choosing how to pay: the influence of foreign backgrounds

Journal of Banking & Finance 37 (2013) 989–998

Contents lists available at SciVerse ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier .com/locate / jbf

Choosing how to pay: The influence of foreign backgrounds

Anneke Kosse a,⇑, David-Jan Jansen b

a De Nederlandsche Bank, Cash and Payment Systems Division, P.O. Box 98, 1000 AB Amsterdam, The Netherlandsb De Nederlandsche Bank, Economics and Research Division, P.O. Box 98, 1000 AB Amsterdam, The Netherlands

a r t i c l e i n f o

Article history:Received 12 January 2012Accepted 5 November 2012Available online 20 November 2012

JEL classification:C25D12

Keywords:Consumer paymentsHabitsDebit cardCashMigration

0378-4266/$ - see front matter � 2012 Elsevier B.V. Ahttp://dx.doi.org/10.1016/j.jbankfin.2012.11.005

⇑ Corresponding author. Tel.: +31 20 524 2827; faxE-mail addresses: [email protected] (A. Kosse), d.

a b s t r a c t

Is having a foreign background a relevant factor in choosing between payment instruments in consumerpoint-of-sale transactions after migration? We analyze this question using a unique diary survey inwhich both participants with a Dutch and a foreign background documented their daily purchases. Wepresent several pieces of evidence suggesting that foreign backgrounds still influence the choice betweenpayment instruments after migration to the Netherlands. For instance, we find that first-generationmigrants from a number of countries that can be seen as cash-oriented are more likely to use cash inthe Netherlands. At the same time, second-generation migrants have similar payment habits as individ-uals with a Dutch background. This finding suggests that payment behavior is not passed on betweengenerations, but affected by host country payment habits. Finally, we suggest that, in this context, specialinformation campaigns to increase debit card usage will not have clear net social benefits.

� 2012 Elsevier B.V. All rights reserved.

1. Introduction

In a point-of-sale (POS) transaction, a typical consumer has thechoice between various payment instruments, such as cash, debitcards, cheques or credit cards. The payments literature (see, forexample, Bolt and Chakravorti (2010) for a synopsis) suggests thechoice will depend on various factors, such as transaction charac-teristics (e.g. the amount), location characteristics (e.g. the avail-ability of a POS terminal), and cost structures (e.g. charges forusing cards). In addition, many studies find that consumer charac-teristics are important. The intensity of using various methods ofpayment is usually related to demographic factors, such as age,education, income and gender (Borzekowski et al., 2008; Klee,2008; Stavins, 2001).

The literature has, so far, paid little attention to payment behav-ior of migrants. This paper studies whether the choices betweenpayment instruments made by individuals with a foreign back-ground are in any way different. If so, can we explain these differ-ences? To this end, we conducted an extensive diary survey among2258 residents of the Netherlands with either a Dutch or a foreignbackground. Thus, our paper sheds light on the role of home coun-try payment habits and on possible changes in payment behaviorafter migration. Various respondents in our survey have ties –

ll rights reserved.

: +31 20 524 [email protected] (D. Jansen).

either directly or through their parents – to countries where con-sumers have payment habits that differ from the Dutch situation.Using our unique dataset, we present several pieces of evidencesuggesting that foreign backgrounds still influence the choice be-tween payment instruments after migration. First, we find thatfirst-generation migrations with a non-western background aremore likely to use cash in Dutch POS transactions. Second, respon-dents from three countries that, compared to the Netherlands, canbe seen as cash oriented (Germany, Morocco, and Turkey) are up to13% points more likely to use cash in POS transactions in the Neth-erlands than respondents with a Dutch background. Third, wecombine our data with information on national payment systemscollected by the World Bank (2008). We find that respondents withbackgrounds in countries with high numbers of card transactionsper capita are less likely to use cash in the Netherlands. These threefindings are robust to including a variety of consumer, transactionand location characteristics.

In focusing on foreign backgrounds, this paper relates to earlierwork that reports differences in payment behavior based on race orethnicity (Borzekowski and Kiser, 2008; Borzekowski et al., 2008;Ching and Hayashi, 2010; Schuh and Stavins, 2010). However,these papers usually do not have detailed information on respon-dents’ country of origin. A second key difference is that these pa-pers are not able to distinguish between different generations. Incontrast, we are able to assess whether payment preferences arepassed on between generations. To this end, we use information

http://dx.doi.org/10.1016/j.jbankfin.2012.11.005

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.jbankfin.2012.11.005

http://www.sciencedirect.com/science/journal/03784266

http://www.elsevier.com/locate/jbf

990 A. Kosse, D. Jansen / Journal of Banking & Finance 37 (2013) 989–998

on whether an individual was born abroad (first-generation mi-grant), or whether she was born in the Netherlands, while one ofher parents was born abroad (second-generation migrant). Here,the key finding is an adjustment to host country modes ofpayment. Differences in payment behavior are only present forfirst-generation migrants. We do not find differences in usage ofpayment instruments between second-generation migrants andrespondents with a Dutch background.

On the basis of our main results, we discuss potential implica-tions for policy. Several studies show that instrument choices inPOS transactions significantly affect the overall efficiency of a pay-ment system.1 In general, substitution of cash by debit cards isfound to reduce social costs. On the one hand, we do find differencesin payment habits for several migrant groups. On the other hand, theresults show that payment habits change over time. Second-genera-tion migrants have not taken over their parents’ habits, but show anadjustment to host country modes of payment instead. We use astylized analysis to assess the net social benefits of specialized pub-licity campaigns or information material targeted at payment habitsof first-generation migrants from cash-oriented countries. Given therelatively high costs of such a campaign, it is not certain that the netsocial benefits will be positive.

This paper proceeds as follows: Section 2 presents a selectivereview of the relevant literature, both in the context of paymentsand use of financial services in general. Section 3 presents back-ground information about Dutch migrant groups and home coun-try payment patterns. Section 4 describes the methodology anddata. Section 5 analyzes the role of foreign backgrounds in choos-ing between payment instruments. Section 6 reports additional re-sults, while Section 7 concludes and discusses policy implications.

