CHAPTER 1: Resource costs of cash and card payments in selected. European countries ... 2.3.1 The Costless but Costly Alternative: Cash p.42. 2.3.2 Optimal ...
TOWARDS A MORE EFFICIENT USE OF PAYMENT INSTRUMENTS
Paul De Grauwe Laura Rinaldi Patrick Van Cayseele
University of Leuven March 2006
Table of Contents
EXECUTIVE SUMMARY
p.3
CHAPTER 1: Resource costs of cash and card payments in selected European countries
p.5
1.1
Introduction
p.5
1.2
Comparing the resource costs of cash versus card payments
p.6
1.2.1 Card and cash usage in Belgium, the Netherlands and Iceland
p.6
1.2.2 Resource costs of cash and cards in Belgium and the Netherlands
p.8
1.3
Comparing the different types of cards
p.12
1.4
Debit cards and e-purses versus cash
p.14
1.5
The allocation of the resource costs among stakeholders
p.18
1.5.1 The resource costs of merchants
p.18
1.5.2 The resource costs of issuers and acquirers
p.19
1.5.3 The resource costs of the central banks
p.21
1.6
Analysis of fixed and variable costs
p.22
1.6.1 Fixed and variable costs
p.22
1.6.2 An analysis of marginal and variable costs
p.23
1.7
Resource costs of cash from a European perspective
p.26
Box A:
Fixed costs and economies of scale
p.30
Appendix 1: Methodological issues
p.32
Appendix 2: Average cost analysis
p.33
Appendix 3: Belgian Costs’ Studies
p.35
CHAPTER 2: Implications of Efficient Pricing of Cash and Debit Cards
p.37
2.1
p.37
Introduction
1
2.2
Some Stylized Facts
p.38
2.3
The Optimal Pricing Structure for Payment Instruments
p.42
2.3.1 The Costless but Costly Alternative: Cash
p.42
2.3.2 Optimal pricing of cash versus cards
p.45
Box B: Two-Sided Markets 2.3.3 Card and Cash Pricing in Reality
2.4
Card Penetration and Pricing
p.47 p.52
p.55
2.4.1 Product Differentiation and Discrete Choice Theory
p.56
2.4.2 Explaining Market Shares
p.57
2.4.3 Empirical Results
p.58
2.4.4 Cost-based pricing: Simulation results
p.62
2.4.5 Cost-based pricing and resource costs
p.63
Box C: European-wide benefits from implementation of efficient pricing
p.67
2.5
p.70
Conclusion
CHAPTER 3: The Shadow Economy and Cash Usage
p.72
3.1
Introduction
p.72
3.2
What is the shadow economy
p.73
3.3
How to measure the shadow economy
p.74
3.4
The theoretical model
p.79
3.5
An empirical application of the model
p.82
3.6
Reduction of the shadow economy and impact on card usage
p.85
3.7
Shadow economy and costs of payment instruments
p.86
Box D: Resource costs at European level
p.89
3.8
p.91
Concluding remarks
REFERENCES LIST
p.93
2
EXECUTIVE SUMMARY
The aim of this study is to analyse the resource costs of cash and debit card payment systems in Europe and to suggest ways to use these in a more cost-efficient way. For this purpose we use the existing evidence on resource costs at the country level in order to identify areas for possible improvement in the provision of payment instruments at the European level. In particular we focus on the role of pricing as signals of resource costs for consumers and on the connection between the shadow economy and cash usage. The first chapter of this report focuses on the resource costs of cash and cards. We use existing studies of Belgium and the Netherlands and more sporadic evidence about other European countries to measure these resource costs. Our main findings are that the use of cash in the payments system absorbs significantly more resources than the use of debit cards. When we allocate these costs among the different stakeholders we find that the resource costs of the use of cash borne by the merchants in general exceed the resource costs of the use of debit cards. We also analyze the fixed and variable cost components in the use of cash and cards. The absence of transparent and cost based pricing in the provision of payment services, and in particular the (almost) free supply of cash to the consumer, leads to important inefficiencies. In the second chapter of this report we first analyze the nature of these efficiencies from a theoretical point of view and we argue that the pricing of cash and cards should be set by taking into account their cost. We then go on estimating an econometric model that explains the market shares of debit cards and cash. The results of this econometric analysis are that consumers react in a significant way to changes in the cost of cash and cards. In particular, an increase in the price of cards leads to a significant decline in its usage. Similarly an increase in the price of cash (as paid by the consumer) leads to a significant decline in its usage. We use the results of this econometric analysis to compute the effects a cost-based pricing system would have on the use of cards and cash in the EU-countries. We find that these effects are quite large, i.e. a switch to cost-based pricing would increase the market share of debit cards from its current level of 4 percent to 19 percent.
3
Finally, we use these estimates of the effects of cost-based pricing to analyse the question of how cost based pricing would affect the resource costs of the payment services provided by cards and cash. Due to a lack of data we could make this analysis for Belgium and the Netherlands only. We find that the introduction of cost-based pricing would lead to a reduction of resource costs of €150 to €200 million in these countries. We also use the average of the Belgian and Dutch costs to extrapolate the cost reduction that would follow the implementation of cost-based pricing in a sample of nineteen European countries. We estimate that on average for the countries in our sample the resource cost reduction would be equal to 0.14 percent of GDP. In the last chapter we analyse the nature and size of the shadow economy in various European countries and its connection with cash usage. We note that the shadow economy is quite large even in developed countries. It is motivated by the high tax level and social security burden, but also by people’s feelings about civic ethics. These findings are very important to set the basis for political action aimed at combating the shadow economy. Given that most activities in the shadow economy are cash-based, it follows that a reduction of the shadow economy is likely to shift an important share of transactions from cash to cards. We test this hypothesis empirically and find strong evidence in favour of the cash-card substitution motivated by hidden activities. On the basis of this result, we estimate the possible impact of a reduction of the shadow economy on the resource costs of payment instruments. We find that in the event of a reduction of the shadow economy by half the current level, the resource costs of cash and cards would be reduced by €40 and €52 million in Belgium and the Netherlands, respectively. Assuming the same policy reduction in the hidden activity and a similar cost structure in a sample of thirteen European countries the reduction in resource costs would be equal to about 0.1 percent of their GDP. Our results clearly show that the shadow economy is intrinsically connected with cash usage. Put differently, hidden economic activities exist mainly because of the anonymity provided by cash. It follows that political action aimed at fighting the shadow economy must operate in the framework of an overall political package that modifies the tax collection systems and that provides incentives to use cards and other alternatives to cash for transactions. Clearly, the main motivation for a reduction in the size of the shadow economy is that it increases government revenues; however, other important benefits are likely to follow. A more efficient and less costly payment system is one of these.
4
CHAPTER 1 Resource costs of cash and card payments in selected European countries
1.1
Introduction
In the last two decades the use of bank-cards has increased very fast leading to a reduction in the demand for currency. Despite this development there is still a significant scope for substitution between cards and cash. Such a substitution is likely to lead to benefits for society as a whole, as it makes the payments system more efficient and thereby stimulates economic activity. Reliable payment systems, however, imply large costs for society so that they should not only be well functioning, but also economically efficient. In order to asses the efficiency of the different payment services it is necessary to identify the resource costs involved, i.e. the amount of resources needed to provide the different payment services. Several studies have estimated the costs and revenues of the cash and card systems and their allocation among the service’s users and providers. Despite the variety of results, these studies conclude that the card system is more cost efficient than the cash-based system. These studies, however, are applied to different countries; they are based on different assumptions and they are carried out using different methodologies. As a result, they are not always directly comparable. Moreover only a few of them gather information about the resource costs involved in the whole payment cycle. Most of these studies focus on the costs borne by only one or a few stakeholders in the payment system.
5
In this study we collect and interpret information on the total resource costs of payment services based on cards and cash in a number of European countries1. We start by analysing the countries for which we have the most complete information and discuss the results from the perspective of the whole society by focusing on the differences and similarities among them. Then we move a step further to analyse the different stakeholders involved in payment services for which detailed information exists to complete and reinforce the general picture.
1.2
Comparing the resource costs of cash versus card payments
The countries for which detailed studies on the resource costs of retail payment services are available are Belgium, Iceland and the Netherlands. We first present a brief overview of the cash and card usage in these three countries, and we then present the evidence about the resource costs in these countries.
1.2.1
Card and cash usage in Belgium, the Netherlands and Iceland
Belgium and the Netherlands appear to be at a similar stage of development of the retail payment system. In these two countries card payments are widely used when compared with the average for the euro area countries, as shown in Figure 1.1. While the card use (both debit and credit) in Belgium and in the Netherlands is comparable, preferences as to the kind of cards differ somewhat (see Figure 1.1). Dutch consumers appear to have a stronger preference for debit cards than Belgian consumers, while the latter use credit cards and e-purses more frequently. Iceland appears to be in a different, more advanced phase in the use of payment services. Figure 1.1 shows that Icelanders use their debit and credit cards much more intensely than Belgium and the Netherlands. In addition, the value of debit/credit card transactions is 4 to 5 times higher than in Belgium and the Netherlands. This higher value is motivated by the fact that in Iceland cards are also used as substitutes of bank transfers (as such they are used to pay bills, taxes, etc.). Note, however, that while in most countries the banking systems are trying 1
Note that what we call resource costs can also be defined as social cost of payment services. In fact the social cost is defined as the sum of the resource costs of the different stakeholders involved in the payment circle. Therefore the two terms could be used interchangeably, however throughout the study we will refer to them as resource costs.
6
to replace small value cash transactions by e-purses, in Iceland an electronic-purse scheme has never been implemented. Iceland is different from Belgium and the Netherlands for another reason: it is very small. With a population of less than 300.000 it is even smaller than Luxembourg. As a result, it is probably not able to exploit economies of scale that exist in the payment systems in the same way as Belgium and the Netherlands that are 30 to 50 times bigger than Iceland. This is likely to lead to a very different cost structure of the use of cash and cards in Iceland as compared to Belgium and the Netherlands as well as other European countries. Therefore we will present and discuss the results for Belgium and the Netherlands in the following section and leave aside the Icelandic case2. Figure 1.1: Card usage euros
no.
70
4000
70
60
3500
60
no.
Netherlands
3000
50
2500
40
euros
Belgium
4000 3500 3000
50
2500
40
2000
2000 30
30
1500
1500
20
1000
10
500
0
0 1998
1999
no.
2000
2001
20
1000
10
500
0
2002
0 1998
1999
euros
Euro area
70
4000
60
3500 3000
50
2500
40
2000
2001
2002
euros
Iceland
150
14000 12000
120 10000 90
8000
2000 30
1500
20
1000
10
6000
60
4000 30 2000
500
0
0 1998
1999
2000
2001
2002
0
0 1998
1999
2000
debit, no.trans per inhabitant
credit, no.trans per inhabitant
debit, no.trans per inhabitant
e-money, no. trans per inhabitant
value of debit/credit transactions per inhab
value of debit/credit transactions per inhab. (euro)
2001
2002
2003
credit, no.trans per inhabitant
2
It should be mentioned that the difference in size between Belgium and the Netherlands on the one hand and Iceland on the other is much higher than the difference in size between the former countries and large EUcountries such as France and Germany. The latter countries are only five to eight times larger than Belgium and the Netherlands. It is very likely that Belgium and the Netherlands exploit economies of scale about as much as France and Germany.
7
1.2.2
Resource costs of cash and cards in Belgium and the Netherlands
After this first glance at card usage, we analyse the costs involved in running payment services in Belgium and the Netherlands. We first compare cash and cards. At this stage we aggregate all cards (debit, credit, e-purse) without making a distinction among them. This will give us a first broad picture of the difference in the cost efficiency of the traditional instrument, cash, versus more advanced alternatives, cards. At a later stage we will make a distinction between the different types of cards and we will see that there are quite large differences between them from a cost-efficiency perspective. We collected information on the same measures of resource costs of cash and cards in the two countries (in appendix we provide more information about data sources and methodologies employed). Note that the data do not refer to the same year: for Belgium the data refer to 1998 and for the Netherlands to 20023. To ease comparability all monetary amounts are expressed in euros. We will use different measures of the resource costs: •
The resource costs as a percent of GDP
•
The resource cost per capita (per head of the population)
•
The resource cost per transaction
•
The resource cost as a percent of transaction value
The first two measures give an indication of the aggregate resource costs. They are influenced by the size of the payment system, as will become clear from our analysis. The last two measures can be considered as average costs. They can, however, give very different results because they are influenced by two types of economies of scale in the provision of payment services. First there are economies of scale that arise when the number of transactions increases. Since there are fixed costs in the provision of payment services, the higher is the number of transactions the lower will be the cost per transaction. Typically, the number of transactions using cash is still significantly higher than the number of transactions using cards. As a result, cash profits more than cards from these economies of scale. Second, there are economies of scale related to the value of the transactions. Given that the cost of making an individual payment is typically unrelated to the value of the transaction, the 3
See Appendix 1 for more detail on data sources.
8
cost of the payment declines as a percent of the value of the payment. From figure 1.2 we observe that the average value of card transactions is much higher than the average value of cash transactions. As a result, cards benefit from this second type of economies of scale much more than cash. We will see that these different economies of scale have an important impact on our measures of average cost. (In box A we analyze the relation between these two types of economies of scale in more detail).
Figure 1.2: Average transaction value (euro)
credit
debit
e-purse
cash 0
20
40
60 Belgium
80
Netherlands
100
120
euro
Note: Data for cards refer to 2002; for cash in Belgium they refer to 1998 and in the Netherlands to 2002.
As our first indicator, we show in Figure 1.3 the resource cost as a percent of GDP. This indicator measures the overall resources that are necessary to provide payment services in the different countries. To make these comparable across countries we divide the total resource costs by GDP. We find very strikingly that the total resource costs of the use of cash are between 3 to 6 times higher than the resources costs of the card payment systems. In Belgium the cash system requires about 6 times more resources than the card system, while in the Netherlands almost 3 times more. The same pattern holds when one takes per capita resource costs of payment services (see figure 1.4). The cash system on average costs a Belgian and a Dutch person respectively 164
9
and 131 euros, while the resources cost of cards per capita is only 23 and 47 euros in Belgium and the Netherlands respectively. Note also that the resource costs of cash are higher in Belgium than in the Netherlands (this can be seen in both figure 1.3 and 1.4), while the opposite holds for cards: resource costs of the Belgian card system appear to be lower in Belgium than in the Netherlands. One of the reasons of this difference is the different methodology used. We discuss this in appendix 1. Figure 1.3: Cost as share of GDP percent 0.8 0.7
0.74
0.6 0.5
0.48
0.4 0.3 0.2
0.17
0.11
0.1 0
Belgium
Netherlands Cash
Cards
Figure 1.4: Resource cost per capita
euro 180 160 140 120
164
131
100 80 60
47
40
23
20 0
Belgium
Netherlands
Cash
Cards
10
Next we consider our measures of average costs. We first discuss average costs expressed as the cost per transaction. We show the results in Figure 1.5. It can be seen that in Belgium a cash transaction costs on average 56 eurocents while a card transaction costs on average 64 eurocents. In the Netherlands these costs are 30 and 64 eurocents respectively. Thus average costs per transaction are lower for cash than for cards in both countries, although the difference is more pronounced in the Netherlands than in Belgium. As mentioned earlier this difference is due to the fact that the number of transactions using cash is still significantly higher than the number of transactions using cards. As a result, cash profits more from this type of economies of scale than cards. Our second measure of the average resource costs of the different payment systems is the cost as a percentage of the transaction value. We show the results in figure 1.6. The results are quite interesting. First, we find that when expressed as a percent of the transaction value, cash payments are clearly more expensive than cards in both Belgium and the Netherlands. This difference is related to the fact that cash use is mostly confined to small value transactions see figure 1.8), and as a result does not profit from the second type of economies of scale, as cards do. Second, we observe that in Belgium cash transactions are almost three times as expensive as in the Netherlands when costs are expressed in terms of the transaction value. Third, the cost of card transactions as a percent of transaction value is of the same order of magnitude in the two countries.
Figure 1.5: Resource cost per transaction
0.7 0.6
euro
0.4
0.64
0.64
0.5 0.56
0.3
0.3
0.2 0.1 0
Belgium
Netherlands Cash
Cards
11
percent
Figure 1.6: Resource cost as a percentage of transaction value
10 9 8 7 6 5 4 3 2 1 0
9.03
3.2 1.45
1.19 Belgium
Netherlands
Cash
Cards
Summarizing the previous results one can conclude, first, that the cash payment systems in Belgium and the Netherlands use significantly more resources than the card payment systems. Second the cash payment system appears to be cost efficient for small value payments, while the card payment system is much more cost efficient for large value payment. We will investigate this issue further in section 1.6 when we analyse the fixed costs and variable costs of the two systems.
1.3
Comparing the different types of cards
Different card schemes have very different costs structures and different rates of usage. Thus, in order to get a clearer picture of the relative cost-efficiency of retail payment services we have to move a step forward in the analysis of resource costs. We now concentrate our attention on the card system and study how the costs related to the functioning of the different card schemes are distributed across types of card. Figure 1.7 shows the average costs per transaction. We find that these average costs are much lower for debit than for credit cards. This is undoubtedly related to the wider use of debit card payment systems in both countries allowing that system to better exploit economies of scale. This is probably not the only source of the difference in average costs. It could be that the debit card system is just more cost efficient for any level of use than credit card systems.
