2020	−0.02***	−0.05***	0.06***	0.01**	−0.03***	−0.05***	0.07***	0.01*
	Lockdown_1_end	0.00	−0.01*	0.03***	−0.02***	0.00	−0.01*	0.03***	−0.02***
	Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
	Lockdown_2_end	0.01*	−0.01	0.01*	−0.02***	0.01*	−0.01	0.01*	−0.02***
	COVID_19					0.00	0.00	0.00	0.00
	COVID_19_June_2020					0.00	0.00	0.00	0.00
Alternative 4	Lockdown_1_start	−0.02***	−0.05***	0.06***	0.01**	−0.02**	−0.06***	0.07***	0.00
	June_15_2020	0.00	−0.02***	0.03***	−0.01***	0.00	−0.02**	0.03***	−0.01
	Lockdown_2_start	0.01	−0.04***	0.06***	−0.03***	0.01	−0.04***	0.06***	−0.03***
	Lockdown_2_end	0.01*	−0.01	0.01*	−0.01***	0.01*	−0.01	0.01**	−0.02***
	COVID_19					0.00	0.00	0.00	0.00
	COVID_19_June_2020					0.00	0.00	0.00	0.00

Note: The dependent variable is Paymentpreference. The same set of control variables is included as in Table 3. The number of observations is 67,113. Data period: January 1 2019 - December 18 2021. Standard errors are clustered at the respondent level. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Conclusion and discussion

We find that COVID-19 and the accompanying containment measures accelerated the already ongoing increase in debit card usage in the Netherlands. A large part of the initial rise in debit card usage appears to be persistent. Initially, the pandemic and the containment measures led to a 12 percentage points increase of the likelihood of debit card usage at the point-of-sale (POS) at the expense of cash usage. The impact of the pandemic on people's payment behaviour appears to be mainly triggered by the measures taken to control the pandemic. Only during the first months of the pandemic, the likelihood to pay with the debit card correlates positively with the number of new COVID-19 infections, but even then the larger part of the change in payment behaviour appears to be triggered by measures and not by increasing infections. In addition, for many people payment behaviour did not return to pre-COVID-19 levels after the end of the first lockdown period in July 2020. The share of cash payments at the POS only partially reversed from its lowest point in April 2020: seven months after the start of the first lockdown debit card usage was still 7 percentage points above its pre-lockdown trend level. In contrast to the first lockdown, the second lockdown which started on December 15 2020 was marked by a small reduction in the likelihood of debit card usage of 2 percentage points, and also the end of the second lockdown period on June 26 2021 showed a small drop in the debit card usage. The cumulative effect of the two lockdowns on the likelihood that people pay with the debit card at the POS ranges between +3 and +4 percentage points. Overall, our findings indicate that the pandemic accelerated the increase in relative debit card usage during 2020, that a large part of the initial change persisted during the second pandemic year, but that the higher share in debit card usage at the POS in 2021 would probably have also occurred in the absence of the pandemic and the accompanying measures as reflected by the positive autonomous trend in debit card usage. Our results are in line with experiences in Australia, Norway and Switzerland where the pandemic has also led to increased usage of card payments at the expense of cash that are expected to last.

The lockdown did not have a homogeneous impact on people's payment behaviour. In some age and income groups debit card usage has hardly changed, for example for people with a low income. For other demographic groups, especially people aged 65 and over and people with a high income, there is evidence of a stronger effect that appears to persist after the end of the second lockdown. The impact of the pandemic also differs by branch. In most branches the pandemic has led to a persisting increase in debit card usage above its trend level, but not in all branches, depending on the initial level of debit card usage and the way the branch has been affected by government measures.

A complete reversal of payment behaviour after the ending of the COVID-19 pandemic seems unlikely. Substantially more people now prefer to pay contactless. The share of people preferring to use their debit card with a PIN code has declined, whereas the share of people preferring cash has also somewhat declined. Compared to before the pandemic, the gap between the share of people preferring to pay with cash and the share of POS transactions actually paid in cash has decreased. Prior research on the Netherlands reported a gap between payment preferences and payment behaviour; cash was used more frequently than one would expect based on stated payment preferences (van der Cruijsen et al., 2017).

Our findings provide a better understanding of how an external health shock and accompanying measures by the government, banks and retailers can shift payment behaviour and payment preferences. Within a few months’ time, a long-lived change in payment behaviour appears to have taken place. However, only time will tell whether the effect of the shock is permanent. Compared to other external shocks, the impact of the pandemic on payment behaviour has been relatively large in magnitude and for some demographic groups and branches long-lasting in duration.