2. Related literature

A substantial amount of empirical research has examined con-sumer payment choice from a micro-perspective.2 Due to the lackof accurate transaction data, most empirical studies are based onself-reported survey data.3 Overall, the literature agrees that con-sumer payment choice at the POS and the adoption of electronicmeans of payment is influenced by consumer, transaction and situ-ational characteristics as well as by financial incentives.4

First, demographic factors are relevant. A common finding isthat the use of electronic means of payment is negatively corre-lated with age and positively related with a consumer’s educationand income level. Younger, more educated consumers with higherincomes are more likely to use electronic payment instruments,either at the POS or in remote transactions. In contrast, the elderly,consumers who have received less education and those with lowerincomes are more prone to using cash or other paper-based instru-ments. The rationale is that young and more educated people aremore open to new technologies and that young people lack the his-tory of relying on paper-based payments. Moreover, educated andhigh-income people have higher opportunity costs, and dislike the

1 See for example Brits and Winder (2005).2 The literature on consumer payment choice starts from the idea of heterogeneous

consumer preferences based on comparative product attributes and distinct con-sumer needs. Each payment instrument differs from the other with respect to costs,safety, anonymity, speed, acceptance and other characteristics and each consumerattaches a different importance to each of these characteristics. In the end, consumerchoice of which payment instrument to use is based on their net benefits received(see Bolt and Chakravorti (2010) and references therein).

3 A few empirical studies use transaction data provided by banks, grocery stores orcredit card companies (Rysman, 2007; Klee, 2008). Others have examined paymentchoice over time using aggregate data supplied by payment systems and data fromindustry sources (e.g. Humphrey et al., 1996; Amromin and Chakravorti, 2009).

4 The many relevant references include Kennickell and Kwast (1997), Jonker(2007), Borzekowski et al. (2008), Borzekowski and Kiser (2008), Klee (2008), Chingand Hayashi (2010), and Schuh and Stavins (2010).

greater amount of time it takes to initiate paper-based versus elec-tronic transactions (e.g. Kennickell and Kwast, 1997). In addition,some studies find a role for gender and region. Women are morelikely than men to use electronic payment media, such as paymentcards or electronic bill payments. Furthermore, the probability ofpaying by cards instead of cash is found to decrease with theurbanization degree of consumers’ living environment (Jonker,2007). This might capture adoption- and acceptance-related deter-minants, such as the regional density of ATMs and POS terminals.Moreover, Stavins (2001) finds that the fraction of other peoplein the region using the same type of payment instrument is alsoaffecting consumers’ usage patterns. This may not only indicate de-mand-related network effects, but also that own use of paymentinstruments is influenced by others’ habits.

Second, consumer payment choice is found to depend on trans-action characteristics, such as the transaction amount and the typeof good purchased. The size of the transaction is found to be a ma-jor determinant of consumers’ payment choice at the POS. Highertransaction amounts are more likely to be paid by cards insteadof cash, while cash is highly preferred for small-value transactions(Bounie and François, 2006; Jonker, 2007; Klee, 2008; Von Kalck-reuth et al., 2009).

Third, location matters. For instance, the absence of a cashier,e.g. at vending machines, usually increases the probability of a cashpayment (Hayashi and Klee, 2003). Bounie and François (2006) andJonker (2007) also show that payment choices differ according tothe location, which most probably reflects the different levels ofpenetration of payment terminals across stores and sectors. Rys-man (2007), for example, demonstrates that consumer paymentchoice is highly correlated with the level of card acceptance byretailers.

Fourth, consumer payment choices are found to be influencedby financial incentives. Bolt et al. (2010) demonstrate that consum-ers react strongly to transaction charges imposed by retailers forparticular payment instruments. In addition, explicit pricing bybanks is shown to affect payment choices (e.g. Borzekowskiet al., 2008). The payments literature has also shown significantlylarge and positive effects of incentive and reward programs (e.g.Ching and Hayashi, 2010) and card discounts, points and cash-backs are generally found to have a positive effect on the use ofpayment cards relative to cash (Carbó-Valverde and Liñares-Zegar-ra, 2011).

In focusing on foreign backgrounds, our paper is related to workthat reports differences in payment behavior based on race or eth-nicity (Borzekowski and Kiser, 2008; Borzekowski et al., 2008;Ching and Hayashi, 2010; Schuh and Stavins, 2010). However,these papers often only include race as additional covariates. Inother fields, the role of foreign backgrounds has been studied moreextensively. A dimension that receives increasing attention is mi-grant participation in financial service markets. Immigrants tendto be less ‘banked’ than the native population. Osili and Paulson(2009), for example, show that immigrants are less likely to owna saving and checking account compared to the native-born. Jan-kowski et al. (2007) analyze currency demand in Chicago and findthat Latin American immigrants demand more $100 bills thanother residents. Since these bills are mainly held as a store of valueinstead of for payment purposes, the results may either indicatebarriers that Latin American immigrants face or their reluctanceto open and maintain bank accounts. Studying the demand forlarge banknotes in Swiss, however, Fischer (2010) finds that immi-grants, due to wealth and age effects, hoard less than natives. Final-ly, Campbell et al. (2012) find that involuntary closure rates forbank accounts are higher in countries with high black populationsand lower in counties with Hispanic and Asian populations.Although the effects are sizeable, it is not immediately clear whatdrives these findings.

Table 1Overview Dutch population (15 years and older) in 2008.

Background Total population (>15 years) %

Turkey 271,660 21Morocco 227,809 18Suriname, Dutch Antilles 365,558 28Other non-western 426,318 33

Total non-western 1,291,345 100Of which:

1st generation non-western 973,214 752nd generation non-western 318,131 25

Indonesia 368,447 29Germany 346,510 27Eastern Europe 185,386 15Other western 360,851 29

Total western 1,261,194 100Of which:

1st generation western 569,030 452nd generation western 692,164 55

Total immigrants 2,552,539 19Total native population 10,917,136 81

Total Dutch population 13,469,675 100

Source: Statistics Netherlands.

A. Kosse, D. Jansen / Journal of Banking & Finance 37 (2013) 989–998 991

There is evidence that the culture of the country of origin influ-ences behavior of immigrants in host countries. Osili and Paulson(2008) find that immigrants from countries with more effectiveinstitutions are more likely than other immigrants to have a rela-tionship with a bank and to use formal financial markets moreextensively. Jankowski et al. (2007) and Osili and Paulson (2009)too, claim that immigrants from countries having a strong institu-tional environment may be more likely to have a bank account intheir host country, whereas immigrants having experienced finan-cial crises in their home countries might be less likely to partici-pate in the host country’s financial system. Kok et al. (2011) findthat a high female participation rate in the home country corre-lates with a high female participation rate in the host country. Thishome country effect, however, vanishes over generations, with sec-ond-generation migrants being more influenced by the culture ofthe host country. Not all the evidence on home country effects isconclusive. For instance, Seto and Bogan (forthcoming) show thatUS immigrant asset market participation rates vary depending oncountry of origin, but find mixed evidence that investment behav-iors are carried over from the country of origin.