12
Note also that in both countries the average cost per transaction of the use of e-purse is of a similar magnitude as the average costs of debit cards, although in the Netherlands the average costs of e-purses appear to be significantly higher than of cash. The average costs as a percent of the transaction values are shown in figure 1.8. Concentrating on the difference between debit and credit cards we find again that the average costs of credit cards are significantly higher than those of debit cards. We also note that the transaction costs of e-purses expressed as a percent of transaction value are extremely high. This has of course to do with the fact that e-purses are used for low value payments (see figure 1.2). The differences, however, are so high that they suggest that the technology may not be costefficient.
Figure 1.7: Average resource cost per transaction
4 3.5
3.6
3
euro
2.5
2.7
2
0.9
1.5 1
0.5
0.4
0.3
0.5 0 Belgium
Netherlands
Debit
Credit
e-Purse
13
Figure 1.8: Average cost as % of transaction value
35
34.32
30
percent
25 20 15
7.02
10 5
0.83
3.11
1.1
2.11
0 Belgium
Netherlands
Debit
1.4
Credit
e-Purse
Debit cards and e-purses versus cash
In this section we compare the resource cost of debit cards with cash. We do this because debit cards come closest to cash as a payment instrument, as it typically does not have a credit (deferred debit) component (in contrast with credit cards). Moreover, credit cards also appear to be used for much larger transactions: in Belgium and in the Netherlands the average transaction value of a credit card is more than twice as large as a debit card transaction. This makes the comparison of the resource costs of credit cards with cash more problematic. We showed the differences in transaction values in figure 1.2. Therefore as far as cards are concerned from now on we focus on debit and e-purse cards, i.e. those cards with the higher potential to substitute cash in retail transactions. In figures 1.9 and 1.10 we present the resource costs of cash, debit cards and e-purse as a share of GDP and on a per capita basis. From the previous analysis it is not surprising that the production and circulation of notes and coins requires more resources than debit cards. Comparing figures 1.9 and 1.10 with figures 1.3 and 1.4, where the cost of cash was compared with the cost of cards all together, we observe that debit cards in Belgium and the Netherlands require fewer resources than cash. In Belgium the difference between the cost of
14
cash and debit cards is particularly pronounced. Clearly, the cost of e-purses is very small because it relates only to a small fraction of total retail transactions.
Figure 1.9: Resource costs as percentage GDP
percent 0.8 0.7
0.74
0.6 0.5
0.48
0.4 0.3
0.12
0.2
0.05
0.1
0.018
0.004
0
Belgium
Netherlands Cash
Debit card
e-purse
Figure 1.10: Resource costs per capita
euro 180 160 140
164 131
120 100 80 60 40
32 12
20 0
5
0.78
Belgium
Netherlands Cash
Debit card
e-purse
15
The average cost per transaction is shown in Figure 1.11. The average cost of a debit card transaction is comparable in both countries, amounting to 0.4 and 0.49 eurocents respectively. In contrast, as we showed earlier, the average cost of cash is significantly higher in Belgium than in the Netherlands. This is also due to the fact that in Belgium, according to the available information, there are proportionally less cash transactions. It is striking that the relationship between the three different payment services in Belgium and in the Netherlands appears inverted. While Figure 1.11 suggests that in the Netherlands cash is managed in a more efficient way, an e-purse transaction is three times more costly than in Belgium. This is indeed very expensive: given that the average e-purse transaction value in the Netherlands is euros 2.72, one e-purse transaction costs society almost one euro!
Figure 1.11: Average cost per transaction euro 1 0.9
0.93
0.8 0.7 0.6 0.5 0.4
0.56
0.49 0.4
0.3
0.28
0.2
0.3
0.1 0
Belgium Cash
Netherlands Debit card
e-purse
Finally figure 1.12 presents the resource cost as a percent of the average transaction value. We find that debit cards are by far the most cost-efficient payment service when we use this indicator of average costs. The resource costs of debit cards are about 1 percent of the transaction value in the three countries. The resource costs of cash as a percent of transaction values are a multiple, 9 percent in Belgium and 3 percent in the Netherlands.
16
As far as e-purses are concerned, the evidence suggests that these schemes fall short of the cost efficiency achieved with debit cards. It is even not clear how potentially efficient they can be in replacing cash in small value transactions. In particular, the evidence for Belgium contrasts with the evidence for the Netherlands. While in Belgium e-purses are cheaper than cash both per transaction and as a share of the transaction value, in the Netherlands cash is much cheaper than e-purses (and to some extent also cheaper with respect to debit cards). This lack of cost efficiency may explain why the use of e-purses has remained very limited in the Netherlands.
Figure 1.12: Resource costs as percentage of transaction value percent 35 34.3
30 25 20 15 10 5 0
9.0
7.0 0.8
3.2
Belgium Cash
1.1
Netherlands Debit card
e-purse
17
1.5
The allocation of the resource costs among stakeholders
In the previous sections we presented and discussed the aggregate resource costs borne by all stakeholders in the different payment systems. In this section we disaggregate these resource costs and we allocate them to the different stakeholders. These stakeholders are •
The merchants
•
The issuers and acquirers
•
The central bank
It should be stressed that since we analyse the resource costs, i.e. the amount of resources needed to provide the payment services by the different stakeholders, we do not include the transfers between these stakeholders (e.g. the interchange fee). We present and discuss the results in the following subsections.
1.5.1
The resource costs of merchants
The resource costs of merchants are presented in figures 1.13 and 1.14. We concentrate on measures of average cost (euro cost per transaction and cost as percent of transaction value). The results here are very similar to the ones obtained for the aggregate resource costs. The resource costs per transaction incurred by merchants are of a similar order of magnitude for cash and cards. In Belgium the use of cash is more expensive than the use of debit cards for merchants, while the opposite holds in the Netherlands4. As far as the average cost as a percent of transaction value is concerned, we find that cash is more resource intensive for the merchants in both Belgium and the Netherlands. This is in a way surprising since the perception of many merchants is that the use of cards, including debit cards, is more costly than the use of cash. This perception has much to do with the fact that merchants, especially the small ones, appear not to consider the labour time used to count and manipulate cash as part of their operating costs.
4
The difference between the two countries, however, could be due to methodological and institutional differences. For instance, certain cash operations might be taken care of by the banks in the Netherlands while they might be performed by merchants in Belgium.
18
Figure 1.13: Merchants’ average cost per transaction eurocents 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
CASH
DEBIT
Belgium
E-PURSE
Nethelands
Figure 1.14: Merchants’ cost as % of transaction value percent 7 6 5 4 3 2 1 0
CASH
DEBIT Belgium
1.5.2
E-PURSE Nethelands
The resource costs of issuers and acquirers
The resource costs of issuers and acquirers are shown in figures 1.15 and 1.16. From these figures the following observations can be made. In Belgium and the Netherlands the cash resource cost per transaction borne by the issuers and acquirers (which include the banks who 19
are active in the processing and distribution of cash) is lower than the debit card cost. As before this result is influenced by the fact that the amount of cash transactions by far exceeds the amount of debit card transactions. When we compare the costs as a percent of the transaction value we obtain the opposite. The cost of cash as a percent of the transaction value is typically more than double the cost of debit cards in Belgium and the Netherlands. Finally, we note again the very high cost of the e-purse as borne by the issuers and acquirers especially in the Netherlands.
Figure 1.15: Issuers and acquirers’ average cost per transaction euro 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
CASH
DEBIT
Belgium
E-PURSE
Nethelands
Figure 1.16: Issuers and acquirers’ cost as % of transaction value percent
10 28.8 8 6 4 2 0
CASH
DEBIT Belgium
E-PURSE Nethelands
20
1.5.3
The resource costs of the central banks
Finally, in figure 1.17 we show the resource cost borne by the central bank in producing, processing and distributing cash5. For Belgium and the Netherlands we find very low resource costs per transaction borne by their central banks. It should also be stressed that we only look at the cost side. The counterpart of the costs borne by the central banks is the large revenues that result from its monopoly in the provision of cash services. As a result, the supply of cash by the central bank is very profitable. This contrasts with the commercial banking system for which the supply of cash services is typically a loss making activity.
Figure 1.17: Central Bank’s average cost per transaction
eurocents 0.02 0.015 0.01 0.005 0
5
Belgium
Nethelands
No information is given on cards since the central banks are not supplying card services.
21
1.6
Analysis of fixed and variable costs
In the previous sections we discussed various measures of total resource costs of payments. In this section we look at the nature of such costs. Similar to the methodology used by the Dutch National Bank (2004) we distinguish between three main types of costs: fixed costs, variable costs related to the number of transactions and variable costs related to the value of the transaction. A comparison of these costs is important when assessing the cost-efficiency of the different payment services. In fact the possibility of exploiting the different economies of scale depends on the relative size of fixed and variable costs. 1.6.1
Fixed and variable costs
We first discuss the distribution of fixed and variable costs in the supply of cash in Belgium and the Netherlands. The results are shown in figure 1.18. It can be seen that the cost structure for cash is very similar in both countries. The fixed costs make up the largest part (42 to 44%); variable costs dependent on the number of transaction represent 37-38%, while variable costs that depend on the transaction value make up about 20%. The structure of fixed versus variable costs of the supply of debit cards differs significantly between the two countries as can be seen from figure 1.19. In particular the fixed cost component appears to be significantly higher in the Netherlands than in Belgium. We have not been able to find out what the source of this difference is. The difference in cost structure of the supply of e-purse is even more pronounced as can be seen from figure 1.20. We observe that in the Netherlands the share of fixed cost is estimated to be close to 100%. Our estimates for Belgium are much lower.
Figure 1.18: Distribution of fixed and variable costs in the supply of cash Cash - Belgium
Cash - Netherlands
18%
21% 42%
44%
38% Fixed
37% Variable-transactions
Variable-sales
Fixed
Variable-transactions
Variable-sales
22
Figure 1.19: Distribution of fixed and variable costs in the supply of debit cards Debit card - Belgium
Debit card - Netherlands 1%
13% 39%
39%
60%
48% Fixed
Variable-transactions
Variable-sales
Fixed
Variable-transactions
Variable-sales
Figure 1.20: Distribution of fixed and variable costs in the supply of e-purse e-Purse - Netherlands
e-Purse - Belgium 2%
4%
0%
41%
57% Fixed
1.6.2
Variable-transactions
Variable-sales
Fixed
96% Variable-transactions
Variable-sales
An analysis of marginal and variable costs
In this section we present an analysis of the marginal and variable costs of the different payment instruments in Belgium and the Netherlands. The way these costs were computed is explained in appendix 2. We use the concept of variable costs as a function of transaction value. This is defined as VCi = ai + bi s where VCi is the variable cost of payment instrument i (e.g. cash, or debit card), ai is the marginal cost associated with one additional transaction using instrument i , bi is the marginal cost associated with one additional euro value of transaction, s is the value of the transaction. Note that we use two concepts of marginal costs. The first one is the additional cost when the
23
number of transactions increases by one unit; the second one is the additional cost when the value of the transaction increases. Both marginal costs are assumed to be constant. We show the results of our calculations for Belgium and the Netherlands in figures 1.21 and 1.22. The intercept of each curve with the y-axis represents the marginal costs with respect to number of transactions (ai). The slope of each curve represents the marginal costs with respect to transaction value (bi). Figures 1.21 and 1.22 allow us to determine the value ranges within which particular payment instruments are more or less cost efficient than the others. Let us concentrate on Belgium first. From figure 1.21 we conclude that debit cards are more cost efficient than cash for all transaction values, while e-purses are more cost efficient than cash and debit cards for all transaction values. There are reasons to believe that our estimate of the variable costs of cash is biased upwards for Belgium because of the important difference with respect to the recent estimations made by the National Bank of Belgium (NBB) for the cost of cash. In order to find out how large this bias could be we show also the variable costs of cash in Belgium as estimated by the NBB (2005). This estimate is represented in figure 1.21 by the line Cash-NBB (dashed line). We now find that for transaction values below €8.7 cash is more cost efficient than debit cards. For transaction values lower than €4.5 cash is more cost efficient than e-purses. Most probably the true variable cost line for cash in Belgium lies somewhere in between the two lines shown in figure 1.21, such that they can be considered as two extremes within which the cost of cash is likely to vary. The results for the Netherlands are presented in figure 1.22. We find that cash is cost efficient for transaction values below €12.4 which makes cash significantly more cost efficient in the Netherlands than in Belgium, whatever the cash estimates considered. In addition, it appears that e-purses are more cost efficient for all transaction values. The analysis of figures 1.21 and 1.22 does not take into account the existence of fixed costs. This is not a problem when the analysis is confined to the short run, during which these fixed costs do not change. For a longer term analysis fixed costs are important.
24
Figure 1.21: Variable costs of different payment instruments (Belgium) 0.5
variable cost (eurocents)
0.4
Cash
Debit
E-purse
5
10
15
Cash-NBB
0.3
0.2
0.1
0 0
20 25 transaction value (s)
Figure 1.22: Variable costs of different payment instruments (Netherlands)
0.5
variable cost (eurocents)
Cash
Debit
E-purse
Credit
0.4
0.3
0.2
0.1
0 0
5
10
15
20
25 30 transaction value (s)
25
1.7
Resource costs of cash from a European perspective
In the previous sections we analysed the resource costs of cash as opposed to cards in three European countries for which detailed studies are available. Even if the evidence from the three countries constitutes a useful reference also for other countries, it is well known that the use of the different payment services as well as the costs and revenues of the different stakeholders involved in payments systems varies widely across the European Union (and the Euro area). In the euro area there is now a common currency in circulation, but the organisation of the production and the provision of cash is still country specific. Thus in the future we may expect a common set of guidelines in order to make cash services more uniform across euro area countries. It would be particularly useful, therefore, to have a general estimation of the resource costs at the level of the euro area. Detailed studies on the resource costs and benefits of cash at European level are not publicly available. However a study by Edgar Dunn & Company has estimated the costs and revenues of ATMs across Europe. The study reveals that ATM withdrawals cost society 0.74 euro cents per transaction with an average withdrawal value between 96 and 108 euro. This estimate of the cost per transaction is quite high when compared to Belgium and Iceland where the resource cost of an ATM transaction is 0.31 and 0.5 eurocents respectively. It is difficult to know why the use of ATMs is so much cheaper in these two countries with respect to the average for European countries. The explanation could be related to the fact that these estimates do not take into account the opportunity cost of holding cash for consumers –while the European estimate does-, hence they constitute an underestimation of real costs. It could also be argues that in small and densely populated countries it may be less costly to maintain and refill ATMs. Moreover, small countries often have a unique and common ATM network. Finally, the explanation could simply be related to a more efficient management. If only the costs of production are considered, i.e. issuers and acquirers’ costs, the cost of ATMs at European level becomes 0.45 euro cents per transaction, which is very close to the equivalent cost estimated for the UK (0.41 eurocents in 1994)6. In any case the acquisition of notes and coins via ATMs is by far more efficient than via withdrawals at the bank-counter which, according to the Edgar Dunn & Company study, costs on average 3.6 euro per transaction. Thus, ATMs allow the banking sector to reduce the costs 6
APACS (1996).
26
related to cash acquisition by consumers and, therefore, to cut the total costs of cash for the whole society. It should be noted that the total costs related to cash acquisition (via ATMs or withdrawals at the counter) constitute only a share of the total resource costs of cash. The share of ATM costs out of total resource costs of cash is equal to 3% in Belgium and 10% in Iceland. Thus this share can differ considerably from one country to another. As a result, it is not possible from the ATM costs to draw a conclusion concerning the total resource costs of cash at European level. The organisation of cash centres as well as the organisation of cash transportation and distribution can be quite different in different countries. The only possible extrapolation we could make would be on the basis of different scenarios, but this would require too many special assumptions making these scenarios very speculative. The only estimates of the aggregate resource cost of cash that exist at multi-country level are the estimates reported in a study carried out by the European Payments Council (EPC) for which, however, we do not dispose of any methodological detail. According to the EPC the total resource cost of cash in the European Union7 amounts to EUR 50 billion, i.e. between 0.4 and 0.6 percent of total GDP. This share of GDP is quite in line with the corresponding estimates for Belgium, Iceland and the Netherlands (0.7, 0.3 and 0.5 percent respectively). However when we take a measure of average costs, which therefore takes into account of the intensity of cash use, the numbers differ widely. According to the EPC estimates, in Europe on average the cost of a cash transaction is 0.15 euro cents. This number is based on an estimated number of cash transactions of 360 billion. This number, however, seems very large: it corresponds to 2.5 cash transaction per day per person as opposed to the 0.8 and 1.2 of Belgium and the Netherlands, respectively. A study carried out in 1994 by APACS for the UK payment system estimated that the number of cash transaction amounts to about 32 billions, which means 1.5 transactions per person per day. A similar study for France referring to the beginning the Nineties also estimated about one cash transaction per person per day8. Therefore from the existing evidence the estimated number of cash transactions put forward by EPC seems to be an overestimation of the actual number. This would explain why the ECP estimations of the resource costs of a cash transaction at European level are so much lower than the estimates we have for different EU-countries. Of course estimating the actual number of cash transactions including person-to-person payments is certainly very difficult. 7 8
Here by European Union is meant EU-15 given that the study refers to 2003. De La Rue, (1993), “The costs of handling Money” European Banking Conference, London 7-8 October.