CRediT authorship contribution statement

Nicole Jonker: Writing – original draft, Writing – review & editing. Carin van der Cruijsen: Writing – original draft, Writing – review & editing. Michiel Bijlsma: Writing – original draft, Writing – review & editing. Wilko Bolt: Writing – original draft, Writing – review & editing.

Declarations of Competing Interest

None.

Table A.1

Description of variables in the payment behaviour regressions.

variable	description	mean	sd	min	max	n
Debitcard	Dummy (1=debit payment, 0=else).	0.70	0.46	0	1	125,651
Explanatory variables
Lockdown_1_start	Dummy (1=March 16 2020, and onwards, 0=before March 16 2020).	0.39	0.49	0	1	125,651
Lockdown_1_end	Dummy (1=July 1 2020, and onwards, 0=before July 1 2020).	0.33	0.47	0	1	125,651
Lockdown_2_start	Dummy (1=December 15 2020, and onwards, 0=before December 15 2020).	0.22	0.41	0	1	125,651
Lockdown_2_end	Dummy (1=June 26 2021, and onwards, 0=before June 26 2021).	0.12	0.32	0	1	125,651
COVID_19	Number of new infections per province per 100,000 inhabitants.	9.58	22.04	0	221.63	122.616
COVID_19_June_2020	COVID-19 interacted with a binary dummy June 1, which equals 1 from June 1 2020.	9.43	22.08	0	221.63	122.616
Male	Dummy (1=male, 0=female).	0.47	0.50	0	1	125,651
Between_12_24	Dummy (1=age between 12 and 24, 0=else). Reference category.	0.10	0.31	0	1	125,651
Between_25_34	Dummy (1=age between 25 and 34, 0=else).	0.10	0.30	0	1	125,651
Between_35_44	Dummy (1=age between 35 and 44, 0=else).	0.16	0.36	0	1	125,651
Between_45_54	Dummy (1=age between 45 and 54, 0=else).	0.21	0.41	0	1	125,651
Between_55_64	Dummy (1=age between 55 and 64, 0=else).	0.20	0.40	0	1	125,651
65_and_more	Dummy (1=age 65 and over, 0=else).	0.23	0.42	0	1	125,651
Education_low	Dummy (1= no education/primary school/VMBO/MBO/MAVO/HAVO/ VWO (first 3 years), 0=else).	0.30	0.54	0	1	125,651
Education_middle	Dummy (1=MBO 2, 3, 4/MBO old or HAVO/VWO, 0=else). Reference category.	0.33	0.55	0	1	125,651
Education_high	Dummy (1=HBO/WO bachelor or WO/HBO, 0=else).	0.41	0.57	0	1	125,651
Income_low	Dummy (1=yearly gross household income less than EUR 23,400, 0=else or unknown).	0.16	0.37	0	1	125,651
Income_middle	Dummy (1=yearly gross household income ≥ EUR 23,400 and < EUR 51,300, 0=else or unknown). Reference category.	0.33	0.47	0	1	125,651
Income_high	Dummy (1=yearly gross household income ≥ EUR 51,300, 0=else or unknown).	0.28	0.45	0	1	125,651
Income_unknown	Dummy (1=yearly gross household income is unknown, 0=income is known).	0.23	0.42	0	1	125,651
Partner	Dummy (1=living together or married, 0=else).	0.64	0.48	0	1	125,651
Children	Dummy (1=household with kids living at home, 0=else).	0.31	0.46	0	1	125,651
Native	Dummy (1=native, 0 = non-native).	0.80	0.40	0	1	125,651
Province_Noord-Holland	Dummy (1=living in the province Noord-Holland, 0=else). Reference category.	0.15	0.36	0	1	125,651
Province_Groningen	Dummy (1=living in the province Groningen, 0=else).	0.03	0.18	0	1	125,651
Province_Friesland	Dummy (1=living in the province Friesland, 0=else).	0.04	0.19	0	1	125,651
Province_Drenthe	Dummy (1=living in the province Drenthe, 0=else).	0.03	0.17	0	1	125,651
Province_Overijssel	Dummy (1=living in the province Overijssel, 0=else).	0.06	0.24	0	1	125,651
Province_Gelderland	Dummy (1=living in the province Gelderland, 0=else).	0.10	0.30	0	1	125,651
Province_Utrecht	Dummy (1=living in the province Utrecht, 0=else).	0.07	0.26	0	1	125,651
Province_Zuid-Holland	Dummy (1=living in the province Zuid-Holland, 0=else).	0.22	0.41	0	1	125,651
Province_Zeeland	Dummy (1=living in the province Zeeland, 0=else).	0.03	0.16	0	1	125,651
Province_Noord-Brabant	Dummy (1=living in the province Noord-Brabant, 0=else).	0.15	0.36	0	1	125,651
Province_Limburg	Dummy (1=living in the province Limburg, 0=else).	0.06	0.24	0	1	125,651
Province_Flevoland	Dummy (1=living in the province Flevoland, 0=else).	0.03	0.17	0	1	125,651
Amount_EUR_5_and_less	Dummy (1=amount paid equals EUR 5 or less, 0=else). Reference category.	0.25	0.44	0	1	125,651
Amount_EUR_5_10	Dummy (1=amount paid lies between EUR 5–EUR 10, 0=else).	0.20	0.40	0	1	125,651
Amount_EUR_10_20	Dummy (1=amount paid lies between EUR 10–EUR 20, 0=else).	0.22	0.41	0	1	125,651
Amount_EUR_20_and_more	Dummy (1=amount paid equals EUR 20 or higher, 0=else).	0.33	0.47	0	1	125,651
Supermarkets	Dummy (1=transaction in a supermarket, 0=else). Reference category.	0.41	0.49	0	1	125,651
Retailstores_food	Dummy (1=transaction in a food store, 0=else).	0.09	0.29	0	1	125,651
Retailstores_non-food	Dummy (1=transaction in a non-food store, 0=else).	0.18	0.38	0	1	125,651
Petrolstations	Dummy (1=transaction at a petrol station, 0=else).	0.06	0.24	0	1	125,651
Vendingmachines	Dummy (1=transaction at a vending machine, 0=else).	0.04	0.20	0	1	125,651
Streetvending	Dummy (1=transaction in street vending, 0=else).	0.04	0.19	0	1	125,651
Cafes_and_restaurants	Dummy (1=transaction in a cafe or restaurant, 0=else).	0.12	0.33	0	1	125,651
Recreation_and_culture	Dummy (1=transaction for recreational or cultural purposes, 0=else).	0.02	0.15	0	1	125,651
Transport	Dummy (1=transaction related to transport, 0=else).	0.01	0.09	0	1	125,651
Services	Dummy (1=transaction in the services sector, 0=else).	0.03	0.16	0	1	125,651
Log_average_transaction	Log of the average transaction value in cash or by debit card at the POS at day t.	2.81	0.86	0.01	8.84	125,651
Log_number_of_transactions	Log of the number of cash and debit card transactions at the POS at day t.	0.98	0.68	0	4.26	125,651