3. Migration and payment habits in the Netherlands

Residents with a foreign background make up around 20% of theDutch population.5 Since years, migrants from non-western coun-tries have made up the largest share of migrants, with the majorityoriginating from Turkey and Morocco, followed by Suriname and theDutch Antilles, as illustrated by Table 1. Migration from Turkey andMorocco started in the 1960s, when there was a high demand forlow-skilled workers which could not be fulfilled by the immigrantworkers from Spain and Italy. Many of the Turkish and Moroccanworkers stayed and their families migrated to the Netherlands dur-ing the subsequent decades (Kok et al., 2011). For instance, it is esti-mated that 65,000 Turkish migrants came to the Netherlands in the1960s and early 1970s. This group often came from Mid- and South-ern Turkey where the unemployment rate was generally at a high le-vel (CGM, 2012).

The inflow of Surinam migrants started in the 1950s and mainlyconcerned young people who came to study. Since the mid-1970s,

5 This estimate by Statistics Netherlands refers to the Dutch population of 15 yearsand older in 2008.

however, migration from Suriname has mainly been driven byother reasons, such as better job opportunities and political rea-sons. In 1975 alone, the prospect of independence led around40,000 people to migrate to the Netherlands (CGM, 2012). Themigration pattern from the Dutch Antilles shows similar patterns.Initially, Antilleans came to the Netherlands to take up a study, butas from the mid-1990s more Antilleans have migrated to the Neth-erlands in the hope of finding prosperity (CGM, 2012).

The largest share of western immigrants originates from Indo-nesia (the then Netherlands East Indies) (Table 1). The official clas-sification by Statistics Netherlands of Indonesia as a westerncountry is strongly related to the colonial linkages between Indo-nesia and the Netherlands. In fact, many individuals from Indone-sia originally have a Dutch background. We will return to this issuein Section 4. Many Indonesians came to the Netherlands after thedecolonization of Indonesia directly after the Second World War(CGM, 2012). Germans make up the second largest group of wes-tern immigrants, followed by people from other European coun-tries (Table 1). Especially the share of migrant workers fromEastern Europe has strongly increased since the entry of Poland,Bulgaria and Romania into the European Union. Since a few years,Poland constitutes the fastest growing group of migrants to theNetherlands (CGM, 2012).

For some migrant groups, payment habits in their respectivehome countries differ substantially from those in the Netherlands.To examine this, ideally, one would like to use a consistent set ofcross-country data on usage of cash and cards. As far as we areaware, the data provided by the World Bank (2008) is the best can-didate source for our purposes. Fig. 1 presents two relevant indica-tors: the number of payment card transactions per capita and thenumber of payment cards per 1000 inhabitants. It compares datafor the Netherlands with those for the home countries of the majormigrant groups living in the Netherlands. Unfortunately, Surinameis not taken up in the World Bank dataset, and these particular twoindicators were not available for the Dutch Antilles. We did try toobtain further data for these countries from other sources, butwithout success.

In the Netherlands, payment cards are used extensively at thePOS. In particular, debit cards are heavily used. In 2006, Dutchinhabitants owned on average 1.6 debit cards per person and theyused their debit card about 90 times a year. Despite being a neigh-boring country, payment behavior in Germany substantially dif-fers. Most Germans own a debit card, but they use it lessextensively than the Dutch. On average, Germans made only 29card payments per person in 2006. The reliance of German con-sumers on cash has also been established elsewhere in the litera-ture (e.g. Von Kalckreuth et al., 2009). In Morocco, Indonesia,Poland and Turkey as well, payment cards are significantly less of-ten used for paying for POS purchases, pointing at a greater reli-ance on cash. In Section 6, we will use the World Bank data in afurther analysis and relate the number of payment card transac-tions in various countries to the choice between cash and debitcards in the Netherlands.

4. Survey methodology and data description

To examine the effect of domestic versus foreign backgrounds,we conducted an extensive survey among Dutch consumers, forwhich the data collection took place between March and July2009. The survey aimed to gather sufficient data on the variousethnicity groups living in the Netherlands. In defining the variousethnic groups, we relied on official classifications by StatisticsNetherlands, as those definitions are also commonly used in Dutchpolicy debates (see Table 2). The same classifications are also usedin other papers, for instance in Kok et al. (2011). To classify individ-

Fig. 1. Payment card possession and usage in different countries – 2006. Source: World Bank (2008).

Table 2Classification of foreign backgrounds.

Description Criteria

Country of birth Country of birth parentsDutch background Not relevant The NetherlandsForeign background At least one of the parents not born in the

Netherlands1st Generation foreign

backgroundNot in the Netherlands At least one of the parents not born in the

Netherlands2nd Generation foreign

backgroundThe Netherlands At least one of the parents not born in the

NetherlandsWestern background A country in Europe (excl. Turkey), North America, Oceania, Indonesia or

JapanAt least one of the parents not born in theNetherlands

Non-western background A country in Africa, South America, Asia (excl. Indonesia and Japan) orTurkey.

At least one of the parents not born in theNetherlands

Notes: This table provides information on the official classification of individuals with foreign backgrounds, as defined by Statistics Netherlands (www.cbs.nl).

6 Using a combination of survey methods may possibly introduce biases. Asdemonstrated by Jonker and Kosse (2009), the survey methodology used maysignificantly affect consumers’ registration of payments. However, given the consis-tency in survey setup, design and length, the effects of using online, face-to-face andpaper-based techniques are expected to be limited.


ual countries, Statistics Netherlands takes the social-economic andsocial-cultural similarities to individuals with a Dutch backgroundinto account, and to a lesser extent their geographical locations.Therefore, some countries are classified under a different categorythan one would initially expect based on their geographicalposition.

The first group our survey targeted was individuals with aDutch background, which means that both parents are born inthe Netherlands. The aim was to have at least 400 observations(individuals) for this group. The second group was individuals witha foreign background, meaning that at least one of the parents isnot born in the Netherlands. For this group, a target of 1600 per-sons was specified. Within this group, it is possible to make twofurther subdivisions on the basis of the Statistics Netherlands clas-sification. If the person itself is born outside the Netherlands, she isclassified as a first-generation individual with a foreign back-ground. If the country of birth is the Netherlands, then she is seenas second-generation. A second distinction is that between wes-tern and non-western. For first-generation individuals, if the coun-try of birth is in Europe (excluding Turkey), North America,Oceania, Indonesia or Japan, that person is classified as western.For second-generation persons with foreign backgrounds, the dis-tinction between western and non-western is first based on thecountry of birth of the mother. If the mother is born in the Nether-lands, then the father’s birthplace is used. The aim was to have atleast 400 persons with a western background, and 1200 individualswith a non-western background. The sample of non-western indi-

viduals was selected in such a way that the four major countries oforigin, i.e. Turkey, Morocco, Suriname and the Dutch Antilles, wereadequately represented.