27
However the implications for the determination of resource costs of payment services are very important. To give an example, if we would assume that in Europe the average number of cash transaction per capita is the same as in the Netherlands9 we would obtain a resource cost of cash of 0.30 cents per transaction, which is in line with the single country estimates available. It is also interesting to compare the distribution of cash cost across stakeholders. As figure 1.25 shows, according to ECP the banking sector bears the largest share of total cash costs followed by the retail sector which bears one-forth of total costs while the central banks only bear 10 percent. As already mentioned we do not dispose of any explanation concerning the methodology used for these estimations, but we assume that the study neglects mutual transfers among stakeholders10. It is quite striking however to see the difference in the costs distribution in the European Union and in the three countries. In the countries the largest part of cash costs are held by the retail sector. However such a different distribution of cash costs across stakeholders could be due to methodological differences. In all the country studies the handling costs of cash for merchants is also included, even though this cost is not always directly visible to merchants. It could be that this cost is not considered in the EPC study and this would then explain the differences.
9
We take as reference the number of cash transactions in the Netherlands because the way it is estimated seem very complete, combining information coming from both merchants and consumers’ surveys (see Dutch National Bank (2004)). 10 This is suggested by the fact that in the EPC study central banks have net costs related to cash and consumers’ costs are not mentioned. This wouldn’t be possible if mutual transfers between stakeholders were also considered. However the total resource costs has to be the same independently of the methodology used.
28
Figure 1.23: Distribution of Cash Costs Belgium
European Union (EU15) 10%
3%
27% 25%
65% 70% Central Bank
Merchants
Banks
Central Bank
Netherlands
Merchants
Banks
Iceland
3%
4%
42%
40%
55%
Central Bank
Merchants
Banks
56%
Central Bank
Merchants
Banks
29
Box A: Fixed costs and economies of scale
In this box we analyze in a stylized way the different sources of economies of scale in the payment systems. We concentrate on the effect of fixed costs on average costs. For the sake of clarity in the analysis, we disregard variable costs. Figure B1 shows the average (fixed) cost on the vertical axis. We have normalized the fixed cost of running the payment system at 1. The x-axis shows the number of transactions; the y-axis shows the value of the transaction. We first concentrate on how the number of transactions affects the average costs. We do this by assuming that the value of the transaction is constant and equal to S1 (say 1 euro). The curve AB then shows how the average cost declines as the number of transactions increases for a given value S1. From each point on the AB curve we can now draw an average cost line moving parallel to the value-axis (y-axis). These curves then define the average cost as a function of the value of the transaction, given the number of transactions. We obtain a nonlinear surface that presents the average cost as a function of the number of transactions and of the value of transactions. We can now illustrate how different payment instruments profit from the two types of economies of scale. Cash profits from the fact that the number of transactions using cash is very large and much larger than cards. However, it is used mainly for low value payments. We have put the average cost of cash on the AB-curve. In contrast, cards are used in fewer transactions. As a result, they cannot exploit economies of scale resulting from a large number of transactions in the same way as cash. However, as cards are used in high value payments they profit from the second type of economies of scale. This is shown in figure B1 by the point “cards”. In figure B1 we have also added a curve through the “cards” point. Its intersection with the AB-line defines the cost per transaction when the transaction value is equal to S1, which is the transaction value assumed for cash. It can be seen that for such a low transaction value, cards are more costly than cash. Thus for low transaction values cash is more efficient than cards. We can find the exact transaction value that makes cards more efficient than cash as follows. The horizontal convex curves define the combinations of number of transactions and transaction value that lead to the same average cost. Thus through the “cash” point there is such a convex curve that leads to a constant average cost of 0.3. The intersection of this 30
convex curve with the red curve through the “cards” point defines the critical point. It corresponds to a transaction value equal to S6. Thus in our theoretical scheme of figure B1 cash will be more efficient than cards for transaction values of S6 or less. For transaction values higher than S6 cash is more efficient.
Figure B1 Average cost as a function of number of transactions and of transaction value
A 1 0,9 0,8 0,7 0,6 cash
0,5 cost
B
0,4 0,3 0,2
S1
0,1
S7 1
4
0 7
10
16
19
22
25
S25
28
S19
13
cards
S13
31
value
number of transactions
31
Appendix 1: Methodological issues In the present study we compare the resource costs of cash and card payments in different countries using two main information sources. The first source is the survey “The costs of payments” by the Dutch National Bank (DNB) (2004). The survey is the result of the activity of a working group organised by DNB and included representatives of the banking industry, Interpay11, merchants and consumers, which supplied the information necessary to carry on the study. The Dutch analysis refers to 2002. The second source is De Grauwe, Buyst and Rinaldi (DBR) (2000) which provides very detailed information about Belgian and Icelandic retail payment systems. The data reported in this study refer to 1998 for Belgium and 1997 for Iceland and were elaborated by the authors on the basis of information coming directly from the different stakeholders (central banks, card companies, banking sector, merchants’ representatives). The original DBR (2000) study did not make a complete distinction between the different types of cards active in the two countries analysed: it collected information for cash versus cards all together. Nevertheless, DBR (2000) study could be extended on the basis of the original data sources. In this report we present two types of extensions. First, the results for debit cards and e-purses are presented separately. Second, we make an analysis of fixed and variables costs of the different payment services. There are some important methodological differences between these two studies. In DBR (2000) for every stakeholder (central bank, issuers, acquirers, merchants and consumers) a complete list of costs and revenues related to each payment service is described and the estimated amounts reported together with the methodology used whenever possible. It is important to notice that that report provides full information on costs and revenues. Obviously in many cases the costs of one stakeholder constitute revenues for another stakeholder (notably in the case of fees), such that the resulting computation includes many cross-transfers which cancel out when taking net resource costs. Conversely, the DNB (2004) study does not consider transfers among stakeholders. Of course, even using this methodology the resulting resource costs are the same. However the resulting distribution of costs across stakeholders is profoundly different in this case, such that from this study it is not possible to make considerations about who bears the costs and who receives the benefits of payment services because only production costs are considered. It 11
Interpay provides the infrastructure for the processing and clearing of electronic payments.
32
follows that in DNB (2004) consumers do not appear among the stakeholders because everything they pay represents revenues for other stakeholders. In order to make the results for Belgium and the Netherlands as more comparable as possible, in the present study we follow the Dutch methodology because the information available for the Netherlands did not allow us to also make a distributive analysis. We however feel that such distributive analysis wouldn’t have been less relevant or interesting. It should also be mentioned that it was not possible to fully verify that the costs considered for every stakeholder in the two country studies are exactly the same because the Dutch study does not provide a very detailed description of the costs considered. From the available information it appears that there should not be major differences in this respect. The Dutch study, however, might consider a few more cost items than the Belgian one (this is the case for example of the merchants’ cost in terms of time to carry on the transaction in the case of cards). In any case it is unlikely that such differences can lead to different results. Another observation is that a perfect match of the costs items considered in the two countries could be impossible due to institutional and market differences. This refers to the fact the certain costs items might have been disregarded in one country because they constitute a revenue for another stakeholder, but they are not in the other country. For instance, in Belgium the cost of POS terminals from merchants is some revenue for Banksys which is another stakeholder. Therefore being a mutual transfer, POS terminal costs have been disregarded in the Belgian analysis, however it is not sure whether the same cost item constitutes a mutual transfer also in the Netherlands. A full computation of cost and revenues (i.e. including mutual transfers) should be able to solve this problem but, as already mentioned, it could not be applied here.
Appendix 2: Average cost analysis From the analysis of fixed and variable costs in section 1.6 we derived two types of average costs curves using some indicators that have also been used in the Dutch study. In this appendix we provide the methodological details on which the figures are based. The indicators used are the marginal cost associated with one additional transaction12 (defined as
12
In the Dutch study this is defined as ‘cost of one additional transaction’.
33
a) and marginal cost associated with one additional euro in value13 (defined as b) which are presented in Table A.1. The exact formulation of these two indicators for any payment instrument i is as follows:
ai =
VN i Ni
bi =
and
VSi Si
where VN i are the total variable costs depending on the number of transactions, Ni it the total number of transactions, VSi are the total variable costs depending on the transaction value and Si is the total value of transactions (sales). From these two indicators we can derive the average cost per average transaction for every payment instrument, which is equal to ai + bi si , with s the average transaction value (represented in figure 1.2). The table below reports the values for the Netherlands14 as well as the corresponding values that have been derived for Belgium. It can be observed that while variable costs related to debit cards across the two countries are very similar, for cash and epurses variable costs are generally larger in Belgium. This of course fully ignores the fixed costs’ part that for cards is smaller in Belgium, as shown in figures 1.19 and 1.20.
Table A1: Variable cost indicators Cash
Debit
e-Purse
Belgium
Netherlands
Belgium
Netherlands
Belgium
Netherlands
Cost of 1 additional transaction, a
0.2145
0.1117
0.1919
0.1903
0.1626
0.0333
Cost of € 1 additional sales, b
0.0166
0.0069
0.0011
0.00014
0.0014
0.00001
Variable costs per av. transaction
0.3167
0.1764
0.2443
0.1965
0.1683
0.0333
The same type of information is also presented in figures 1.21 and 1.22, which show the average variable costs curve of each payment instrument for any transaction value, s. Hence the curves represented in figures 1.21 and 1.2215 are based on the following relationship:
VCi = ai + bi s .
13
In the Dutch study this is defined as ‘the cost of one additional euro of sales’. See Table 4 on p.9 in Dutch National Bank (2004). 15 This figure corresponds to Chart 2 in DNB (2004). 14
34
Appendix 3: Belgian costs’ studies
We have to acknowledge that when this study was completed, a new study estimating the costs of payment instruments in Belgium was released. The new study titled “Coûts, avantages et inconvénients des différents moyens de paiement” is the result of the set up of a Round Table organised in Belgium in 2004 with the participation of the financial sector and consumers and merchants representatives and coordinated by the National Bank of Belgium (NBB). It estimates the resource costs of retail payment instruments in 2003. The study closely replicates the analysis made by the DNB (2004) for the Netherlands. Therefore, the methodological differences existing between De Grauwe, Buyst and Rinaldi (DBR) (2000) and DNB (2004) are also valid when considering the DBR (2000) and the NBB (2005) study (see appendix 1 for a description of methodological differences). In terms of results it has to be noted that DBR (2000) estimates the costs of the different payment services in 1998 while NBB (2005) in 2003. The resulting differences in resource costs, however, cannot be fully imputed to the different observation year, even though the role of time in explaining some of the variation in costs cannot be underestimated, particularly as far as cards are concerned. In fact, during the period 1998-2003 the usage of the different types of cards has increased. An immediate consequence of this is that the average transaction value of e-purses and credit cards has decreased in the meantime. For debit cards however the variation has been minimal which witnesses the fact that the debit card scheme in 1998 was already mature. Let us consider in more detail the origin of the main differences in the two costs estimates. A first and very large difference concerns the value of cash transactions. Despite the fact that the estimated number of transactions and the relative resource costs in 1998 and in 2003 are very close, the value of cash transactions was estimated to 19 billion euros in 1998 against the 52 billions in 2003. It follows that the resulting average transaction value of cash is also very different (€6.2 in DBR (2000) against €17.6 in NBB (2005)). This fact clearly explains a large proportion of the difference in the estimated costs of cash. In fact, as shown in the table below, while the cost of cash on a per transaction basis is comparable (€0.56 and €0.53 respectively), the cost as a percentage of the transaction values varies a great deal (9 percent and 3 percent). The difference in the analysis of variable costs in the two studies has also to be connected with the different payment value of cash.
35
Thus, the issue arises of the size of cash transactions in Belgium. The NBB estimation is obtained from private consumption (cf. note (1) on page 25). The DBR’s estimation is obtained by multiplying the average transaction value by the total number of cash transactions. Although the NBB procedure appears more ‘scientific’, the resulting average transaction value (€17) appears abnormally high. As a term of comparison, the same number is estimated to be €9.4 in the Netherlands. In the end the truth lies probably somewhere in between the two Belgian estimates, a definitive estimate being impossible to make. Therefore, to bypass the problem in our analysis of variable costs in section 1.6.2 we used them both. As far as the costs of debit cards are concerned, the increase in the number and value of transactions between 1998 and 2003 has been reflected in a more or less proportional increase in the resource costs involved. To give an idea about the size of such increase the volume and value of debit transactions increased by increased by 180 percent while total costs involved increased by 240 percent, which explains the large difference in the first two measures shown in table A.2. Moreover, not surprisingly, the increased usage has had a stronger impact on the fixed than on the variable costs. Despite this difference, the analysis of variable costs for debit cards is relatively similar in the two studies. The same pattern also applies to credit cards and e-purses.
Table A2: Main cost indicators CASH
DEBIT CARDS
DBR (2000)
NBB (2005)
DBR (2000)
NBB (2005)
Resource cost per capita
€ 164
€ 153
€ 12
€ 29
Resource cost as % GDP
0.74
0.59
0.05
0.11
Cost per transaction
0.56
0.53
0.40
0.54
Cost as % transaction value
9.0 %
3.0 %
0.8 %
1.1 %
36
CHAPTER 2
Implications of Efficient Pricing of Cash and Debit Cards
2.1
Introduction
Although it is well known that the use of cash to make payments is more costly to society than other means of payment, cash continues to be widely used. Moreover, there are quite a few differences between different countries in the EU regarding the use of alternative payment systems like debit and credit cards. While undoubtedly taste differences can explain some of the differences regarding the penetration of payment services, another (larger) part of the differences must be driven by economic factors such as pricing, the number of alternatives available or institutional (legal) features. The present study is an attempt to explain the differences in usage of current account payment services by focussing on these economic factors, in particular pricing. For this purpose, we need to look into the pricing structure of all payment instruments. One characteristic of this pricing structure is that some instruments carry a price, while others such as cash are provided (nearly) for free, although there is a cost associated to using them. Given that it is impossible to make a detailed analysis of all payment instruments, we will focus on retail payment services associated with current accounts, notably cash and debit cards. These are the two most common alternatives offered by current accounts: cash because it is legal tender and, debit cards because they are typically included in the standard current account package. Current accounts sometimes also include other payment services, like epurses or cheques, however only cash and debit cards are common across all the countries
37
considered16. Credit cards are not included in the analysis because they are not part of the current account package and do not only allow payment services. We will use the recent theoretical literature to analyze the optimal pricing of cards and cash. We first summarize the arguments that follow from the most commonly used models. We then ask what additional elements that are not present in the models could account for differences in the penetration of payment instruments. Finally, we conclude the theoretical part by analysing some policy issues. Although many of the arguments apply to both debit and credit cards, the focus is on the first type of card. The ultimate purpose of this part is to suggest how a current price structure for payment instruments should look like. Next, we take an empirical approach to explain the share of cards and cash usage. Recently, industrial organisation has relied heavily on micro-econometric modelling to explain the market shares in product differentiated industries. A similar approach will be followed here. This will lead us to specify a model that explains the market shares of cash and debit cards by price variables and by variables expressing the convenience of cash and cards. The model will then be estimated using data of European countries. Finally, the results of this econometric analysis are used to compute the extent to which the use of a cost-based pricing system will affect the market shares and the resource cost of payment services.
2.2
Some stylized facts
A variety of solutions exist to execute a payment. Payments themselves can be put into different categories: large amounts or small transactions, domestic or international, between households (individuals) and companies, …., etc.. In the present study, we will exclusively focus on the so-called C2B segment, see table 2.1 for a classification of these segments and the most common payment instruments used in each segment.
16
In many of the countries we analyse e-purse schemes are not in place and even when they are, their use is marginal. Conversely, in most cases cheques are no longer used for retail transactions.
38
Table 2.1: Market Definition and Research Focus TO
CONSUMER
MERCHANT
FROM CONSUMER
C2C
B2C
Cash C2B MERCHANT
B2B
cards cash
electronic money transfer
electronic money transfer
This so-called C2B segment is characterized by many small value payments. In this segment we still see the use of a lot of cash, and moreover notice pronounced differences between the penetration of cards between EU countries. To give an idea about the relative use of these instruments, table 2.2 and figure 2.1 show the number of transactions of cash and cards in some selected countries. In this respect, it is important to note that official statistics for the number of cash transactions in the different countries are not available. The reason is related to the fact that this number is very difficult to estimate because no record is kept of many cash transactions at the point of sales. Moreover, in contrast with cards, cash is also used for person-to-person transactions, which are even more difficult to estimate. For a few countries broad estimates exist for a particular year, but they are always rough estimations deduced from merchant and banking surveys. Therefore, we estimated the number of cash transactions throughout the years in a homogeneous way across the different countries. The estimation has been obtained as follows. We assumed the total number of transactions (cash + cashless) per day made by each inhabitant to be constant across countries and equal to two transactions per day per person. Given that official statistics on the number and amount of cashless transactions are available, we subtract this number form the total so as to obtain the estimated total number of cash transactions. We are aware that this type of estimation cannot be precise because there are likely to be important differences in the frequency and nature of payment in the different countries; however, our estimates at least give an indication of the rate of use of cash and non-
39
cash instruments which are undoubtedly related17. We show the results of this estimation in table 2.2 and figure 2.1. Focusing on the penetration of cards we observe a quite diverging picture. Table 2.2 and the associated figure 2.1 indicate that in Belgium, France, the Netherlands and the UK, as compared to Italy and Spain, citizens make at least three times as many transactions with debit cards, pointing to a more pronounced use of cash in Italy and Spain. Similar findings hold for the year 2000, 2001 and 2002.
