RDP 2009-03: Competition Between Payment Systems: Results 6. Conclusions

In this paper we have used our ECR model and several others to study competition between payment systems. Because of the complexity of such competition, all models necessarily involve some simplifications to make them tractable. For example, our ECR model does not incorporate ‘business stealing’ considerations, nor does it allow for separate issuers and acquirers (being a model of competing three-party rather than four-party payment networks).

Our ECR model does, however, overcome several key limitations of earlier models used in the literature. Most notably it allows for flat rather than per-transaction pricing by platforms to consumers, and endogenises card holding and acceptance decisions by both consumers and merchants. These represent steps towards a more realistic treatment of payments system competition in a range of practical settings of interest. A number of key messages and subsidiary findings emerge from our analysis, which we would expect to be robust to the use of alternative modelling approaches.

The main message is that platforms' use of flat rather than per-transaction pricing to consumers has a potentially major impact on the behaviour of platforms, merchants and consumers. Caution is therefore called for in drawing any firm conclusions about market behaviour from models which assume purely per-transaction pricing. For example, our results show that introducing flat pricing by platforms to consumers calls into question what had been a consistent finding in the literature: that an increase in the propensity of consumers to single-home will necessarily lead competing platforms to bias their prices in favour of consumers relative to merchants.

One reason we would expect this general message to be robust to plausible changes to our specific ECR model is that, in a world where both sides of the market may choose to multi-home, the use of flat pricing leads to different consumers facing different effective per-transaction prices for the use of a given platform's card. This is not the case when platforms' pricing to both sides is on a purely per-transaction basis, or when (as in the CR model) consumers are prohibited by fiat from multi-homing. It thus represents a fundamental change (and one that better matches reality in, say, the case of competing credit card schemes).

A second, subsidiary finding of our analysis of flat versus per-transaction pricing is that, even in a world where platforms use purely per-transaction pricing, care needs to be taken in describing the relationship between a side's propensity to single-home and the relative attractiveness of the pricing it will be offered by competing networks. Various papers, in reporting there to be a positive link, have stated it in quite general terms. For example, Rysman (2007) asserts that ‘the literature on two-sided markets establishes that, in a competitive market for payment networks, the side that multi-homes subsidizes the side that single-homes’ (p 10).

Our results, however, not only show the need to qualify such statements as applying only to situations where platforms use solely per-transaction pricing, but also demonstrate the need for a further caveat. This relates to which side holds the choice of instrument at the moment of sale. Specifically, our results show that a positive link only holds between the propensity to single-home on the side with the final choice of payment instrument, and the relative attractiveness of platforms' pricing to that side. Where the propensity to single-home increases on the side without the final choice of payment instrument, this can lead to less attractive pricing to that side, in proportional terms, even where platforms' pricing to both sides is on a purely per-transaction basis.

Finally, as noted previously by Hermalin and Katz (2006), possessing the final choice of payment instrument may not be an unalloyed benefit to consumers because it may shift competing platforms' price structures in favour of merchants and against consumers. Of course, other factors – including some not incorporated in our various models, such as ‘business stealing’ considerations – will likely work to push these price structures in the opposite direction. However, even if these other factors dominate, it is still potentially helpful to be aware of this possible influence on platforms' pricing behaviour, and to understand how it would change with variations in key elements of platforms' cost structures or the preferences of market participants.

In this regard, the results from our basic PTP model, and various generalisations of it, extend the findings of Hermalin and Katz. These results show that the force of the Hermalin and Katz phenomenon falls off not only as competing platforms' per-transaction costs rise, but also as their per-subscriber costs increase, and as consumers' innate predisposition to single-home rises (at least above a certain low threshold level).