Ethnic minorities are usually underrepresented in consumersurveys, due to the complexity and high costs of reaching andapproaching them. In order to accommodate to the specific charac-teristics and attitudes of these groups and to minimize non-re-sponse, we used a combination of survey techniques.6

Respondents with a Dutch or western background were mainly se-lected using an existing internet panel. This group also answeredthe questionnaire online. However, to mitigate a potential bias,non-internet users were contacted by sending out letters. If theywere willing to participate, the respondents were subsequently ap-proached for face-to-face interviews. Virtually all respondents witha Turkish, Moroccan, Surinam and Antillean background were se-lected using a quota procedure, where the interviewers used theirown networks and visited specific places with a high probability ofencountering the targeted respondents. The Surinam and Antilleanrespondents were subsequently surveyed in a face-to-face interview.Respondents with a Moroccan and Turkish background, however,were more reluctant about participating in a face-to-face survey,

http://www.cbs.nl

Table 3Characteristics of survey participants.

(1) (2) (3) (4) (5) (6) (7) (8)Population Full sample Background

Dutch ForeignAll Western Non-western

1st Generation 2nd Generation 1st Generation 2nd Generation

Female 0.49 0.53 0.54 0.53 0.56 0.53 0.52 0.52Age 46.9 41.4 48.4 38.7 50.0 48.1 42.0 26.5

EducationNone 0.04 0.01 0.04 0.02 0.00 0.09 0.01Primary 0.08 0.15 0.12 0.16 0.03 0.11 0.23 0.13Secondary 0.64 0.6 0.61 0.59 0.59 0.63 0.55 0.63BA 0.17 0.17 0.2 0.15 0.22 0.18 0.1 0.19MA 0.10 0.05 0.05 0.05 0.15 0.07 0.04 0.04

IncomeNone 0.09 0.08 0.09 0.04 0.07 0.09 0.12<1000 0.27 0.22 0.29 0.29 0.21 0.26 0.371000–2000 0.33 0.32 0.33 0.34 0.27 0.42 0.252000–3000 0.19 0.22 0.17 0.15 0.26 0.15 0.14>3000 0.13 0.16 0.12 0.18 0.19 0.08 0.12

With partner 0.62 0.64 0.66 0.63 0.83 0.79 0.71 0.4Urban 3.0 3.8 3.2 4.0 3.7 3.5 4.2 4.2

RegionWest 0.58 0.43 0.64 0.56 0.51 0.66 0.7North 0.06 0.1 0.05 0.11 0.05 0.05 0.03East 0.16 0.2 0.14 0.2 0.15 0.13 0.14South 0.2 0.27 0.17 0.13 0.29 0.16 0.13

Bank account (NL) 0.98 1.00 0.97 0.99 1.00 0.96 0.97# Bank accounts (NL) 1.6 1.7 1.5 1.9 1.8 1.4 1.4Bank account (abroad) 0.06 0.01 0.07 0.11 0.01 0.11 0.04Ever use debit card? 0.94 0.96 0.93 0.97 0.99 0.89 0.94# Respondents 2258 620 1638 123 272 724 519

Fraction of totalIn sample 0.27 0.73 0.05 0.12 0.32 0.23In population 0.80 0.20 0.04 0.05 0.06 0.05

Notes: This table summarizes the various control variables for the respondents to the survey. Column 2 describes the full sample, while columns 3–8 show summaries forvarious background categories. Column 1 gives data for the Dutch population based on Statistics Netherlands (www.cbs.nl, population aged over 15). Numbers representfractions, with the exception of age and the number of bank accounts, which are shown in averages, and the number of respondents, which are shown in totals. The datashown from the survey are un-weighted.


even in case of interviewers with a Moroccan or Turkish background.In particular, the respondents had reservations about providing per-sonal and financial information to the interviewer. Also, there wasfear of making mistakes because of insufficient command of theDutch language. To address these concerns, paper-based interviewtechniques were used for these particular groups.

The survey consisted of two parts. First, respondents docu-mented their expenses during one day in a transaction diary.7

The request was to register the time of the purchase, the location,the method of payment, the transaction amount and whether thetransaction was business or private. Regarding the location, a pre-defined set of twenty types was given. The second part of the surveywas a list of detailed background questions.

Turning to a description of the data, Table 3 presents an over-view. The target number of respondents was met for all the groups.The sample includes 620 individuals with a Dutch background and1638 respondents with a foreign background. Column 2 summa-rizes consumer characteristics for all 2258 individuals in our sam-ple. As a benchmark, column 1 and the last line of the table presentinformation on the Dutch population based on data from Statistics

7 Jonker and Kosse (2009) demonstrate that one-day transaction diaries are thepreferred methodology for assessing payment behavior. One-week registrationmethods and retrospective interviews are shown to lead to a significant increase ofincomplete recall and zero observations due to diary fatigue and diary exhaustion.

Netherlands. Columns 3–8 present a breakdown based onbackgrounds.

The sample differs on a number of dimensions from the Dutchpopulation. On average, the respondents are younger, more oftenfemale, less highly educated, and more likely to live in an urbanenvironment in the western part of the Netherlands. Also, giventhe survey design, persons with a foreign background are overrep-resented. Given these differences, we constructed samplingweights based on gender, age, education, degree of urbanization,region where the individual lived, and the ethnic background,which we will use in all regressions.

Table 4 shows information on the number of cash and debitcard payments recorded by the various groups.8 For the full sample,the average number of cash payments was roughly 1.6 per personper day, compared to 0.7 for the number of debit card payments.The ratio of cash payments versus the total number of payments isfairly equal across the various groups, with one clear exception.For first-generation non-western respondents, the ratio is around0.75, compared to 0.66 for the other groups.

8 We focus on cash and debit cards only, as these two payment instruments wereby far the most common in our survey. This is not surprising since other instruments,such as the credit card and the e-purse are rarely used in the Netherlands: about 97%of total POS transactions are paid by cash or debit card (Jonker et al., 2012).

http://www.cbs.nl

Table 4Payment characteristics.

(1) (2) (3) (4) (5) (6) (7)Full sample Background

Dutch ForeignAll Western Non-western

1st Generation 2nd Generation 1st Generation 2nd Generation

# Cash payments 1.61 1.50 1.66 1.70 1.45 1.80 1.56# Debit card payments 0.73 0.74 0.73 0.86 0.75 0.64 0.80Ratio cash/total 0.69 0.67 0.69 0.66 0.66 0.74 0.66Cash at start (EUR) 26.4 28.0 25.5 35.4 30.3 30.5 20.0

Notes: This table gives information on the number of cash and debit card payments recorded per person per day for various categories. Column 1 describes the full sample,while columns 2–7 show summaries for various background categories. Numbers represent un-weighted averages.