Table 2.2: Transactions per person per year (2003)18
CASH
DEBIT
Belgium
564
53
France
509
73
Germany
568
20
Italy
674
11
The Netherlands
510
71
Spain
649
15
United Kingdom
526
57
Figure 2.1: Transactions per inhabitant (2003)
700 600 500 400 300 200 100 0
BE
FR
GE
IT
NE
Debit
Cash
SP
UK
17
It should also be mentioned that comparing the estimated amounts to the few country estimates available we could note that the numbers are quite comparable. 18 The data presented in table 2.2 and figure 2.1 are taken from a common source, i.e. from the Blue Book of the ECB which standardized the definition of debit card transactions across countries. For this reason, these figures might not coincide with the corresponding figures coming from country sources.
40
Another way to represent the findings of figure 2.1 is to denote for each country the percentage “cash”, i.e. the percentage of payments made by cash, and to do the same for the percentage “non-cash”. The result is the parallelogram showed in figure 2.2. The closer the graph comes to a rectangle, the less pronounced the between country differences in the use of the different payment instruments. For Europe, pronounced differences exist.
Figure 2.2: Pronounced Differences in the Use of Cash and other Payment Instruments in Europe
Max:
92
Italy
CASH
Min:
70
8
NON CASH
The Netherlands/ France
30
As argued in the introduction, a variety of factors should be called upon to explain these differences. We investigate first the economic variable “par excellence”: the price of the payment instrument. Next, we include other factors, such as the “convenience” provided by each instrument. By convenience is meant that payment instruments are easier to use than others. Or they are “qualitatively superior”, at least in particular circumstances. One of the crucial factors in this respect of course is who accepts the payment instruments to clear a purchase. This immediately brings us to the “two-sided” characteristics of the market, and the recent theoretical developments in that area, as well as other features of product differentiation that may matter in this respect.
41
2.3
The optimal pricing structure for payment instruments
In this section, we show the consequences of mispricing of the different payment instruments. In a first subsection, we document the immediate consequences of the underpricing of cash. In a second subsection, we show that the implications of not pricing cash properly may reach even beyond those documented in the first subsection.
2.3.1
The costless but costly alternative: Cash
The detrimental effects of having cash almost for free can be shown graphically. The exposition follows a number of stages. First, we make some assumptions. Then we illustrate the distortions using a series of figures. The comparison we focus on is in terms of number of the transactions paid for by a consumer. Alternatively, the analysis could be done in terms of the size of the transaction involved, see Brits and Winder (2005). Essentially, both outcomes yield the same conclusion. Initially, we focus on cash and debit cards in the absence of multi-homing, (where a consumer uses both cash and cards). In order for a consumer to execute a transaction in cash, a central banker has to print notes, and the banking system as a whole has to distribute the notes to branches. Each time a merchant accepts a payment in cash, he has to get the notes back in to the banking system, and so on. This involves numerous operations of counting, transporting …, notes. A detailed description of the processes performed by a cash centre of a mid-size retail bank in Europe is given by Dab (2005). For the moment, we assume that each time a consumer makes a payment in cash, this involves a cost c to the economy as a whole. In addition, making available the notes comes at a fixed cost equal to z. The other possibility is to use a debit card. Setting up the possibility to pay with such an instrument also involves some costs. These however are mainly fixed: setting up the IT platform, connecting the servers with the banks and so on. We assume that these costs, denominated as w, are more substantial than the fixed costs associated with cash payments (w > z) but they can process virtually any number of transactions with very small increase in costs. In figure 2.3, we set out these different cost components (vertical axis) as a function of the number of transactions (horizontal axis). Note that this is a long-term analysis therefore it is
42
necessary to consider total costs of the two payment systems. The line CASHSOC represents the total cost of cash for society as a function of the number of transactions. It will be remembered that these were called the resource costs of cash in the previous chapter. The line CARDSOC represents the total costs of cards to society as a function of the number of transactions. From panel a of figure 2.3 we find that there is a point qS beyond which the use of cash is cost inefficient, i.e. uses up more scarce resources than cards. It becomes clear that from a welfare point of view, cash should not be used beyond this point. Thus if cash is priced correctly, i.e. it reflects all resource costs only a limited number of consumers will use cash. This is the result of the expensive nature of cash in terms of transacting, as shown by the steep slope of the “CASHSOC” curve. In panel b of figure 2.3, we show what happens when cash is not priced properly, as it is the case today. We move from the social optimal (CASHSOC) curve to the private equilibrium (CASHPRIV). The CASHPRIV curve has no intercept (for we assume that the public does not pay for the cost, z, of printing notes), and a flatter slope because we assume that in order to get a note to pay, the only cost the consumer takes into account is the cost of visiting the branch of a bank. (If the consumer at each visit to his bank branch withdraws more notes at a time to make several payments, as in the well-known Baumol-Tobin (1952, 1956) model of demand for money, the CASHPRIV-curve would be step shaped, but nothing fundamental to our analysis would change). Therefore, the slope of CASHPRIV is flatter than the slope of CASHSOC. The result is clear to the public, cash is the better priced alternative up to a number of transactions equal to qP. The implication is that too many consumers keep using cash. Panel c of figure 2.3 shows what happens if banks try to convince the public to engage in multi-homing, by including a debit card for free in the current account package and by installing ATM’s which also can be used for withdrawing cash for free. These are the standard options available to current account holders. The result is an even flatter CASHATM-curve with qA even further away from qS. This is because cash is made available even at a cheaper rate, for instead of visiting the bank branch during opening hours, the consumer now can withdraw cash from the wall everywhere an ATM is available, 24 hours a day, 7 days a week.
43
Figure 2.3: Social and Private Optima
Panel a CASHSOC CARDSOC w z
0
qS
q
Panel b CASHPRIV CARDSOC w
0
qP
q
Panel c CASHATM CARDSOC w
0
qA
q
44
The arguments in favour of debit cards may be taken one step further. Once it becomes clear that the majority of the consumers in the economy should use a debit card (as figure 2.3 indicates, only those who do a very limited number of transactions, below qS, should use cash (this is likely to be a very limited group), the existence of a different “mode” (i.e. cash) may be detrimental for still another reason. In the intermodal competition literature, it is documented how the co-existence of networks may lead to the underprovision of a service, if the latter to an insufficient degree can cover the fixed costs involved. The economics of intermodal competition often leads to the suggestion of “prohibitive measures” on behalf of policy making. We will not pursue these any further here, as in reality it has become clear that debit card networks are viable, even in co-existence with cash. But the point made above should be taken seriously: debit cards could be more expensive than necessary because the use of free cash prevents the debit card network from reaching its full potential. This will be even more the case when the free alternative, viz. cash, in fact is costly. This in reality is the case and it is well known that the costs of cash are borne by other parties than consumers, mainly the tax payer (central bank), share- and stakeholders of private banks, and merchants. Especially the stake-holders of private banks can be affected if they are required to cross-subsidize the free provision of cash. Cross-subsidization is a term used by economists to indicate that the price of one item is too high vis-à-vis what it costs, and the reason is that it needs to cover up for another product, for which the price falls short of covering the costs.
2.3.2
Optimal pricing of cash versus cards
In this section we analyze the question of how an optimal pricing scheme for cash and cards should be designed. The viewpoint taken here is one where society provides different means of payment to the consumers. Each and every payment instrument carries a cost and the public has a willingness to pay for each instrument, given its preferences. Without loss of generality we use the subscript 1 for cash and 2 for a debit card. A payment executed by cash then costs c1 while one done by card costs c2. At the same time ε1 denotes the price elasticity of demand for cash (the percentage sales lost as a result of a price increase of cash by one percent), while ε2 is the corresponding number for cards. It then follows from
45
maximizing social welfare subject to a break-even constraint (indicating that payments cannot incur a loss upon the economy) that:
p2 − c2 p2
2 p1 − c1 1 ε = p1 ε
(2.1)
In formula (2.1) above, p1 and p2 of course denote the price charged for making a payment with cash, respectively a debit card. Under the simplifying assumption that the price elasticity of demand is the same for both instruments (we will elaborate on this assumption in section 2.4 when investigating the use of payment instruments empirically), equation (2.1) can be simplified to:
p1 c1 = p2 c2
(2.2)
Formula (2.2) is remarkably simple and transparent and therefore easy to implement. It says that society, if maximizing social welfare will require that the ratio of the price charged for using cash over the price for using a card is equal to the ratio of the costs that are incurred. Equation (2.2) also may be written as:
p1 p 2 = c1 c 2
(2.3)
indicating that both prices should be in the same relation to the respective costs. Besides being simple and transparent, equation (2.3) also points to the necessity for pricing to be cost-based. By cost-based we mean that prices should be set proportionally to their underlying cost. In other words, a pricing system based on the provision of some payment services below cost and others above cost (cross-subsidization) is not a welfare maximizing strategy. Note that ‘cost-based’ pricing does not necessarily mean ‘price equal to cost’, but rather that the cost should be taken into account in determining the price. The payment industry is also characterized by strong network effects. These network externalities extend from one side of the market to the other, and vice versa. In box B we analyze the economics of two-sided markets. This allows us to study how these externalities can be incorporated in our framework.
46
Box B: Two-Sided Markets The recent literature on “Platform Competition” has investigated in detail what the pricing rules should be when a player in a two-sided market wants to maximize profits, see Rochet and Tirole (2003). In this section, we first explain what is meant by a two-sided market. We discuss the optimal pricing schemes derived in the literature for these markets. As the insights put forward hold for both debit and credit cards alike, we do not distinguish between the two at this point. Remarkably, not only profit maximizing players, but also platforms that maximize social welfare and/or act as associations follow in equilibrium pretty much the same rules. To the extent that in reality pricing rules deviate from these characterised by the optimum (whether it be private profits or social welfare), an obstacle for the less than optimal penetration of cards exists. Hence it is important to understand properly the business model of a two-sided market. Two-sided Markets: Business Model The most important feature of a two-sided market is precisely captured by the name: there are two sides, or even better: “it takes two to tango”. In order for a cardholder to be able to pay a purchase with his card, the merchant where he intends to buy needs to accept the card. And vice versa, in order for the merchant to invest in a card-reader, sufficient consumers must be willing to use the card19,20. This interaction of both sides of the market over the platform (= the card) is crucial in a twosided market. Without cardholders, the platform cannot make a value proposition to the merchant. And with no merchants accepting, the card is worthless to the consumer. From an economic perspective, it is this “interaction over the platform” element that is the distinguishing fact of this business. For example, a platform is not an intermediary, who resells what it has bought. Intermediaries provide value added by distributing items, by providing liquidity, and so on. But the value added of a platform differs from that provided by an intermediary. In order to see this, suppose a platform mistakenly perceives its business as 19
Other considerations on both sides of the market that interfere with the decision to use the card or not are discussed below. 20 A distinction can be made between “having” the card, and “using” it. Economically, both approaches can be reconciled, see Rochet and Tirole (2004). When a fee is charged for having the card, for example on an annual basis, whereas also for using it one has to pay, a two-part tariff is in place.
47
one of an intermediary. Moreover assume it regards merchants as the sellers of a payment service, since by accepting the card they provide a facility to the buyer who does not have to carry notes and coins any longer. The customer of the merchant then is the buyer of the service and the platform perceiving itself as an intermediary would charge the buyers when they use the card, and pay a certain amount to the merchant who accepted the card and “provided the payment facility”21. In this business model, the merchant will try to get the best price for providing this payment facility, just in the same way he tries to get the best price for the goods he sells. Inevitably, increasing the cost for using a card will reduce demand and both the card company and the merchant collect less revenues from card usage. This in turn will reduce the attractiveness to merchants to invest in a card reader, to provide the payment service. But if less merchants accept the card, it is not providing a lot of utility to the cardholder. He will object to high charges for a card, reducing the revenues further, a.s.o.. In a formal approach, it is easy to show that if the platform recognizes that it operates in a truly two-sided market, it will charge both sides of the market a certain amount. Whereas when it is perceived as an intermediary, it – by definition – will charge to one side of the market and pay the other. So clearly a first possible mistake is to confuse a platform with an intermediary. In general, a platform requires contributions from both sides of the market22. To understand the optimal pricing rules for a platform, we start from an analysis of pricing in a multiproduct company, as in section 2.3.1. But instead of identifying the different product as “cash” and “debit card”, we now re-interpret the setting. Instead we conceive of “inducing payments with the card (using the card by a consumer)” and “making the card accepted” as the two activities the multiproduct company engages in. It is a model of a company that serves several groups of consumers, subject to a break-even constraint. Or the firm does not maximize profits, but it cannot afford losses either. The bestknown examples are electricity producers and telecommunication companies before deregulation.
21
The platform keeps the difference, equivalent to the well-known bid-ask difference (“spread”) in financial intermediation. 22 In equilibrium, negative prizes on one side of the market could result. This is as if the platform operates as an intermediary, charging one side to pay to cross-subsidise the other side. Or, more commonly observed, it could be that the platform achieves “full participation” on one side of the market, to which it doesn’t charge. Bolt and Tieman (2005) show that this outcome is optimal given particular assumptions – likely too hold in cards – are satisfied.
48
The optimal pricing rules for a multiproduct company (e.g. public utility) involve a positive price charged to all (both) groups of consumers. In that respect, it is closer to the two-sided platform model that is the appropriate business model for the cards industry. And it can be shown that both business models are closely related to each other. Yet if applied directly to cards, it overlooks important elements. The reason is that according to this business model, high prices will be charged on the inelastic group of consumers. To see this, suppose for a moment that these are the cardholders. (Merchants then are the elastic group). Suppose further that by a 10percent increase in prices, with a charge going from 100 euro per year, the price of the card becomes 110 euro. Whereas before, the card company had 1000 clients (and revenues equal to 100.000 euro), the number of cardholders as a result of this price increase drops to 910. Revenues then are equal to 100.100 euro, or they increase by 1000 euro. At first sight, the price increase from 100 to 110 euro was then a good idea. But this calculus overlooks the fact that in a two-sided market, 90 cardholders less (a drop of 9 in the customer base) reduces the attractiveness of the card to the merchant. The latter who also pays for the card (perhaps by renting a card-reader) finds that he will be forced to accept 90 additional payments in cash (assuming every cardholder purchased once a year with this merchant). Even though the merchant belongs to the elastic group and is charged a low price for the card reader, the device becomes less attractive to him for fewer of his customers demand it. Again, the substitution of the appropriate two-sided platform model by the economy business model leads to mistakes in pricing. By now it then should be sufficiently clear that the appropriate approach of two-sided markets should be followed and the question we investigate next is: what then are the implications that follow from the optimal pricing rules? Even if the platform correctly realizes that it operates in a two-sided market, the pricing rules that should be implemented carry counter-intuitive elements. In this section, with a minimal degree of formalism, these formulae are discussed, and it is explained what counter-intuitive logic underlies them. Failure to recognize this inevitably results in erroneous pricing. Two-Sided Markets: Optimal Pricing Schemes The equilibrium profit maximising pricing rules for a platform are obtained from solving a system of first order conditions. In the specific setting pioneered by Rochet and Tirole (2003), this leads to the so-called “proportional” pricing rules. More in particular, when p c denotes
49
the price charged to the cardholder and p m denotes the price to the merchant, whereas ε c and
ε m denote the corresponding own price classification of demand, we have: pc εc = pm ε m
(2.4)
The formula clearly indicates how the price charged to the cardholder is proportional to his demand elasticity. This implies that if cardholders are sensitive to high prices, they pay a lot, contradicting received theory and common sense. Further investigation however reveals the logic beyond (2.4). To illustrate this, first denote p = p c + p m and ε c + ε m = ε . Then it can be shown that
p−c 1 = p ε
(2.5)
Where c is the cost to the platform for an interaction. Equation (2.5) is the well-known Lerner or standard “inverse” elasticity rule. Or regarding the overall price level, nothing changes. The (apparent) difference is in the allocation of the charges to the different groups vis-à-vis received theory. And to show further that the difference of the proportionality rule is only artificial, again contrast the platform competition model with the better-known model of multiproduct public utilities, discussed informally in the previous section, when deriving equation (2.3). The resulting optimal pricing rules for that business model are: pm − c 1 = m m p ε
(2.6)
pc − c 1 = pc εc
(2.7)
and
Or the standard inverse elasticity rules apply on each consumer group. That is, if merchants are inelastic and cardholders elastic, then the first group will face high charges and the latter will pay low prices. This logic apparently seemed not to hold for platforms. But it can be shown that (2.4) and (2.5) imply:
50
1 p m − (c − p c ) = pm εm
(2.8)
p c − (c − p m ) 1 = c pc ε
(2.9)
and
Clearly, equations (2.8) and (2.9) are the exact equivalents of (2.6) and (2.7), except for the appearance of the price charged to the other side of the markets in the formulas. The so-called price-cost margins (the left hand sides of equations (2.5) to (2.9)), use in the case of equations (2.8) and (2.9) a different cost concept. Besides the marginal cost which is also present in equations (2.5) to (2.7), also the price charged to the other side of the market enters the picture. The interpretation is as follows: when an additional card is sold, this has a cost. But it generates a revenue on the other side of the market as well, since for example an additional merchant will be involved to become a member of the network, because cardholder demand has increased. The result is that the marginal cost is corrected by subtracting the revenue generated on the other side of the market. Some additional intuition can be provided. Suppose initially membership nearly is free for merchants. They will pay a symbolic charge of 1 euro. The result is that many merchants will accept the card, a strong and important advantage to the cardholders. Next, assume that for some reason or another, it becomes clear to the platform that somewhat more should be charged to the merchant. For instance, it becomes clear by some study that the optimal charge is 2 euro. Consider what happens under two different scenarios. In one case, the merchants are the elastic group whereas the cardholders are inelastic. In the other case, the opposite assumption is made. When the platform increases its charges from 1 to 2 euro to the merchant side, which is elastic, many merchants will drop out. In order to compensate for the reduced the number of merchants, the platform will need to make the service more attractive, by attracting more cardholders. The latter are inelastic, so a large decrease in the price they need to pay will be needed to do the job. Or a 1 euro increase in the merchant price (from 1 to 2, or 100 percent) is matched by a substantial decrease in the
51
charges to the cardholders. The result is a low ratio of cardholder to merchant prices, in line with ε c / ε m which is inelastic/elastic or low/high or low. Under the alternative assumption of elastic cardholders and inelastic merchants only few merchants drop out, so in order to restore their number, only a few additional cardholders will do. Moreover, this group is elastic, so a small price decrease already gets the required number. Or doubling the merchant price only leads to a small increase in the price charged to the cardholder. The resulting equilibrium ratio of prices is high, in line with the ratio of elasticities which is elastic/inelastic or high/low or high.