Note: This table describes the variables used in the regressions reported in Table 1. The mean, standard deviation (sd), minimum (min), maximum (max), and number of payment transactions used (N) are reported for the sample included in these regressions. Unreported are summary statistics of trend and the dummies to control for the week of the day.

Table A.2

Description of dependent variable in payment preference regressions.

Variable	Description	Mean	Sd	Min	Max	N
Paymentpreference	Variable capturing payment preferences (1=prefers to pay in cash, 2= prefers to pay by debit card with PIN code, 3=prefers to pay contactless by debit card, 4=prefers to pay contactless by mobile phone). To construct these variables, we use the answers to three survey questions. The first one is ‘Under normal circumstances do you prefer paying by debit card or paying cash?’ with answer options ‘preference debit card’, ‘preference cash’, and ‘no preference/I cannot say’. People who answered ‘no preference/I cannot say’ got the follow-up question ‘And if you had to choose between paying by debit card and cash, what would you prefer?’ with possible answers ‘preference debit card’ and ‘preference cash’. Based on these two questions the respondents with a preference for the debit card where asked what they prefer: ‘paying by debit card using a PIN code’, ‘paying contactless with a debit card’, ‘paying contactless with a mobile phone’, ‘no preference’.	2.42	0.91	1	4	67,133

Note: This table describes the dependent variable used in the regressions reported in Table 3. The mean, standard deviation (sd), minimum (min), maximum (max), and number of payment diaries used (N) are reported for the sample included in these regressions.