5. Cash or debit card: Results

We now turn to an econometric analysis of the choice betweencash or debit card payments. The dependent variable is a binarydummy equal to one in case the respondent used cash in thepoint-of-sale transaction. In total, we analyze 4225 transactions.9

We first run a benchmark probit regression without variables on for-eign backgrounds. We run the probit regressions using a rich set ofcovariates, consisting of consumer, transaction and location charac-teristics suggested by the literature.10 The consumer characteristicsinclude the following variables: gender, age, education, income cat-egory, marital status, household size, homeownership, region (north,east or south), ZIP code, the number of bank accounts abroad, andthe amount of cash in pocket at the start of the day. The referencecategory is a debit card payment made by a male, from the age groupbetween 35 and 44, whose highest qualification is primary school.The transaction characteristics include the amount of the transac-tion, whether the transaction was private or business and the dayof the week (Monday to Saturday). In addition, we use dummiesfor the various pre-defined locations where the transaction occurred.As a benchmark location, we use the supermarket, which accountedfor 28% of all transactions. As a final control, we include the type ofsurvey instrument used (internet, face-to-face or paper-based).

Table 5 shows parameter estimates and standard errors (in ital-ics) for the benchmark probit model. We only show results for se-lected covariates. Regarding the consumer characteristics, thereare a number of intuitive results. In accordance with the existingpayments literature, we find that age and education play a signif-icant role, with younger, more educated consumers being morelikely to pay electronically. However, we find no differences be-tween males and females. The negative parameter suggests that fe-males are less likely to use cash, but the effect is not significantlydifferent from zero. Our results further show that consumers hav-ing a partner are less likely to use cash. One explanation could be awealth effect, which would be in line with earlier findings of higherincome people being more prone to use electronic payment instru-

9 Our analysis is at the transaction-level rather than at the consumer-level. Of thepeople who reported at least one transaction, 25% reported one, 26% reported two and18% reported three purchases. We cluster standard errors by respondent to allow forpossible correlation across transactions by the same individual.

10 As suggested by the literature (e.g. Bolt et al., 2010), financial incentives may playa role as well. Although Dutch consumers are not faced with explicit bank imposedtransaction fees or incentive and reward programs, POS transactions may carry someretailer imposed costs, as retailers in the Netherlands are allowed to apply a surchargefor specific payment instruments. Bolt et al. (2010) found that 22% of the Dutchretailers applied a surcharge for small payments in 2006. Since then, the relativesafety and efficiency advantage of debit cards has led to a strong decline in the usageof surcharges. At the time of our survey, it is estimated that only 5% of the debit cardaccepting retailers was applying a surcharge in 2009 (Hoofdbedrijfschap Detailhan-del, 2011). Still, it would have been useful to use a dummy variable to control for thepotential effect of surcharges. Unfortunately, due to the unavailability of data, wewere not able to do so.

ments. Moreover, the amount of cash that people carry with themhas a significant positive relation to cash usage during the day.Regarding the transaction characteristics, the findings are as ex-pected. Higher transaction values are less likely to be paid in cash.In terms of location, cash is more often used at street vendors, atsmall food stores and in restaurants and bars. Purchases made atfashion and shoe stores and at gas stations, by contrast, are moreoften paid by debit card. Since we have already controlled for dif-ferences in transaction values explicitly, these findings most prob-ably reflect the different levels of debit card acceptance across thedifferent types of stores.11 Finally, there is an indication that con-sumers are more likely to use their debit cards on Wednesday.

In the next step, we include variables measuring foreign back-ground into the benchmark probit model. We do this in two parts.First, we use binary dummies following the official classificationsby Statistics Netherlands, which were discussed in the previoussection and summarized in Table 2. So, we use dummies that mea-sure whether an individual has a foreign background, from whichregion she stems, and whether she is a first or second-generationmigrant. Second, we further refine the analysis by using binarydummies defined – as much as possible – on a person’s countryof origin.12 The results for the first exercise are in Table 6, whilethe results for the second analysis are in Table 7. Both tables presentaverage marginal effects rather than parameter estimates.

First of all, we find no overall difference between individualswith Dutch and foreign backgrounds. Someone with a foreignbackground is around 1.8% points (pp.) more likely to use cash,but the effect is not significantly different from zero (Table 6, col-umn 1). When we split the data based on generations, again wefind no significant differences, although there is a hint of differ-ences between generations (column 2). For second-generation mi-grants, the marginal effect is essentially zero, whereas for first-generation migrants, the probability of choosing cash is around3 pp. higher. When we take the region of origin into account (col-umn 3), there is an indication that persons with non-western back-grounds are more likely to use cash. Again, though, the differencesare not significant. When we combine the information on regionand generation (column 4), we find one clear difference betweenforeign and domestic backgrounds. For first-generation migrantswith a non-western origin, the probability of using cash is 6 pp.higher compared to persons with a Dutch background. For sec-ond-generation non-westerners, the chances of using cash areactually smaller, by 0.7 pp., although not significantly different.

11 About 90% of all Dutch retail traders are able to accept debit card payments. Thelevels of penetration of payment terminals, however, considerably differs acrossdifferent types of stores: supermarkets (100%), gas stations (100%), fashion stores(97%), specialized food stores (82%), catering (64%) and street vendors (28%)(Hoofdbedrijfschap Detailhandel, 2011).

12 We still aggregate Eastern European and other European countries, as we have alimited number of observations from individual countries in these groups.

Table 5Using cash over debit card: the role of consumer, transaction andlocation characteristics.

Consumer characteristicsFemale �0.07

0.11Age

15–24 �0.030.16

25–34 0.010.15

45–54 0.130.15

55–64 0.35**

0.1665 and older 0.16

0.16Education

Secondary �0.28*

0.15Bachelor �0.66***

0.21Master �0.49**

0.23Partner �0.21*

0.12Cash at start of day 0.00**

0.00

Transaction and location characteristicsAmount �0.01***

0.00Location

Street vendor 1.46***

0.24Food (small shop) 0.67***

0.15Fashion/shoes �0.89***

0.27Restaurant/bar 0.76***

0.14Gas station �0.69***

0.15

Day of weekWednesday �0.43**

0.19Constant 0.62

0.55

Notes: Parameter estimates and standard errors (in italics) for aprobit regression, where the dependent variable is a binarydummy equal to 1 in case of cash payments and 0 in case of debitcard payments. The results are based on 4225 transactions. Themodel includes a full set of variables, described in the main text.The table shows results for selected covariates. Observations areweighted on the basis of gender, age, education, ethnic back-ground, degree of urbanization and region. Standard errors areclustered by respondent.

* Denotes significance at the 10% level.** Denotes significance at the 5% level.