2.3.3
Card Pricing in Reality
In Europe current accounts usually include a card, which allows both to withdraw notes and coins from ATMs and to pay at points of sale. Generally banks issue debit cards in combination with ATM cards without an explicit choice on the part of the consumer. In this sense debit cards are also distinguished from credit cards which are not included in the current account package and as such require an explicit choice on the part of consumers. Because of this reason we do not consider charge or credit cards in this section. In reality, the pricing schemes that are used for current account payment services are not the same everywhere. On the contrary, a high variation is observed between countries and sometimes also within countries. Moreover, the ATM and debit card being included in the current account package, they are not always priced as individual services, but are sometimes included in the general current account fees. This certainly does not contribute to the transparency of the market. Probably, this is the result of regulatory interference in the market. As the sensitivity of the users of payment instruments differs across countries, it is likely that providers will be inclined to use tariff structures that avoid objections. It then becomes to some extent politics that will determine who pays what. Although in general, it is known that competition in twopart tariffs can lead to multiple equilibria too.
52
Let us first review pricing for debit cards associated to the current account across a sample of thirty European countries (including all euro area countries)23. Consumers are charged an annual fee for the card, ranging from 2 to 43 euros with an average cost of about 10 euros24. There is evidence that in the last years the annual cost of a card has slightly increased (Retail Banking Research, 2005). The UK, Greece and Slovenia are the only countries in Europe where cardholders are not charged an annual fee. Generally, the use of debit cards at POS does not involve a per-transaction fee. The only exceptions are Iceland and Norway, where cardholders are charged a fixed fee by the issuing bank every time the card is used at a POS. In a couple of other countries (Austria and Ireland) a fixed fee is applied only after a given number of free current account operations has been exhausted. Considering the costs of cash, consumers’ cost of cash provision is here approximated by the fees for money withdrawals at ATMs25. Typically, ATM fees are always flat, i.e. they are a fixed amount26. Pricing practices for ATM services are of three different types. •
In the majority of countries, withdrawals are free at the ATMs of the cardholder’s bank (‘on-us’ transactions), while withdrawals from other banks’ ATMs are charged a fee27. Denmark and Norway are two special cases within this group. In these two countries “onus” ATM withdrawals are free only during office hours and are charged outside office hours.
•
In some countries ATM services are free for cardholders for all banks and in the whole country. This is generally the case of small countries in which banks share a common ATM infrastructure, being Austria, Netherlands, Portugal, Sweden, Iceland, Slovenia and until recently also Luxembourg.
23
It should be mentioned that the pricing policies described in this section might not be exhaustive. There exists large differences in the fees applied by different banks and for some countries we could only survey the rates applied by few banks, which therefore might not be fully representative of the whole country. 24 However in some countries (Portugal, Germany, Netherlands, and Poland) cards might be issued without yearly fee for students and other particular categories of consumers. 25 The cost of provision of cash by other way than via ATMs is not considered here for lack of information. However it is important to note that despite the different pricing policies across countries, the common practice is such that services requiring manual intervention are subject to higher fees than ATMs. This practice contributed to reduce non-standardised transactions, which recently became marginal, at least in some countries. 26 Only in Greece there are variable fees (1 percent of the amount in the case of not ‘on-us’ transactions), while in Spain, Hungary and Latvia the fee is constituted by both a fixed and a variable part. 27 Sometimes the free service is also extended to other banks with which the cardholder’s bank has a special agreement or shares a common infrastructure.
53
•
Finally, in Ireland, Czech Republic, Romania and sometimes also in Spain, all ATMs withdrawals are charged whatever the ATM’s ownership.
Let us now consider payment services costs for merchants. As far as cash is concerned, it is very difficult to summarize the type of costs associated with it. Cash costs might be relatively visible and explicit for large retailers, but are very difficult to determine for small ones. Conversely, for debit card acceptance merchants incur three types of costs. •
First, there is the cost of the terminal, which is sometimes rented and sometimes bought. Typically the same terminal will be able to process more than one type of card, which makes it difficult to allocate a precise share of this cost to debit cards only.
•
Second, merchants sometimes pay entry fees to enter the card scheme while sometimes they pay a monthly or annual subscription to the scheme. In some cases merchants pay both, the entry fee and the yearly/monthly subscription which might include the terminal rent and telecommunication costs. Very little information is available on these two categories of costs because merchant agreements are usually confidential28.
•
Third, merchants pay a fee per transaction called Merchant Service Charge (MSC). MSCs can vary a lot from one merchant to another. A first dimension of variation is between countries and mainly depends on the debit card scheme(s) in place. Another dimension of variation is related to the merchant’s sector of activity; such sector differentiation is present in all countries. Also the nature of the MSC varies; sometimes it is a fixed amount while sometimes it is a percentage of the transaction value (see the table 2.3). Despite these differences, during the last years in Europe MSCs have slightly decreased following market pressure and regulatory actions (Retail Banking Research, 2005). It is also worth noting that in Denmark and Norway debit cards do not have a merchant fee (recall that in Norway cardholders pay per-transaction fee).
28
For this reason later in the paper we will not consider these costs in the empirical analysis.
54
Table 2.3: Merchant Service Charges in Europe FIXED AMOUNT
VARIABLE AMOUNT
NO MSC
Belgium, Germany (POZ),
Austria, Finland, France,
Denmark,
Ireland, Netherlands, Switzerland,
Germany (e-Cash), Greece, Italy,
Norway (Bank Axept)
Sweden, UK
Luxembourg, Portugal, Spain, Czech Republic, Hungary, Malta, Poland, Slovakia, Slovenia
Trying to put into perspective the cost of cash and debit cards for consumers, from what we just reviewed we can conclude that the current pricing structure is not very transparent and is certainly not related to the underlining costs of the service. As we mentioned, the pricing structure may vary a lot across different financial institutions in the same country and is often unknown to the users. As such, the prices certainly cannot act as “signals” so as to favour a rational and cost-efficient use of such instruments. On the merchant side, the situation is even more complex. In the case of cash, in most cases the costs are hidden and therefore impossible to quantify. In the case of cards, the pricing structure is in principle clearly specified by contract. However, because of different fee structures for different cards, and because of several other costs involved (terminal, communication, network participation) it might be quite difficult for merchants to figure out the actual costs of an instrument on a per-transactions basis. One can conclude that there is room for potential benefits from transparency and a rational reorganization of the pricing policies both for cash and for cards.
2.4
Card penetration and pricing
The last three decades have witnessed an expanding literature on discrete choice theory. This subdiscipline of economics deals with explaining how consumers pick a certain brand, and therefore it ultimately explains what market share the manufacturer of a certain product (or service) has.
55
More recently, these models have been used to simulate the outcome of particular mergers. Nowadays, an important merger case will trigger a review process that heavily relies on such merger simulation, see Van Bergeijck and Kloosterhuis (2005). The application of these models to payment instruments and technologies is not trivial, yet when explaining why cards are a success (or not), one wants to know what determines the choice of the consumer and hence how the market shares are achieved. In this section, we first discuss the features of payment instruments in a non-formal manner. This is to identify the factors that could matter in a quantitative analysis. Next, we borrow from the existing discrete choice theory to set up a model of choosing to execute a payment either with a card or with cash. Finally, we try to capture especially the impact of several determinants, including pricing, of the different payment instruments on their market share.
2.4.1
Product Differentiation and Discrete Choice Theory
Economists distinguish between horizontal and vertical product differentiation models. In the first category, consumers have heterogeneous tastes. Some like fancy metallic colours on their cars or bikes, whereas others prefer more classic or traditional paints. The second category of product differentiation has consumers with homogeneous tastes: everyone prefers a certain brand of a car above another one. But the consumers differ in other respects, i.e. mainly in their income or what they can and/or will pay for the product. The problem here is to ensure that the consumers who can afford it pick the best quality provided by the market. Among other things, this implies charging enough for the lowest quality and not too much for the highest. Most empirical models used nowadays combine elements of above mentioned theoretical work. The idea is to explain why a given population chooses the way they do, i.e. to explain market shares, taking into account differences in the product, its prices, and tastes. Since the latter are unobserved, the typical approach will distinguish between observed and unobserved characteristics. The first group results from the product design of the constructor (the car is a sedan (4 doors), rear-wheel traction, x kilowatt of horsepower (or y seconds for acceleration to a certain speed limit), full consumption of z litres per 100 kilometres, a.s.o.), as well as the pricing decision taken by the car manufacturer.
56
Unobserved characteristics on the other hand relate to the ignorance of the analyst on the importance/unimportance of particular features. Does the consumer value a particular design? How does he like the interior of the car? ….. For payment instruments as a whole, there has to our knowledge not been a study on the importance of product differentiation. Nevertheless, several informal studies have divided the spectrum of payment facilities and they constitute an interesting part of departure. The studies in question either tried to reflect on the type of service provided, or tried to get an impression of the competitive forces that one instrument of payment exerts on the other.
2.4.2
Explaining market shares
In this section we summarize the informal discussion made in the previous sections so as to arrive at a more formal specification, to test against the data in the next section. Recently a discrete choice theory approach has been used to explain consumer choice in cards, see Rysman (2004). This article mainly focuses on consumer choice within credit cards. The approach followed here will be broader in that it looks at consumer choice over categories of payment instruments, more in the spirit of Hyytinen and Takalo (2004). In particular, we consider current account-based retail payment instruments. But the “convenience” element that comes up in Rysman in the form of a positive feedback loop is withheld. From the product differentiation literature, it becomes clear that when considering the choice of consumers with respect to payment services we need to control for product differentiation, if possible. This can be observed and/or unobserved to the researcher. Next, the relative prices are important. A discrete choice is any choice made from among a limited range of mutually exclusive options. The basic principle of these models is utility maximization where each individual tries to make choices maximizing his well-being. The underlying utility function is a numerical representation of the preferences. The testable empirical model is obtained by assuming that the maximized utility function is not totally deterministic, but contains a probabilistic term of known statistical distribution. This probabilistic term reflects the fact that part of the choice is not explainable by the analyst for various reasons (information availability, lack of consumer rationality, imperfect information on the options, etc.).
57
Formally, it is assumed that the utility of choice option i for individual n is Uin which is considered to be the sum of a deterministic and probabilistic term: Uin = Vin + ein where Vin is the deterministic term and ein is the probabilistic term of individual n and option i. In our specific case We denote the options for consumers with i=(Debit, Cash) representing the mutually exclusive choice of paying by debit card or by cash. Usually, the literature relates the market shares of one product (payment service) to the market share of a benchmark variety. In the present, we try to explain the relative use of debit cards to cash. This is not entirely the same as the relative market share of debit cards to cash, but the explanatory variables remain relative prices and product characteristics. For the last category the main focus is on “convenience”, i.e. whether the card gets accepted. To this extent we computed the number of points of sale (a measure of places where the debit card can be used) and divided it by the total number of merchants in a country. It gives an indication of the penetration of card readers in the merchant population, and thus how easy it is to use the card. Other possible characteristics of cards are captured by including the number of debit cards per capita. If a card offers many facilities most people will have one. This variable is also included to enable the use of a debit card to pay, conditional on its presence. If in a particular country very few people have a debit card, it is unlikely to see many payments being executed by debit card. Finally, price variables will determine how much debit cards will be used. This holds in a negative sense for the price charged for using the card at the points of sale as well as the annual fee. Both are included in the regressions as separate variables or as a unique price variable defined as total fees paid per year per capita. When the alternative, cash, becomes more expensive, debit cards will be used more frequently to pay. Since in many cases, paying with cash is not charged, and it is hard to incorporate the opportunity cost in terms of visiting a branch to withdraw money as well as to calculate the foregone interest, a cash payment often carries a cost equal to zero to the consumer. In these cases it is impossible to regress the ratio of debit card payments over cash payments on the ratio of the respective prices, for it would imply dividing by zero. In some cases, however, consumers are charged when they withdraw cash at ATMs so that this variable is used as the price for cash. Similarly to debit card fees, we define the variable as total ATM fees paid per year per capita. In the countries 58
where there are such fees for cash withdrawals, we also derive the relative price of debit card with respect to cash (the price of debit card divided by the price of cash) to use as additional variable. The data on the variable discussed are obtained from a panel of 25 European countries for which we have yearly observations for the period 1998-200329. Unfortunately we could not obtain full information for all the countries, therefore the panel is unbalanced.
2.4.3
Empirical results
The specification, using the variables described in the previous section, has been estimated by means of panel data techniques, notably, the fixed effect estimation. Table 2.4 below presents the results of several regressions, which are nested versions of the following equation:
DebitVol POS Cards DebitFees ATMFee = α + β1 + β2 + β3 + β4 +µ CashVol Merch pop pop pop
where
(2.10)
DebitVol = the ratio of debit cards to cash transactions CashVol
POS = the number of points of sale per merchant Merch Cards = the number of debit cards per capita pop DebitFees = debit cards fees paid by consumer per capita pop ATMFee = ATM fees per capita pop
In some specifications we will use these two price variables as a ratio, i.e.
ATMFee/DebitFees.
29
The countries included in the analysis are: Austria, Belgium, Denmark, France, Finland, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, UK, Czech Republic, Estonia, Hungary, Latvia, Slovenia, Poland, Romania, Switzerland, Norway and Iceland.
59
Table 2.4: Estimation of Debit Card Usage Dependent variable: Debit over Cash volume
(1)
(2)
(3)
(4)
(5)
(6)
POS over Merchants
0.006** (0.003)
0.026*** (0.007)
0.014** (0.007)
0.010* (0.006)
0.010* (0.006)
0.009*** (0.004)
Debit cards per capita
0.091*** (0.011)
0.066*** (0.016)
0.048*** (0.016)
0.052*** (0.016)
0.051*** (0.016)
0.078*** (0.013)
Per-transaction Price of Debit per capita
-0.029* (0.016)
-0.019 (0.015)
Yearly Price of Debit per capita
0.002* (0.001)
0.003*** (0.001)
Total Price of Debit per capita Price ATM per capita Price Debit/Price Cash
0.002*** (0.001) 0.002*** (0.001)
0.001*** (0.001)
0.001*** (0.001)
0.001*** (0.001) -0.002*** (0.001)
constant
-0.023*** -0.032*** -0.026*** -0.025*** -0.025*** -0.018*** (0.007) (0.008) (0.008) (0.008) (0.008) (0.009) N 25 20 19 19 19 13 R2 0.50 0.64 0.68 0.67 0.68 0.73 Note: ***, ** and * indicate respectively 1%, 5% and 10% significance rate. The standard errors are in parenthesis.
Considering the different regressions, all coefficients except one have the expected sign and are significant at least at the 10 percent level. In particular, we find that most price variables have the expected sign. An increase in the price of retrieving cash from ATM reduces the use of cash. Similarly an increase in the per-transaction fee of debit cards reduces the use of debit cards. We also introduced the relative price of debit cards versus cash (column (6)). This variable also indicates that when the price of using cash increases relative to debit cards consumers switch away from cash in favour of cards. The one odd finding is that the fixed annual fee on debit cards has a positive effect on the usage of cards. It is as if the consumer tries to reduce the cost per transaction of the card by increasing the number of transactions. This result contrasts with the effect of the charges for using the card, which have the expected sign, although in the last regression they appear not to be significant30. When the total cost of the card, i.e. the annual fee together with the charge per transaction is withheld, the total 30
Note that this series has value zero for most countries given that only in three countries out of the sample there are per-transaction fees on debit cards. This might explain the statistical non-significance of the estimated parameter.
60
effect is positive and significant (see column (4)). This could be explained by a misperception of sunk costs by the consumer. In particular, the consumer who has paid for his card does not realize that he should evaluate at each occasion what is the most appropriate way of paying, given the opportunity and the cost of the payment instrument. In other words, it is likely that the consumer thinks that “he should make the most of it” given that he has already paid for having the card. A number of interesting conclusions therefore emerge from our empirical results: 1.
Consumers respond to changes in the relative price of cash versus cards. This implies that
a policy aimed at pricing cash so as to reflect its true costs will induce consumers to switch towards the use of debit cards. In the next section we analyse how strong this switching effect is likely to be. 2.