*** Denotes significance at the 1% level.

Table 6Using cash over debit card: the role of region of origin and generation.

(1) (2) (3) (4)

Foreign background 0.0180.023

1st Generation 0.0290.029

2nd Generation 0.0010.024

Western 0.0050.028

Non-western 0.0380.029

1st Generation western �0.0030.042

1st Generation non-western 0.060*

0.031

2nd Generation western 0.0120.029

2nd Generation non-western �0.0070.035

Notes: Marginal effects and standard errors (in italics) based on probit regressions.The dependent variable is a binary dummy equal to 1 in case of cash payments and0 in case of debit card payments. This table classifies individuals with a foreignbackground according to the official definitions of Statistics Netherlands. Theregression includes a full set of control variables, described in Section 5 of the maintext. The reference category is a debit card payment in a supermarket by a malewith a Dutch background, from the age group between 35 and 44, whose highestqualification is primary school. Results are based on 4225 observations for columns1 and 3, and on 4214 observations for columns 2 and 4. Observations wereweighted on the basis of gender, age, education, ethnic background, degree ofurbanization and region. Standard errors are clustered by respondent.* Denotes significance at the 10% level.⁄⁄Denotes significance at the 5% level.⁄⁄⁄Denotes significance at the 1% level.

13 One tentative explanation is that a higher degree of familiarity with the Dutchlanguage has facilitated the adjustment to Dutch payment habits. Analyzing the roleof language skills in payment instrument choice seems an interesting extension forfuture research. Another explanation may be that migrants from Indonesia are usuallyfrom strata with strong social-economic backgrounds and are already more familiarwith card payments. We thank Dr. Irma Hindrayanto for discussions on these twopoints.


For individuals from western countries, there are no significantdifferences.

Overall, the results in Table 6 suggest that region of origin andgeneration are relevant factors for payment behavior. Table 7 fur-ther expands the analysis by using information on the country oforigin (column 1) and a further break-down by generations (col-umn 2). Participants with a German background are around9.3 pp. more likely to choose cash, while persons with Turkishand Moroccan origins are around 7.5 pp. more probable to paycash. Once again, differences between generations are present (col-umn 2). For individuals with German, Turkish or Moroccan back-grounds, the higher usage of cash is restricted to the firstgeneration. In fact, for second-generation persons with Turkish or

Moroccan origins, the differences with the reference group are neg-ligible. For those participants with German origins, the difference isstill around 6 pp. for the second generation, but no longersignificant.

To summarize, we find indications that first-generation mi-grants from three countries that can be seen as cash oriented (asdiscussed in Section 3) continue to have strong preferences forcash after migration to the Netherlands. However, for anothercash-oriented economy (Indonesia), we find no clear significantdifferences.13 Therefore, we will further explore the role of cash ori-entation using data from the World Bank (2008) on national pay-ment systems. This exercise is described in Section 6.

6. Additional results

This section presents three additional analyses. As a first check,we focus on payments exclusively made in supermarkets and gasstations. This serves as an additional check on any supply-relatedfactors, such as the (un)availability of POS terminals. Both sectorsare homogeneous in terms of payment acceptance as in virtuallyall supermarkets and gas stations consumers have the opportunityto choose between cash or debit cards. As shown in Table 8, we find

Table 7Using cash over debit card: results using home countries.

(1) (2)

Germany 0.093**

0.0381st Generation 0.129**

0.0572nd Generation 0.063

0.046

Eastern Europe 0.0230.051

1st generation 0.0030.071

2nd generation 0.0570.063

Europe (other) �0.0540.048

1st Generation �0.1140.092

2nd Generation �0.0200.048

Indonesia �0.0330.044

1st Generation �0.0200.078


Suriname 0.0240.034



Dutch Antilles �0.0120.041



Turkey 0.075*

0.0391st Generation 0.089**


0.053

Morocco 0.074*

0.0441st Generation 0.088*


0.057

Notes: Marginal effects and standard errors (in italics) based on probit regressions.The dependent variable is a binary dummy equal to 1 in case of cash payments and0 in case of debit card payments. This table classifies individuals with a foreignbackground according to their country of origin (column 1). Column 2 is based on afurther analysis of the data based on generation (column 2). The regression includesa full set of control variables, described in Section 5 of the main text. The referencecategory is a debit card payment in a supermarket by a male with a Dutch back-ground, from the age group between 35 and 44, whose highest qualification isprimary school. Results are based on 4225 transactions in column 1 and on 4214transactions in column 2. Observations are weighted on the basis of gender, age,education, ethnic background, degree of urbanization and region. Standard errorsare clustered by respondent.* Denotes significance at the 10% level.** Denotes significance at the 5% level.⁄⁄⁄Denotes significance at the 1% level.

Table 8Using cash over debit card: focusing on payments in supermarkets and gas stations.

(1) (2)Supermarket Gas station

1st GenerationWestern �0.010 0.160

0.070 0.107Non-western 0.047 0.132**

0.052 0.061

2nd GenerationWestern 0.009 0.075

0.052 0.068Non-western �0.050 0.048

0.060 0.075

Notes: Marginal effects and standard errors (in italics) based on probit regressions.The dependent variable is a binary dummy equal to 1 in case of cash payments and0 in case of debit card payments. Column 1 uses 1208 purchases made in super-markets, while column 2 uses 273 payments made at gas stations. The regressionsinclude a full set of control variables, described in Section 5 of the main text. Thereference category is a debit card payment made by a male with a Dutch back-ground, from the age group between 35 and 44, whose highest qualification isprimary school. Observations are weighted on the basis of gender, age, education,ethnic background, degree of urbanization and region. Standard errors are clusteredby respondent.⁄Denotes significance at the 10% level.** Denotes significance at the 5% level.⁄⁄⁄Denotes significance at the 1% level.

14 We thank the referee for suggesting this extension.


no differences between persons with Dutch and foreign back-grounds when it comes to paying in supermarkets (column 1). Nei-ther the region of origin nor the generation are relevant factors forpayment choices made for supermarket purchases. First-genera-tion non-westerners are 4.7 pp. more likely to pay in cash, butthe differences are not significant.

Turning to payments at gas stations (column 2), however, we dofind a significant difference: the probability of using cash is 13 pp.higher among non-westerners compared to individuals with Dutchbackgrounds. For the other groups as well, the likelihood of usingcash is higher, although not significantly different. The results con-firm our earlier finding. Again, differences in payment methods areonly present for first-generation non-westerners. Since we haveaccounted for one of the most important location-related charac-teristics, i.e. the availability of payment terminals, this differenceis most probably consumer-related. The reason why we only finddifferent payment patterns in gas stations and not in supermarketscan be attributed to the fact that they are two very different sectorsregarding the type and size of purchases done and the share of cashand debit cards used. Overall, 60% of all supermarket transactionsand 37% of total supermarket sales are paid by cash, whereas in gasstations the share of cash is significantly lower; 44% of all transac-tions and 18% of total sales (Jonker et al., 2012). The stronger pref-erence for cash among first-generation non-westerners is thereforeclearly emphasized in this latter debit card prevailing sector.