“Convenience” as measured by the number of merchants that accept the card (as proxied
by the number of POS terminals) is an important determinant. The positive sign of this coefficient points to a positive feedback loop (see Rysman). It indicates that this market is a truly two-sided one and hence that penetration on the merchant side is crucial. Some further research should refine our findings. Especially the number of debit cards could be endogenous. Looking at the regression coefficients in table 2.4, it appears that the number of cards is positively correlated with the annual fee, but negatively correlated with the pertransaction fee. This could suggest that consumers’ choice to take a card is not discouraged by its annual fee (recall that usually the same card is also used at ATMs), but the per-transaction fee might discourage its use. However, even if it was the case that the number of cards are endogenous, our results would not change, see column (5) of table 2.4 which excludes the per-transaction fee from the regression. But the issue of how the consumer decides about “having” a card versus “using” it certainly merits further investigation, especially in a world where multi-homing is important. Another important area for further research is explaining the “ownership” decision on the other side of the market. As it turned out, it seems quite important that merchants have a cardreader technology. Also the distinction between “owning” and “using” the technology may be important, especially in terms of surcharge rules.
61
2.4.4
Cost-based pricing: Simulation results
In the previous section we analysed the determinants of consumers’ choice of debit card versus cash and found that price variables are important determinants of that choice. In this section we will use the results from the estimated price elasticities to simulate what would be the impact of introducing cost-based pricing policies for both cash and cards. The way we precede is as follows. We first compute the ratio of debit card to cash transactions (in volume terms) using the estimated coefficients from table 2.4, column (4) and the sample means of the explanatory variables. This yields the ratio of debit cards versus cash as estimated by the model, i.e. 4.1 percent. We show this number in table 2.5 in the second column. In the next step we assume that cost-based pricing is introduced both for cash and debit cards, and we compute the effect of this pricing on the share of cards versus cash in volume terms. In order to do so we substitute the costs of cards and cash as obtained in chapter 1 into the equation. For the sake of simplicity here we interpret cost-based as ‘price equal to cost’, although this does not need to be the case in order to respond to the cost-based pricing principle as derived in section 2.3.2. We show the results in columns 3 to 5 in table 2.5. Column 3 and 4 show the result using the Belgian respectively the Dutch estimates of the resource costs of cash and cards; column 5 shows the results using an average of the Belgian and Dutch cost estimates. The results are quite spectacular. A system that would price the use of cards and cash based on resource costs would lead to an increase of the ratio of debit card to cash from 4.1 percent to 23-24 percent. Thus a switch to cost-based pricing would lead to a very large increase in the market share of debit cards with respect to current levels. As the last row in table 2.5 shows, currently the market share of debit cards in the 19 countries included in our database corresponds on average to 3.9 percent. Note that the market share of debit cards is defined as:
Number of card transactions _______________________________________________ Number of cash transactions + number of card transactions
62
The implementation of cost-based pricing policy would significantly reduce the share of cash transactions, such that debit cards would reach a market share of about 19 percent31 (based on the actual cost structure in Belgium and the Netherlands). This means that while in these countries the current usage rate of debit cards vis à vis cash is about four out of hundred transactions, cost-efficient pricing would lead to the use of debit cards in 1 out of 4 transactions. Having determined the potential shift in cash and card transactions’ volume due to efficient pricing, it is also interesting to compute how a switch to cost-based pricing would affect the market shares in terms of value. Unfortunately we cannot do this type of simulation because we do not have information on the value of cash transactions in the different countries. In any case we can expect something similar to what we found for the volume market shares.
Table 2.5: Cash and cards use with and without cost-based pricing
Debit Card volume / Cash volume Debit
Card
volume terms
2.4.5
Market
share
in
Assuming cost-based pricing
Current pricing structure as observed in our sample of 19 countries
Based on Belgian Costs
Based on Dutch Costs
Based on average of Belgian and Dutch costs
4.1%
22.9%
23.9%
23.4%
3.9%
18.6%
19.3%
18.9%
Cost-based pricing and resource costs
In the previous section we found that a cost-based pricing system would lead to a large increase in the market share of debit cards and a corresponding decline in the market share of cash transactions. The question that arises now is whether this shift in market share would affect the resource cost of the payment system. In other words, will this shift make the overall 31
Note that the resulting market shares of cash and debit cards sum up to 100 percent. In other words we assumed that retail payment instruments based on the current account are one market composed by cash and debit cards. As already mentioned in the introductory section, this assumption is valid throughout the whole paper and constitutes a simplification of reality in so far as it does not take into account of electronic purse schemes and cheques. However, we believe that such simplification has negligible implications on our results.
63
payment system more cost efficient? In order to analyze this question we use the change in cash and card usage that is generated by the implementation of cost-based pricing policies to measure the corresponding costs savings. Again, we can only do this for Belgium and the Netherlands - for which we have detailed information on payment services and the related costs -, and use these two countries as benchmarks. We proceed as follows. For each country we estimate the market shares of cash and debit cards in volume terms as the empirical model based on regression (4) of table 2.4 predicts. As shown in table 2.6, the model predicts that in Belgium 96 percent of transactions are made by cash while only the remaining 4 percent are made by means of a debit card. In the Netherlands cash is used slightly less: the market share of cash is 93.5 percent and the share of cards is 6.5 percent32. Then, we estimate the market share the model predicts when the observed prices (their sample mean) are substituted with the cost-based prices in these two countries. The result of this calculation is shown in table 2.6 (first and second row). We observe that following the introduction of cost-based pricing, the market share of cash transactions (in volume) would decrease from 96 to 81 percent in Belgium and from 94 to 78 percent in the Netherlands. In other words, cost-based pricing would imply a decrease of about 15 percent in the current volume of cash transactions in favour of debit cards. The use of debit cards would increase by a factor of 3 to 4. In a similar way we estimate the current and cost-based pricing market shares in terms of transactions’ value. As shown in table 2.6 (rows 3 and 4), the model predicts a very large change in value terms. The value of cash transactions would be reduced from 73 to 34 percent of the total in Belgium and from 76 to 45 percent in the Netherlands, implying that, according to our empirical model, in both countries debit card use would double in value terms with respect the current level.
32
Notice that these market shares are the one the model predicts on the basis of mean values and intercepts for Belgium and the Netherlands, which are different from the real market shares in these countries. In Belgium, the real market share of cash and cards are 91 and 9% in volume terms and 56 and 44% in value terms. In the Netherlands, the real market share of cash and cards are 87 and 13% in volume terms and 59 and 41% in value terms.
64
Table 2.6: Potential gain in resource costs Belgium
Netherlands
Cash
Debit card
Cash
Debit card
Market share predicted by model (volume)
95.6%
4.4%
93.5%
6.5%
Cost-based pricing Market share as predicted by model (volume)
80.6%
19.4%
78.3%
21.7%
Market share predicted by model (value)
73.4%
26.6%
75.5%
24.5%
Cost-based pricing Market share as predicted by model (value)
34.4%
65.5%
43.4%
56.6%
Gains in resource costs (Million euros)
211
147
Gains in resource costs (% current resource costs)
11.8
5.6
The last step in the analysis is to compute the effect of these changes in the usage of cards and cash on the resource costs of these payment services. In the first chapter we extensively discussed the nature of resource costs of payment services, which can be divided in three categories: fixed costs, variable costs depending on the volume of transactions and variable costs depending on the value of transactions. We use these different cost elements as estimated in chapter 1 for Belgium and the Netherlands, and apply them to the market shares (volume and value) obtained under cost-based pricing (rows 2 and 4 of table 2.6). In particular the way this is computed is as follows. From the two upper panels of table 2.6 we can see that, for instance in Belgium, the model predicts a change in the market shares of cards and cash due to the introduction of cost based pricing equal to 15 percent (in volume terms) and 39 percent (in value terms)33. Therefore, we add (subtract) these obtained changes to the real market shares of cards (cash) in the country and estimate the corresponding resource costs34. We then compare these estimated resource costs under cost-based pricing with the currently prevailing resource costs, and compute the cost saving realized by cost based pricing. The results are shown in the last two rows of table 2.6.
33
We follow the same procedure for the Netherlands. The real market shares of cards in volume terms are 9 and 13 percent in Belgium and the Netherlands respectively. The real market shares in value terms are 44 and 41 percent.
34
65
We find that the gain in resource costs associated with the introduction of cost based pricing would amount to more than 200 million euros in Belgium and about 150 million euros in the Netherlands. Putting these numbers into perspective, these cost savings in Belgium correspond to 12 percent of current resource costs of cash and debit cards together, while in the Netherlands these cost saving represent 6 percent of actual resource costs. The gains we have computed in table 2.6 can be called the static cost gains. They do not take into account any changes in the cost structures that cost-based pricing generate. In particular, they do not take into account that the shift in card use induced by the introduction of more efficient pricing policies will generate economies of scale within the card payment system. These economies of scale are likely to lead to additional cost reductions and thus to a further decline in the resource costs of payment services. Recently Brits and Winder (2005) from the Dutch National Bank estimated the costs of the different payment instruments and, on the basis of the distribution of fixed and variable costs, derived the most efficient payment instrument for different transaction values. As we already discussed in chapter 1, they find that although most small value transactions are made by cash, they should more efficiently be replaced by transaction with e-purses and, to a smaller extent, with debit cards. On the basis of this finding, they also estimate the possible cost savings from moving away from cash. In particular, they assume that 500 million small-value cash transactions are replaced by e-purse transactions and 1 billion mid-value cash transactions are replaced by debit card transactions. This scenario for a shift from cash to cards would lead to around 100 million euros in costs savings. While these findings are certainly very interesting, the approach followed by the Brits and Winder is based on an artificial shift that does not take into account of the possible response of the consumers (i.e. their price elasticity). In this sense our approach is more complete because it takes into account the price sensitivity of consumers with respect to payment services. The same conclusion is valid with regard to the costs savings estimated by the National Bank of Belgium (2005). These costs savings are obtained with the same methodology used by Brits and Winder, but assuming the replacement of a different number of cash transactions of a given value (500 million transactions made with debit cards and 250 million with e-purses). Also in this case the shift from cash to cards is exogenously imposed without considering how it can (or cannot) be achieved in practice, hence, disregarding the role played by the stakeholders in driving or opposing such a shift.
66
BOX C: European-wide benefits from implementation of efficient pricing In this section we extend the previous analysis to investigate the possible impact of the implementation of cost-based pricing for payment instruments to other European countries. Gaining some knowledge of the resource costs of payment instruments at the European level is particularly relevant in the light of the creation of a Single Euro Payment Area (SEPA) that will become effective in a few years. The SEPA will bring large benefits, but also large costs at the European (Euro Area) level. However it is extremely difficult to estimate those benefits and costs in monetary terms because the exact usage and the resource costs of the current retail payment system are not known at the European level. As we already mentioned, because of the different organisation of payment systems and the institutional framework in the different countries the studies of the costs of payment systems are typically restricted to single-country. For this reason we cannot directly extend the analysis from the previous section to the Euro Area. However we believe that even a broad estimation can be useful to give an indication of the order of magnitude of the possible impact of the implementation of transparent and cost-based pricing. In order to extend the analysis we necessarily have to make some assumptions, which we base on the available information from Belgium and the Netherlands, the two countries that we used as benchmarks. We follow the same methodology used before, proceeding by stages. In a first stage, we use the volume ratio of cards to cash usage as estimated by our econometric model to obtain the market shares of cash and cards in volume terms and in value terms. We then compare the latter shares with the corresponding share obtained from the same model but under the assumption of cost-based pricing for both cash and cards. The difference between the two market shares - before and after the policy change - is then added to the real market shares for the full sample in 2003. The market shares referring to the average of the 19 European countries of the panel are illustrated in the table below (table B1). In these countries, the introduction of cost-based pricing would lead to a decrease in cash usage from 96 to 81 percent of all transactions. At the same time, even though the reduction in terms of number of cash transaction is not very large, in value terms this means
67
approximately halving the current value of total cash usage (from 78 percent to 38 percent of the total value of retail transactions35).
Table B1: Implementing cost-based pricing in Europe36 Average results for panel of 19 European Countries Cash
Debit card
Observed market share (volume) in 2003
93 %
7%
Observed market share (value) in 2003
66 %
34 %
Market share as predicted by model (volume)
96 %
4%
Market share as predicted by model (value)
78 %
22 %
Market share with cost-based pricing as predicted by model (volume)
81 %
19 %
Market share with cost-based pricing as predicted by model (value)
38 %
62 %
From a methodological point of view, it is important to recall that the volume of cash transactions in the different countries of the dataset is not observed, but, rather is derived from a common assumption, as explained in section 2.2. In addition, it should be noted that in order to derive the market shares of cash and cards in value term we needed information regarding the average value of cash and card transactions. In the case of cards we could derive this from the available statistics (euro 54.8), while in the case of cash we had to make some assumptions, since the total value of cash transactions is unknown for the countries considered. We assumed the mean value of a cash transaction to be equal to the average of the Belgian and Dutch mean values, i.e. euro 7.837. This number could possibly be smaller than the actual one, given that some of the countries in our database are largely cash based. However since the same assumption is made both for the observed market share in value terms and the market share of cash and cards after implementation of cost-based pricing and 35
Notice that these market shares implicitly assume that only cash and debit cards are used in retail transactions. The countries are Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, United Kingdom, Czech Republic, Estonia, Hungary, Poland and Switzerland. 37 In Belgium the mean value of a cash transaction is euro 6.2 and in the Netherlands euro 9.4. 36
68
that since it is the difference between the two that matters, the potential bias is likely to be negligible. In a second stage, we use the newly obtained market shares of cash and cards to estimate how the fixed and variable costs connected with the cash and debit card payment systems would change following the introduction of cost-based pricing. Once more, because of a lack on information we have to make some assumptions in order to carry out the simulation. In particular we do not know the value of the total resource costs of payment instruments and the proportions of fixed and by variables costs. For the variable costs that depend on the volume and on the value of transactions we take the weighted average of Belgian and Dutch variable costs. The fixed costs are assumed to be equal to the average of the fixed costs in Belgium and in the Netherlands. We make the same assumptions both in the case of cash and of cards. All the remaining numbers we used to carry out the simulations are the sample means of various variables included in our dataset relative to 2003. On the basis of this exercise we estimate the current resource costs of cash and cards to be equal to 1.22 percent of GDP (see table B2). However the introduction of cost based pricing would lead to savings amounting to 0.14 percent of GDP, which is not a negligible amount. Note that our simulation is based on the mean value of the nineteen countries of the dataset; therefore the estimated total resource costs in million euros should be interpreted very carefully. They should be considered as referring to an average of the nineteen European countries and not to the total of the nineteen countries (the same is valid for the cost savings in million euros). Putting these numbers in perspective, we can consider that for the whole euro area the costs saving would amount to approximately 10 billion euros38.
38
In 2003 the GDP of the Euro Area amounted to 7300 billion euros (source: ECB statistics).
69
Table B2: Resource costs of payment instruments in Europe
Average results for panel of 19 European Countries Cash + Debit cards
Estimated current resource costs (total, 2003)
Gains in resource costs from cost-based pricing
2.5
Million euros
6376
% GDP
1.22
Million euros
750
% GDP
0.14
Conclusion
The absence of transparent and cost based pricing in the provision of payment services, and in particular the (almost) free supply of cash to the consumer, leads to important inefficiencies. In a first part of this chapter we analyzed the nature of these efficiencies. We then moved on surveying the theoretical literature for what it has to say about the optimal pricing in the provision of payment services. We find that in order to maximize social welfare the ratio of the prices of cash and card services should be equal to the ratio of their costs. At the same time, both prices should be in the same relation to the respective costs. This suggests, therefore that similar principles should guide pricing of the different payment services and that pricing should take cost as a reference. In a second part we estimated an econometric model that explains the shares of debit cards versus cash. We used discrete choice theory to specify this model. The results of this econometric analysis are that consumers react in a significant way to changes in the cost of cash and cards. In particular, an increase in the price of cards leads to a significant decline in its usage. Similarly an increase in the price of cash (as paid by the consumer) leads to a significant decline in its usage. 70
We used the results of this econometric analysis to compute the effects a cost-based pricing system would have on the use of cards and cash in the EU-countries. We find that these effects are quite large, i.e. a switch to cost-based pricing would increase the market share of debit cards from its current level of 4% to 19%. Finally, we used these estimates of the effects of cost-based pricing to analyse the question of how cost based pricing would affect the resource costs of the payment services provided by cards and cash. Due to a lack of data we could make this analysis for Belgium and the Netherlands only. We found that the introduction of cost-based pricing would lead to a reduction of resource costs of €150 to €200 million in these countries. Trying to extrapolate this analysis at European level, whenever it is assumed that the European countries have a cost structure of resource costs of payment instrument similar to Belgium and the Netherlands, we estimate the possible saving from the implementation of cost-based pricing as amounting to 0.14 percent of GDP.
71
CHAPTER 3 The Shadow Economy and Card Usage
3.1
Introduction
The “shadow economy” continues to be of great importance. The anonymity provided by cash ensures that this form of payment will continue to be the privileged one for transacting in the shadow economy. The question that arises then is the extent to which the movement towards the use of cards can contribute to reducing the importance of the shadow economy. In this part of the research project we want to analyze, first the importance of the black economy in the industrialized world. We will use information that has been compiled by Friedrich Schneider (2005). Second, we will go into the issue of how the substitution of cash by cards can contribute to reducing the importance of the black economy. This will then also allow us to shed light on the possible budgetary gains for the governments. In the remainder of the chapter when referring to cards we will mean only debit cards, which are the most common alternative to cash in Europe. Although most of our discussion could be extended to other types of cards, we will not do this here as such an extension would go beyond the scope of the study. In the next section we define what is meant by black economy, and we discuss the different methodologies used in the economic literature to quantify the shadow economy. Next, we develop a simple theoretical framework to analyse the choice of paying by card versus cash in the light of the existence of a hidden economy. On the basis of this framework, in section 3.4 we estimate the empirical impact of the shadow economy on the use of payment instruments.
72
In section 3.5 and 3.6 we use the prediction of the empirical analysis to estimate the benefits of a reduction in the shadow economy rate. We formulate conclusions in the section 3.7.
3.2
What is the shadow economy
The “shadow” economy is a well known and widespread phenomenon. Every year, in the countries around the world, an important share of payments takes place without ever being recorded or taxed. The real size of this phenomenon, however, may depend on several factors like the type of institutions in place, the degree of social/state trust, the level of taxation, the incentives or disincentives to act illicitly. In order to discuss these issues it is necessary to define what is meant by shadow economy. Very often when talking about the shadow economy people think of illicit and criminal activities like drugs or prostitution. In reality the shadow economy is much more than that. In economic research typically the shadow economy is defined as consisting of all the unregistered economic activities that contribute to Gross National Product. In this study we base our analysis on the work of Schneider (2005), who estimated the size of the shadow economy in a very large set of countries (about 120 countries) using a unique and advanced technique. Therefore we will adopt his definition of the shadow economy. Schneider considers as shadow economy “all market-based legal production of goods and services that are deliberately concealed from public authorities (i) to avoid payment of taxes; (ii) to avoid payment of social security contributions; (iii) to avoid having to meet certain legal labour market standards; (iv) to avoid complying with certain administrative procedures”. This definition, therefore, excludes from the analysis criminal illegal activity and the informal household economy. In general, the tax and social security burden seem to be the two main causes behind the existence of the shadow economy. The incentive to remain in the shadow is higher the larger the difference between the total cost of official labour and the after-tax earnings from work. However even large tax reductions might not lead to a proportional reduction in the shadow economies because there might be important costs of switching from the shadow to the official activity. An implication of the existence of the shadow economy is a loss of potential revenue for the government, which may have an impact on the quality and quantity of public services provided to the public. If, in order to compensate for this loss, the government raises the tax 73
rate on the official economic activity it actually increases the incentive to participate in the shadow economy. Johnson et al. (1998) analyse this issue and find that higher tax revenues are associated with smaller size of the shadow economies with smaller tax rates and lower regulatory burden. Moreover, the countries where tax revenues are used to establish a better rule of law are characterized by a low shadow economy. This is generally the case of European Union countries. Conversely, in countries in transition, like for instance Russia, there might be a bad equilibrium between tax revenues and shadow economy because of a heavy and discretionary regulatory burden on firms and high share of activity within the unofficial economy. In this study we put forward the argument that part of a government strategy to reduce the size of the shadow economy could be to favour the use of card payments so as to reduce the incentives to act within the hidden economy. Such a strategy would also have the effect of increasing government revenues.
3.3
How to measure the shadow economy
Measuring the size of the shadow economy is certainly not an easy task. In the economic literature several methods have been used for this purpose. In particular, three methods appear to be quite common. We briefly review them in this section. A first approach is the direct method, which is usually based on voluntary surveys. This is a useful way of measuring the shadow economy. However, like in other survey-based analysis, the issue arises of the interpretation of unanswered questions and the fact that people might not give true answers on sensitive matters. Alternatively, the direct approach can also be based on an analysis of the difference between fiscal declarations and selective fiscal controls. This procedure, however, suffers from the fact that usually fiscal controls are not based on a random sample.
Another drawback of the direct procedure is that it only allows for
estimating the shadow economy at one particular moment in time and not how it evolves. A second approach is indirect and is based on macroeconomic and other economic indicators. One common type of indirect method is based on the comparison of different statistics, e.g. a comparison between income statistics at the national level and expenditure from the national account statistics. If expenditures are systematically larger than income, this can be an indication of hidden activity. One drawback of this approach, however, is that the discrepancy
74
between these two statistics is not only due to the shadow economy but also to statistical errors. Similarly, one can measure the shadow economy by examining the statistics of labour force and labour participation. If labour force is assumed to be constant, then a decrease in labour participation can be caused by a rising shadow economy. However this could also be due to other factors. A third way of estimating the shadow economy is the currency demand approach. This approach is based on the estimation of a demand for currency depending on income, interest rates and the tax rate. This approach was first introduced by Tanzi (1983) and the idea behind is that, assuming that the shadow economy is a cash-based activity, any excess increase in currency in circulation that cannot be explained by conventional factors is to be attributed to the shadow economy. The main criticism raised to this approach relates to the fact that not all shadow economy related transactions are cash-based; some are based on barter. This criticism is particularly relevant when considering developing countries. However, in industrialized countries non-cash transactions related to the shadow economy are likely to be marginal. Nevertheless, a study by Isachsen and Strom (1985) based on a survey revealed that in Norway only about 80 percent of hidden transactions were paid by cash, which leaves an important share of transactions unaccounted for. Another limitation of the demand for currency approach is that it intrinsically links the shadow economy to the tax rate. However, while other factors might be relevant in explaining hidden activity (like regulation) they might be less measurable. Moreover, the Tanzi method is often based on the implicit assumption that the velocity on money in the official and in the hidden economy is the same. This may or may not be the case. All the methods described above are based on one particular indicator that captures the effect of the hidden economy, i.e. the labour market, money demand or expenditure, and consider tax evasion as the main driver of the hidden economy. However the shadow economy is likely to have several effects. As a response to this criticism the model approach has been developed. This approach takes into account several possible causes and various potential indicators of the hidden economy at the same time. The empirical methodology used to measure the shadow economy, in this case is the theory of unobserved factors, such that the factor-analytical procedure estimates the hidden economy as an unobserved (latent) variable. This procedure is called DYMIMIC39 model. 39
Dynamic multiple-indicators multiple-causes.
75
The DYMIMIC model approach as explained by Schneider (2005) works as follows. The procedure is composed of two parts. The coefficients of the model are estimated in a set of structural equations within which the unobserved variables cannot be directly derived. First, a measurement model links the unobserved variables to observable indicators. Then, the structural equations model specifies the causal relationships among the unobserved variables. In particular, there is one unobserved variable, i.e. the size of the shadow economy. Thus, the size of the shadow economy is assumed to be influenced by several indicators able to capture the structural dependence of the shadow economy on variables that can be used as predictors of its development overtime. Figure 3.1 clarifies the way the model works. The causal variables Zit (i=1,2,…k) determine the development of the shadow economy Xt at time t which, in turn, is reflected in the indicators Yjt (j=1,2,…p).
Figure 3.1: Structure of DYMIMIC model
CAUSES
Xt-1
INDICATORS
Z1t
Development of the
Y1t
Z2t
shadow economy over time
Y2t
…
Xt
Zkt
… Ypt
Source: Schneider (2005), Fig. 6.1.
Let us now discuss the different causes and indicators used in the model. Three factors causing the development of the shadow economy are usually considered in the literature. First, the burden of direct and indirect taxation, being actual of perceived. This burden increases the incentive to participate in the shadow economy. Second, the burden of regulation, which also has a negative impact on the unknown variable. Third, the ‘tax
76
morality’, as Schneider defines it, which refers to the attitude the citizens have with respect to the state. A declining tax morality will tend to increase the participation to the shadow economy to the detriment of the official activity. The main indicators are the movements of monetary aggregates, the development in the labour market and in production. From this Dymimic procedure one can obtain a time series index of the hidden-to- regular output. Then, the estimation of the cash demand model allows converting the index from the Dymimic approach into percentage units. In fact, the cash demand model provides the longrun average value of hidden-to-regular output so that the index for this ratio estimated by means of the Dymimic approach can then be used to obtain the level and percentage ratio of the shadow economy. The combination of these two methodologies, therefore, allows overcoming several problems related to the estimation of the shadow economy. Schneider (2005) uses the approach just described to estimate the size of the shadow economy overtime in a very large set of countries among which also most OECD countries and he estimates this for a few points in time. In particular for the group of OECD countries, Schneider uses the following causal variables: share of direct and indirect taxation and of social security payments (as percentage of GDP), burden of state regulation, unemployment, tax morale and GDP per capita. The indicator variables he uses are the unemployment rate, the average working time per week, the rate of change of GDP and the change in currency in circulation per capita. Among the seven causal variables the ones which appear to be quantitatively more relevant are the tax and the social security burdens and the tax morale. Among the indicators variables, employment and the change in currency per capita are the most relevant. From this latent factor analysis the author combines the size of the shadow economy as obtained from the currency demand approach, which gives the size of the shadow economy as percentage of GDP, with the DYMIMIC approach. For our empirical analysis we use his estimation of the size of the shadow economy. Below we report a table with the estimated sizes of the shadow economy by country as estimated by Schneider (2005). We note that the size of the shadow economy (expressed as a percent of GDP) is relatively high in the European countries, with the lowest rates observed in Austria and Switzerland and the highest in Greece, Italy, Portugal and Spain and, to a smaller extent, Belgium. Between the second half of the 1990s and the beginning of 2000 the general tendency has been a small decline of the shadow economy. Surprisingly enough, the two
77
countries with the smallest sizes and Germany experienced an increase in the shadow economy according to these estimates, while France saw some up and down movement.
Table 3.1: Size of the shadow economy as % of GDP Average
Average
Average
Average
1997-1998
1997-2000
2001-2002
2002-2003
Austria
9.0
9.8
10.6
10.8
Belgium
22.5
22.2
22.0
21.5
Denmark
18.3
18
17.9
17.5
Finland
18.9
18.1
18.0
17.6
France
14.9
15.2
15.0
14.8
Germany
14.9
16.0
16.3
16.8
Greece
29.0
28.7
28.5
28.3
Netherlands
13.5
13.1
13.0
12.8
Norway
19.6
19.1
19.0
18.7
Italy
27.3
27.1
27.0
26.2
Portugal
23.1
22.7
22.5
22.3
Spain
23.1
22.7
22.5
22.3
Sweden
19.9
19.2
19.1
18.7
Switzerland
8.1
8.6
9.4
9.5
UK
13.0
12.7
12.5
12.3
Ireland
16.2
15.9
15.7
15.5
Source: Schneider (2005). The estimates are based on the currency demand and Dymimic methods.
We will use these estimates for the empirical analysis in section 3.4, but before doing so, we present a simple model linking the shadow economy to the choice of payment instruments. This model will provide a useful theoretical background for the estimations and simulations that will follow.
78
3.4
The theoretical model
Consider the following economy. Consumers live along a Hotelling line of length 1 (see figure 3.2). At the extremes (that is in 0 and 1) two shops are located selling a particular good of which the consumers buy one unit per unit of time. These shops compete by posting the price for which the good sells. We will denote these merchants by the subscripts L and R (Left and Right respectively).
Figure 3.2: The Hotelling (linear-city) model
ML 0
MR
x
1
ATM / Bank
In order to reach the shop, the consumers have to travel. Per unit of distance travelled, they incur a cost equal to t. The two shops are located in different tax districts. In the left district, tax controllers are unable to reconstruct the sales of a shop because it sells goods which are obtained from a supply chain that remains undercover, and payments are executed using cash. So, the shopkeeper who wants to minimize his tax contribution forces his consumers to pay by means of cash. In order to avoid that consumers incur an extra cost from this, we assume an ATM or a bank branch is located in 0, so that consumers can withdraw cash on the spot. The merchant located in 1 cannot attract goods from a black economy supply chain. Therefore there is no reason why he would refuse cards. Hence he accepts them, and therefore card usage in this shop only depends upon consumer tastes for using cash and cards. We denote the fraction of the population that uses a card by u. Therefore, without tax distortions, the ratio of cards to cash r is given by:
79
TRUE =
u =r 1- u
(3.1)
This ratio can also be considered as the true ratio of cards to cash transactions. If we denote the indifferent consumer by x, that is the consumer who needs to travel a distance of x to reach the left shop and a distance (1-x) to reach the right shop, we can easily show that the actual ratio of cards to cash is different. As a matter of fact, we have:
ACTUAL =
u (1 - x ) u (1 - x ) = x + (1 - u )(1 - x ) 1 - u(1 - x )
(3.2)
where the distance x corresponds to the demand of the left-merchant, i.e. the sum of all consumers buying from the merchant who only accepts cash, while (1-x) corresponds to the demand of the right-merchant, i.e. sum of all consumers buying from the merchant who accepts cards. We assume that both shopkeepers can acquire the goods at the same price and normalize it to be equal to 0. The tax controller in the right district charges a tax equal to g per unit sold (alternatively one can consider ad valorem taxes or corporate income taxes). The tax controller in the left district cannot collect anything. However, rather than doing his job properly, he can tax the left shopkeeper by benchmarking him to the right shopkeeper, and charge the same amount of taxes. This would be a “lump-sum tax”, a common practice. This assumption is not necessary; however leaving the left tax controller without any means to obtain tax money only would strengthen our results. It is now possible to show the following:
Proposition 1: In equilibrium, prices are respectively shares are equal to
3t + g 6t
g (3t − g ) 3t
t+g 3
on the left and
going to the left shopkeeper, and
3t − g 6t
t +2 g 3
on the right, whereas market
to the right. Government collects taxes
(remember the lump-sum tax inflicted upon “left”), whereas it should obtain
g (the market remains covered so everybody buys so 1 times g=g). Hence, the tax loss is g2/3t. Percentage wise this is g/3t, or LOSS=g/3t.
80
These results allow us to prove the following proposition:
Proposition 2: As the per unit transport cost t increases, the ACTUAL CARDS USAGE ratio increases whereas the TAX LOSS ratio decreases. At the same time, increases in the per unit tax g increases the TAX LOSS RATIO while it decreases the ACTUAL CARDS USAGE ratio. In other words, tax losses go hand in hand with reduced usage of cards. An implication of Proposition 2 is that because of the existence of hidden economic activity, the share of cash payments is likely to be higher than it would be otherwise. In this sense we can think of the share of cash payments as consisting of two components. First, there is a given proportion of cash payments that derives from consumers’ preferences and merchants’ supply of different payment facilities, i.e. it is the result of the combination of merchants’ and consumers’ preferences. We can call this the ‘preferred use of cash’. Second, besides the ‘preferred’ cash use there is also the ‘induced’ cash use, where by ‘induced’ we mean all the transactions that are explicitly made by cash with the intention of exploiting its anonymity such as to leave the transaction unrecorded and hence within the hidden economy. Therefore we have a situation like the one illustrated in Figure 3.3. Figure 3.3: Nature of cash usage
“INDUCED”
=
TRUE CASH USE
=
“PREFERRED” CASH USE IN THE OFFICIAL ECONOMY
+
CASH USE FOR PURCHASES IN THE SHADOW ECONOMY
Going back to our model, given that we have two shops it results that one is in the official economy while the other one is the shadow economy. It follows that the respective markets
81
shares of the two shops correspond with the size of the regular and shadow economy respectively. In other words, we can rewrite the ratio of cards to cash transaction (3.2) as
CARDS USE u (size OE) = , CASH USE (1 − u )(size OE) + (size SE)
(3.3)
where OE is the official economy and SE is the shadow economy. The advantage of this formulation is that it can be tested empirically. We will do this in the next section.
3.5
An empirical application of the model
In this section, we estimate a model of how consumers choose between payment instruments. We are particularly interested in the role played by the shadow economy in affecting consumers and merchants’ choice of payment instruments. To do this, we follow the procedure explained in the previous chapter where we used a discrete choice theory approach to derive the preference for debit card payments as opposed to cash payments as the result of several determinants, like the payments’ infrastructure and their costs. As explained in the previous chapter a discrete choice is any choice made from among a limited range of mutually exclusive options. The basic principle of these models is utility maximization where each individual tries to make choices maximizing his well-being. Note that in a discrete choice model, the level of utility is not identified. Only the relative utility is identified, which is approximated by the market share. In our case this leads us to the testing the following relationship: DebitVol CashVol
= α + β1
POS Merch
+ β2
Cards pop
+ β3
DebitFees pop
+ β4
ATMFee pop
+ β5 SEratio + µ
(3.4)
Note that equation (3.4) corresponds to equation (2.10) in the previous chapter except for the inclusion of an additional term, i.e. the “SEratio”. The latter is the share of the shadow economy (SE) over of the official economy or, formally, it equals SE/(1-SE). Note that SE is
82
defined as the share of the shadow economy as a percentage of GDP. As this series is not available in official statistics, it is obtained from the study by Schneider (2005) (see table 3.1 for country data). The other factors influencing the choice of debit cards relative to cash are the number of POS over the number of merchants, the number of debit cards per capita, the annual fees paid on debit cards per capita and the annual fees paid on ATM withdrawals per capita. A discussion on the impact played by these variables is to be found in the previous chapter40. Intuitively, it could be argued that there might be high negative correlation between the POSover-merchant variable and the shadow economy ratio to reflect the idea that in the countries where the shadow economy is more widespread, merchants could be more reticent to invest in an payment technology alternative to cash. We checked the correlation between the two variables for an assessment and we noticed that it is true that there is negative correlation, but that it does not appear particularly high (-0.34). We estimate equation (3.4) for a panel of 13 European countries41 using the fixed effect estimation technique for panel data. Unfortunately, we could not extend the analysis to a larger set of countries because we did not dispose of data for the shadow economy that vary over-time during the sample period (a necessary condition to run the fixed effect estimation). The results of the estimation are shown in the following table 3.2. From table 3.2 it can be observed that the coefficient for the first four variables listed in the first column are very close to the coefficients obtained for the same variables in the previous chapter. As a result, the conclusions are similar and we will not detail them here. The interesting part of the results here is that the ratio of the shadow economy over the regular economy has a strong and negative impact on the choice of a debit card payment relative to cash. This means that the larger the proportion of hidden activity in a country, the more people will use cash for transactions at the expenses of cards. Conversely, when the proportion of activity in shadow economy decreases because of political action, regulatory or cultural changes and so on, the rate of cash usage in retail transactions will also decrease and the volume of cards payments increase.