Second, there may be an endogeneity issue regarding theexplanatory variable ‘cash-at-start’, which we include to modelpayment choice. Consumers having a higher preference for cashcan be expected to carry larger amounts of cash. To explore thispossible endogeneity, we re-ran the regressions in Tables 5–7without the variable ‘cash-at-start’. Overall, there were no substan-tial changes to our conclusions on the role of foreign backgrounds.Both marginal effects and standard errors were broadly similar.Biases due to the possibly endogenous nature of this variable aretherefore concluded to be limited.

Third, we use the World Bank (2008) data to study more di-rectly the role of cash orientation in the home country.14 Table 9uses data on the number of card transactions per capita to fleshout the distinction between cash and non-cash oriented countries.Of the available World Bank indicators, this variable is the most intu-itive way to measure cash orientation, as it is based on actual pay-ment choices. One caveat to the analysis is the missing data forSuriname and the Dutch Antilles, which means we lose around

Table 9Using cash over debit card: link with number of card transactions per capita in home country.

(1) (2) (3)Debit card transactions Credit card transactions Card transactions

Marginal effect �0.0017** �0.0056*** �0.0025***

0.0008 0.0015 0.0007Number of observations 2040 1215 1215

Notes: Marginal effects and standard errors (in italics) based on three probit regressions. In each case, the dependent variable is a binary dummy equal to 1 in case of cashpayments and 0 in case of debit card payments. The right hand side variables are the number of debit card transactions per capita (column 1), the number of credit cardtransactions per capita (column 2) or the total number of card payments per capita (column 3) in the country of origin in 2006 as reported in World Bank (2008). Theregressions include a full set of control variables, described in Section 5 of the main text. Observations are weighted on the basis of gender, age, education, ethnic background,degree of urbanization and region. The samples include all respondents with a foreign background. Standard errors are clustered by respondent.⁄Denotes significance at the 10% level.

** Denotes significance at the 5% level.*** Denotes significance at the 1% level.

15 See for example Brits and Winder (2005).16 For this calculation, we combine cost estimates reported in Brits and Winder

(2005), McKinsey&Company (2006) and EIM (2007). Following Brits and Winder(2005), we only consider variable costs and distinguish between costs varying withthe number of transactions and costs varying with the value of the transactions.Further details of this calculation are available upon request.

17 In 2002, the total costs of all POS payments in the Netherlands added up to EUR2.9 billion or, equivalently, 0.65% of GDP (Brits and Winder, 2005).


250 observations. Another caveat is the fact that for some countrieseither the number of debit card or the number of credit card trans-actions is missing. Therefore, Table 9 presents three columns whichuse either debit or credit cards, or the sum of both. In each case, weuse data from the 2006 vintage of the World Bank data as the num-ber of missing observations is smallest for that year. The probitregressions use all respondents with a foreign background.

For each of the three probit regressions in Table 9, we find anegative relationship between card use in the home country andthe likelihood of a cash payment by a respondent with a foreignbackground in the Netherlands. So, a more intensive use of cardsin countries of origin is associated with a lower likelihood of usingcash in the Netherlands. This finding supports the earlier results inTables 6–8 that payment habits related to foreign backgroundscontinue to influence the choice of payment instruments aftermigration. Although the marginal effects are significant, we wouldnote that the estimated economic significance is relatively moder-ate. For instance, the result in column 3 suggests that ten addi-tional card payments per capita would only decrease thelikelihood of cash payments by around 0.025%. With more (precise)data, future research will be better equipped to further explore therole of home country cash orientation.

7. Conclusions and policy implications

This paper examines a detailed set of transaction and consumerdata that was collected through a diary survey. The innovative as-pect of this diary survey is that we have detailed information onboth individuals with a Dutch and a foreign background. Overall,we find a number of results suggesting that foreign backgroundsare relevant for payment behavior after migration. First, we findthat first-generation migrants from non-western countries aremore likely to use cash. Second, we find that persons with back-grounds in three countries that, compared to the Netherlands,can be classified as cash-oriented are still more likely to use cashin point-of-sale transactions in the Netherlands. Third, we find evi-dence that the number of card transactions per capita in countriesof origin is negatively related to the likelihood of using cash in theNetherlands. These findings are robust to controlling for a range ofconsumer, transaction and location characteristics. At the sametime, we also find that payment habits of the migrant populationchange over time. For second-generation migrants, there is nolonger evidence of different payment habits compared to individu-als with a Dutch background.

Turning to policy implications, a possible question is whetherthere is a case for targeted information campaigns promoting theuse of debit cards. For instance, in the Netherlands one recent cam-paign set up by banks and retailers encouraged consumers to usedebit cards also for small transactions. A further replacement ofcash by debit card payments may foster the social efficiency and

safety of the payment system.15 Debit card transactions are usuallycheaper for banks and retailers and for safety reasons, many retailersprefer to have a minimum of cash in their till. By enticing people toturn away from cash to debit cards, potential efficiency gains couldbe reaped.

Based on our results, would it be worthwhile to develop spe-cially targeted programs for first-generation migrants from cash-oriented economies? First, our findings suggest that habits relatedto foreign backgrounds, which may be difficult to change, play animportant role in payment choices. In order to engineer a substan-tial shift in habits, a publicity campaign would presumably be rel-atively costly and long-term. Second, as noted in Section 4, it isgenerally difficult to approach the particular groups that showhigher preferences for cash. Again, this would imply that a cam-paign would be relatively costly. Third, one needs to considerhow much substitution from cash to debit card payments a cam-paign could generate. In all likelihood, an educational campaignwill only affect payment habits of a fraction of the target groupwith high cash preferences. A stylized analysis is useful here toput potential net benefits in perspective.16 In the most optimisticscenario, suppose the campaign is fully successful by reducing thecash ratio of individuals from cash-oriented economies from 74%to the average level of 66%. This means a substitution of 0.2 cashpayments per person per day, adding up to a replacement of 36.5million cash payments a year. Assuming an average value of cashpayments of EUR 12.19 (Jonker et al., 2012), the total direct socialcosts saved would amount to EUR 3.3 million. In light of the total so-cial costs of POS payments in the Netherlands, this is still a modestefficiency gain.17 Overall, given the relatively high costs of a special-ized education or information campaign, it is not certain that the netsocial benefits would be positive.

Acknowledgments

We gratefully acknowledge the important contributions to thisproject by Cindy Horbach, Lin Feei Lee and Maarten Jansen. Wethank Ike Mathur (the editor), an anonymous referee, Wilko Bolt,Hans Brits, Irma Hindrayanto, Nicole Jonker and seminar partici-pants at various venues for useful guidance, comments and discus-sions which greatly improved the paper. Any errors and omissionsare our responsibility. The views expressed in this article do notnecessarily coincide with those of the Nederlandsche Bank.


References

Amromin, G., Chakravorti, S., 2009. Whither loose change? The diminishing demandfor small-denomination currency. Journal of Money, Credit and Banking 41,315–335.

Bolt, W., Jonker, N., Van Renselaar, C., 2010. Incentives at the counter: an empiricalanalysis of surcharging card payments and payment behaviour in theNetherlands. Journal of Banking and Finance 34, 1738–1744.

Bolt, W., Chakravorti, S., 2010. Digitization of Retail Payments. Working Paper, DeNederlandsche Bank.

Bounie, D., François, A., 2006. Cash, check or bank card? The Effects of TransactionCharacteristics on the use of Payment Instruments. Working Paper, TelecomParis Economics and Social Sciences.

Borzekowski, R., Kiser, E.K., 2008. The choice at the checkout: quantifying demandacross payment instruments. International Journal of Industrial Organization26, 889–902.

Borzekowski, R., Kiser, E.K., Ahmed, S., 2008. Consumers’ use of debit cards:patterns, preferences and price response. Journal of Money, Credit and Banking40, 149–172.

Brits, J.H., Winder, C.C.A., 2005. Payments are no Free Lunch. Occasional Studies, DeNederlandsche Bank.

Campbell, D., Martínez-Jerez, F.A., Tufano, P., 2012. Bouncing out of the bankingsystem: an empirical analysis of involuntary bank account closures. Journal ofBanking and Finance 36, 1224–1235.

Carbó-Valverde, S., Liñares-Zegarra, J.M., 2011. How effective are rewards programsin promoting payment card usage? Empirical evidence. Journal of Banking andFinance 35, 3275–3291.

CGM, 2012. Centrum voor de Geschiedenis van Migranten. <http://www.vijfeeuwenmigratie.nl>.

Ching, A.T., Hayashi, F., 2010. Payment card rewards programs and consumerpayment choice. Journal of Banking and Finance 34, 1773–1787.

EIM, 2007. Het toonbankbetalingsverkeer in Nederland: Kosten en opbrengsten vantoonbankinstellingen in kaart gebracht (Point-of-sale payments in theNetherlands: Costs and revenues of merchants). Final Report, Zoetermeer.

Fischer, A.M., 2010. Immigration and Large Banknotes. Working Paper, SwissNational Bank.

Hayashi, F., Klee, E., 2003. Technology adoption and consumer payments: evidencefrom survey data. Review of Network Economics 2, 175–190.

Hoofdbedrijfschap Detailhandel, 2011. Monitor betalingsverkeer 2011 (Paymentsystem monitor 2011). Final Report, Den Haag.

Humphrey, D.B., Pully, L.B., Vesala, J.M., 1996. Cash, paper and electronic payments:a cross-country analysis. Journal of Money, Credit and Banking 28, 914–939.

Jankowski, C., Porter, R.D., Rice, T., 2007. Against the Tide: Currency use AmongLatin American Immigrants in Chicago. Federal Reserve Bank of ChicagoEconomic Perspectives QII, 2–21.

Jonker, N., 2007. Payment instruments as perceived by consumers: results from ahousehold survey. De Economist 155, 271–303.

Jonker, N., Kosse, A., 2009. The Impact of Survey Design on Research Outcomes: ACase Study of Seven Pilots Measuring Cash Usage in the Netherlands. WorkingPaper, De Nederlandsche Bank.

Jonker, N., Kosse, A., Hernández, L., 2012. Cash Usage in the Netherlands: HowMuch, Where, When, Who and Whenever One Wants? Occasional Studies, DeNederlandsche Bank.

Von Kalckreuth, U., Schmidt, T., Stix, H., 2009. Choosing and Using PaymentInstruments: Evidence from German Microdata. Discussion Paper, Bundesbank.

Kennickell, A.B., Kwast, M.L., 1997. Who uses electronic banking? Results from the1995 – survey of consumer finances. In: Proceedings from the 33rd AnnualConference on Bank Structure and Competition, Federal Reserve Bank ofChicago.

Klee, E., 2008. How people pay: evidence from grocery store data. Journal ofMonetary Economics 55, 526–541.

Kok, S., Bosch, N., Deelen, A., Euwals, R., 2011. Migrant Women on the LabourMarket: On the Role of Home- and Host-Country Participation. DiscussionPaper, Centraal Planbureau.

McKinsey&Company, 2006. Betalingsverkeer in Nederland: Een onderzoek naar deopbrengsten en kosten voor het bankwezen (Payment Systems in theNetherlands: A Study into Revenues and Costs for the Banking Sector). FinalReport, Amsterdam.

Osili, U.O., Paulson, A.L., 2008. What can we learn about financial access from USimmigrants? The role of country of origin institutions and immigrant beliefs.World Bank Economic Review 22, 431–455.

Osili, U.O., Paulson, A.L., 2009. Immigrants’ access to financial services & assetaccumulation. In: Bland, R.M., Barr, M.S. (Eds.), Insufficient Funds. Russell SageFoundation, New York.

Rysman, M., 2007. An empirical analysis of payment card usage. Journal ofIndustrial Economics 55, 1–36.

Schuh, S., Stavins, J., 2010. Why are (some) consumers (finally) writing fewerchecks? The role of payment characteristics. Journal of Banking and Finance 34,1745–1758.

Seto, S., Bogan, V.L., forthcoming. Immigrant household investment behaviour andcountry of origin: A study of immigrants to the United States. InternationalJournal of Finance and Economics.

Stavins, J., 2001. Effect of consumer characteristics on the use of paymentinstruments. New England Economic Review 3, 19–31.

World Bank, 2008. Payment Systems Worldwide: A Snapshot. Outcomes of theGlobal Payment Systems Survey 2008. The World Bank. <http://siteresources.worldbank.org/INTPAYMENTREMMITTANCE/Resources/Global_Survey_Appendix.pdf>.

http://www.vijfeeuwenmigratie.nl

http://www.vijfeeuwenmigratie.nl

http://siteresources.worldbank.org/INTPAYMENTREMMITTANCE/Resources/Global_Survey_Appendix.pdf



choosing how to pay: the influence of foreign backgrounds

Documents