40
The previous chapter also provides details on the exact series used. Austria, Belgium, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, UK.
41
83
Table 3.2: Debit card usage Dependent variable: Debit over Cash volume POS over Merchants
0.013*
(0.008)
Debit cards per capita
0.059***
(0.011)
Total Price of Debit per capita
0.002**
(0.000)
Price ATM per capita
0.001**
(0.001)
SE ratio
-0.816***
(0.222)
constant
0.143***
(0.047)
FE (Belgium)
-0.033
FE (Netherlands)
-0.025
N = 63 R2 = 0.72 Notes: i) ***, ** and * indicate respectively 1%, 5% and 10% significance rate. The standard errors are in parenthesis. ii) The table shows only two of the country fixed effects. However it should be noted that the fixed effects estimates sum to zero and should be interpreted as deviations from the overall mean. The estimates of the fixed effects, however, do not have reported standard errors since the program treats them as nuisance parameters for the purposes of estimation.
This result, therefore, provides strong evidence for the existence of what we labelled as ‘induced’ cash use in box D. In the next section we will use this result to evaluate the impact of a policy change on cards and cash volume of transactions. Notice also that in table 3.2, the coefficient for the constant should be interpreted as an average country effect that cannot be captured by the other explanatory variables included in the regression. Thus, the fact that the constant is significant and with positive sign suggests that there are some country factors not observable within our explanatory variables that favour the debit-cash substitution. Finally, the last two rows of table 3.2 display the coefficients of the specific fixed effect for Belgium and the Netherlands that have to be interpreted as deviations from the sample mean and will be useful for the simulation analysis in the next section.
84
3.6
Reduction of the shadow economy and impact on card usage
In the previous section we analysed the impact of the size of the shadow economy on the determination of the card to cash volume ratio. In this section we evaluate the impact of a policy change aimed at reducing the size of the shadow economic activity on the relative usage of cash and cards for payments. In order to do so we proceed in a way similar to the methodology used in the previous chapter. First, we derive the ratio of cards to cash in volume terms as our empirical model presented in section 3.4 predicts for the sample of 13 European countries. We obtain this ratio from the estimated coefficients (reported in table 3.2) and the sample means of the respective explanatory variables. As a result of this calculation we obtain a ratio of debit to cash volume of transactions estimated by the empirical model (basis scenario), which is equal to almost 2 percent (see table 3.3).
Table 3.3: Cash and cards use with and without SE reduction Debit Card volume Cash volume
Base scenario
Halving SE rate scenario
Observed in sample of 13 European countries
1.6 %
Observed in Belgium
14 %
Observed in the Netherlands
14 %
Estimated for sample of 13 European countries
13 %
Estimated in Belgium
22 %
Estimated in the Netherlands
22 %
As we discussed in section 3.3, the proportion of the shadow economy can vary considerably over time. The main reason for a change is likely to be due to the implementation of some targeted political action, for instance, a convincing reform of the tax structure and/or level, a more determined enforcement of regulation or an intensification of fiscal investigations. Therefore, in a second stage we assume that the government implements policies that lead to a decrease by half of the size of shadow activity as percentage of GDP (Halving SE rate scenario). Hence, we substitute this change in the size of the shadow economy (SE) into our model and compute how this affects the ratio of debit to cash volume. We find that for our 85
sample the policy change increases the volume ratio of card to cash transactions from almost 2 percent to 13 percent. Table 3.3 also shows what the impact is of a policy change aimed at halving the hidden economy on the relative use of cash and cards in Belgium and the Netherlands. The second and the fifth rows show that using the estimated coefficients in combination with Belgian sample means and fixed effect, the model predicts a ratio of debit to cash volume equal to 14 percent given the current shadow economy rate42 while when the ratio of the shadow economy is halved the corresponding ratio jumps to 22 percent. This same policy change in the Netherlands produces a similar shift43. It is not surprising that the shift from the observed scenario to the scenario involving halving the shadow economy rate is so much larger for the full panel made of 13 European countries than for the single cases of Belgium and the Netherlands. This is due to the fact that debit cards usage is by far more important in the Benelux countries than in the average of the 13 countries considered in the empirical analysis.
3.7
Shadow economy and costs of payment instruments
An important motivation for taking strong political action aimed at reducing the hidden economic activity is to raise government revenues, or more precisely, to reduce government losses related to tax collection. Estimating such a gain for the government would be beyond the scope of this paper, but from our theoretical model discussed in box D there are good reasons to think that it can be quite important. Besides the higher tax revenues, another favourable implication of such political action is a reduction in resource costs of the payment system. Hence, in this section we will use the results from the previous section to estimate the possible reduction in the resource costs of cash and debit cards. As was the case for the simulations in chapter two, we are only able to compute the possible gains in terms of resource costs of payment instruments for Belgium and the Netherlands because only for these countries we dispose of detailed estimates of the amounts of resource costs and their nature (fixed or variable). 42
Note that this ratio is quite close to the real value of this ratio for Belgium, equal to 10 percent (based on 1998 figures from De Grauwe et al. (2000)). 43 The effective debit to cash ratio in the Netherlands is equal to 15 percent (based on 2002 figures from DNB (2004)), thus extremely close to the one predicted by our empirical model.
86
The procedure to carry out these estimations requires first, to derive the market shares of cash and cards in volume and value terms as the empirical model discussed in section 3.5 predicts. The definition of market share in volume terms and in value terms that we use are as follows:
D Market share (volume) =
Volume of D transactions Volume of D transactions + Volume of C transactions and,
D Market share (value) =
Volume of D trans. * Mean Value D Volume of D trans. * Mean Value D + Volume of C trans. * Mean Value C
where D and C stand for debit cards and cash respectively. To obtain the market shares for cash one can just take the same expressions and substitute the D with the C and vice versa. These market shares are displayed in table 3.4. In terms of volume, the model predicts a market share for cards equal to 12 percent in Belgium and 13 percent in the Netherlands. In value terms, the debit cards market shares are respectively 52 and 40 percent. The large difference between the value and the volume market share has to be imputed to the fact that the average transaction value for debit cards is much larger than for cash. Thus although the number of debit transactions is only a fraction of the number of cash transactions, in terms of value they amount to a very large share of retail transactions, at least in these countries. The next step consists in estimating the same market shares predicted by the model under the assumption of a reduction by half of the shadow economy rate. As shown in table 3.4, the volume market share of debit goes from 12 or 13 percent to 18 percent. As far as the value of transactions, the market share of cards increases to 59 percent in Belgium and to 51 percent in the Netherlands. In the next step, we use these changes in the market share of cash and cards as resulting from the reduction of hidden economic activity to compute the impact on the fixed and variable resource costs of cash and cards44. We find that the policy induced decline in the shadow economy decreases the costs of payment instruments by 40 million euros in Belgium and 52
44
For a detailed discussion on resource costs see chapter 1.
87
million euros in the Netherlands45. In terms of percentage of current resource costs this means a reduction of about 2 percent. Note that this cost reduction does not imply a reduction of economic activity at an aggregate level, therefore, the overall value and volume of transactions (cash plus debit card) is unchanged. Table 3.4: Potential gain in resource costs Belgium
Netherlands
Cash
Debit card
Cash
Debit card
Market share predicted by model (volume)
88 %
12 %
87 %
13 %
Market share with halving SE as predicted by model (volume)
82 %
18 %
82 %
18 %
Market share predicted by model (value)
48 %
52 %
60 %
40 %
Market share with halving SE as predicted by model (value)
41 %
59 %
49 %
51 %
Gains in resource costs (Million euros)
40
52
Gains in resource costs (% current resource costs)
2.3
2.0
Some additional discussion of these results is necessary. First, the difference in the market shares of cash and cards due to the decrease in the shadow economy is only to be imputed to a shift of a share of the economic activity from hidden to official. This means that people’s preferences vis à vis cash and cards are not affected, but only the ‘induced’ cash use is affected (see Figure B3). Second, although the decrease in the market shares of cash is quite important, the large majority of retail transactions would still take place in cash (82 percent), because the simulation assumes that preferences of the agents in the regular economy are unchanged. As a result, the costs associated with cash cannot decrease by a large amount.
45
These numbers are obtained in the following way. From the model we obtained the change in the market shares in volume and value terms produced by policy action. Then we add this change to the real market shares in the two countries and estimate the relative changes in resource costs.
88
Third, as already mentioned, the social benefit of smaller resource costs related to payment instruments will be associated with larger government income from fiscal revenues, which however is not considered here. Thus, the efficiency gains shown in the bottom part of table 3.4 represent only a fraction of total gains from a decrease in the size of the shadow economy.
Box D: Resource costs at European level As in the previous chapter, we make an extrapolation of the Belgian-Dutch analysis of the savings in resource costs of payment instruments to euro area level. We use the same methodology. We first derive the volume and value market shares for cash and debit card payments. The assumption made regarding the mean value of cash needed to obtain the value market share of cash is similar to the one made in the previous chapter. Thus, the mean value of cash transactions used is again equal to 7.8 euro, while because of the smaller set of countries included in the analysis the mean value of cards is slightly higher (56.9 euro). In the same way, we derive the market shares of cash and cards under the assumption of a reduced shadow economy. Table B3 shows that reducing by half the shadow economy in Europe would lead to doubling debit card transactions (both in volume and in value terms). Such a shift away from cash, even though still limited (88 percent of transactions would still be by cash), can lead to considerable savings in resource costs of payment instruments as we show next. We now estimate the resource costs of cash and cards based on the respective value and volume market shares obtained under the assumption of a reduction of the shadow economy. As before, we assume the same cost structure for cash and cards as in the Belgian and Dutch payment systems (see Box C in the previous chapter for more details). Thus we obtain the numbers illustrated in table B3. For this group of European countries, the total resource cost of cash and cards is estimated to be 1.1 percent of GDP on average. Decreasing the size of the shadow economy by half would produce savings equal to approximately 0.1 percent of GDP, which at the euro area level would mean saving about 7 billion euros.
89
Table B3: Impact of Reduced Shadow Economy in Europe46 Average results for panel of 13 European Countries
Actual market share (volume) Actual market share (value) Observed market share as predicted by model (volume) Observed share as predicted by model (value) Market share with halving SE as predicted by model (volume) Market share with halving SE as predicted by model (value)
Cash
Debit card
92 %
8%
63%
37%
95%
5%
72 %
28%
88 %
12 %
46 %
54 %
It is worth mentioning again that these numbers should be interpreted with care as they are obtained on the basis of a number of assumptions, in the absence of detailed information at supranational level. Nonetheless we think that the results of our simulations can give a useful indication of the real resource costs of payment instruments and the possible savings that could be made from a better usage of the payment system. Table B4: Resource costs of payment instruments in Europe Average results for panel of 13 European Countries Cash + Debit card Estimated current resource costs (total, 2003)
Gains in resource costs from reducing shadow economy
Million euros
7751
% GDP
1.08
Million euros
636
% GDP
0.09
46
The countries are Austria, Belgium, Finland, France, Germany, Ireland, Italy, the Netherlands, Portugal, Spain, Sweden, Switzerland and the UK.
90
3.8
Concluding remarks
In this chapter we analysed the nature and size of the shadow economy in various European countries and its connection with cash usage. We noted that the shadow economy is quite large even in developed countries. It is motivated by the high tax level and social security burden, but also by people’s feelings about civic ethics, or tax moral, as Schneider defines it. These findings are very important to set the basis for political action aimed at combating the shadow economy. Given that most activities in the shadow economy are cash-based, it follows that a reduction of the shadow economy is likely to shift an important share of transactions from cash to cards. We tested this hypothesis empirically and found strong evidence in favour of the cash-card substitution motivated by hidden activities. On the basis of this result, we estimated the possible impact of a reduction of the shadow economy on the resource costs of payment instruments. Clearly, the main motivation for a reduction in the size of the shadow economy is that it increases government revenues; however, other important benefits are likely to follow. A more efficient and less costly payment system is one of these. In this perspective, a careful and coordinated government action which would simultaneously fight the shadow economy and stimulate cards usage can be particularly effective in reducing the size of hidden activities and in producing widespread benefits. Such a policy would be all the more effective if it concentrates its attention on strategic sectors that are particularly sensitive to secure and efficient payments. The scope of this study is not to explore how to fight the shadow economy, but rather to understand what drives the shadow economy and to analyse some of the benefits connected to its reduction. Nevertheless, from our analysis some conclusions about how to reduce the shadow economy can be drawn. Our results clearly show that the shadow economy is intrinsically connected with cash usage. Put differently, hidden economic activities exist mainly because of the anonymity provided by cash. As we showed in our theoretical analysis (Box D), the shadow economy needs cash to operate. It follows that political action aimed at fighting the shadow economy must operate in the framework of an overall political package that modifies the tax collection systems and that provides incentives to use cards and other alternatives to cash for transactions.
91
When considering competing payment services with their connected costs and benefits, it becomes clear that the existing situation gives cash a structural advantage over other payment instruments because it facilitates transactions in the shadow economy. This is the reason why any policy aiming at reducing the shadow economy must take this factor into account in order to be effective. In the context of such a comprehensive policy package one could envisage giving incentives to the business sector to use alternative payment instruments in order to avoid the possibility of hidden activities. This could be done, for example, in the form of VAT reductions for those transactions using cards as a payment instrument. In a competitive environment such a tax incentive would ultimately be passed on to the consumer in the form of lower prices for card based transactions. At the same time it would make hidden activities relatively more costly and thus reduce their importance. Two further considerations are necessary in considering the benefits from combating the shadow economy. First, there might be important costs of switching from hidden to official activities. As a result, political action aimed at favouring such transition might have smaller effects than expected. That is why a coordinated political action of the type just mentioned would be particularly useful. A second factor to keep in mind is that some unregistered activities may not be viable when they are forced to move to the official sector. Therefore, when considering the benefits of moving economic activities from unregistered to official ones, one should take into account the fact that part of these activities are likely to disappear. This however is likely to attenuate to benefits in term of fiscal revenues, but not the benefits in terms of a more efficient payment system.
92
REFERENCES
APACS (1996), The costs of money transmission, March. Baumol W. (1952), The transactions demand for cash- and inventory theoretic approach, Quarterly Journal of Economics, vol. 66, 545-556. Baumol W. and D. Bradford (1970), Optimal Departures from Marginal Cost Pricing, The American Economic Review, vol. 60, 265-83. Bolt W. and A. Tieman (2005), Skewed Pricing in Two-Sided Markets: An IO approach, mimeo, 21 p. Brautigam R. (1979), Optimal Pricing with Intermodal Competition, The American Economic Review, vol. 69, 38-49. Brits H. and C. Winder (2005), Payments are no free lunch, De Nederlandsche Bank, 36 p. Dab D. (2005), Debit Cards and E-purses as Substitutes for Cash, mimeo, MFTS programme, 25 p. De Grauwe P., Buyst E. and L. Rinaldi (2000), The costs of cash and cards compared. The cases of Iceland and Belgium, University of Leuven. Dutch National Bank (2004), The costs of payments, Working Group on Costs of POS Payment Products. ECB Blue Book (2004), Payments and securities settlement system in the European Union. Addendum incorporating 2002 Figures, April. ECB Blue Book (2005), Payments and securities settlement system in the European Union. Addendum incorporating 2003 Figures, August. Edgar Dunn & Company (2003), Cost of Cash. European Payments Council (2003), Cash Working Group. Summary of Findings & Recommendations. Hyytinen A. and T. Takalo (2004), Multihoming in the Market for Payment Media: Evidence from Young Finnish Consumers, Bank of Finland Discussion Papers, nr. 25. Isachsen A. J. and Strom S. (1985), “The size and growth of the hidden economy in Norway”, Review of Income and Wealth 31, 21-38. National Bank of Belgium (2005), Coûts, avantages et inconvénients des différents moyens de paiement, December.
93
Retail Banking Research (2005), Study on the impact of Regulation 2560/2001 on bank charges for national Payments, prepared for the European Commission. Rysman M. (2004), An Empirical Analysis of Payment Card Usage, mimeo, Boston University. Rochet J.C. and J. Tirole (2003), Platform Competition in Two-Sided Markets, Journal of the European Economic Association, vol. 1(4), 990-1029. Santomero A. and J. Seater (1996), Alternative Monies and the Demand for Media of Exchange, Journal of Money, Credit and Banking, vol. 28, 942-960. Schneider, F. (2005), “Shadow economies around the world: what do we really know?” European Journal of Political Economy, vol. 21, pp. 598-642. Tobin, J. (1956), The interest elasticity of transaction demand for cash, Review of Economics and Statistics, vol. 38, 241-247. Van Bergeijk, P. and E. Kloosterhuis (2005), Modelling European Mergers: Theory, Competition Policy and Case Studies, Edward Elgar. Van Cayseele, P. (1994), De Belgische wet op de Mededinging: Concentraties in een Industrieel Economisch en Internationaal Juridisch Perspectief, Maklu, Antwerpen. Van Cayseele, P. (2005), What Merger Simulation is Not: The Merger between Hessenatie and Noord Natie in Retrospect, in P. Van Bergeijk and E. Kloosterhuis (eds.), Modelling European Mergers: Theory, Competition Policy and Case Studies, Edward Elgar.
94