RDP 2012-03: ATM Fees, Pricing and Consumer Behaviour: An Analysis of ATM Network Reform in Australia 3. Theoretical Models of ATM Fees and Empirical Findings

The literature on ATM networks explores a number of inter-related aspects of banks' profits, including decisions over pricing, the number and location of ATMs, the linking of networks, and strategies in closely related markets such as retail deposits and general banking services. Current theoretical research is focused on generating testable predictions regarding equilibrium outcomes in ATM markets. Empirical work has been more limited, in large measure due to data availability (McAndrews 2003). Empirical support for the most recent theoretical work has generally relied on qualitative results.

Two of the earliest papers to examine the profit and welfare implications of different ATM fee regimes were Salop (1990) and Gilbert (1991). Gilbert asserts that a regime that only allows jointly determined interchange fees will yield a more efficient outcome than a regime where banks are also independently allowed to set foreign and direct fees. He argues that because banks sell complementary products – ATM services and general deposit banking services – independent fee setting will lead to higher prices than in either the joint profit-maximising setting or the social welfare-maximising setting. This is because, when setting fees, each bank will not consider the decline in demand for the complementary products offered by other banks. In contrast, Salop reasons that as operating costs and consumers' utility from ATMs vary by location, regimes that only allow interchange fees are unlikely to yield an optimal allocation of ATM services. He also argues that if banks are allowed to charge both foreign and direct fees for ATM use, the price faced by consumers will be unaffected by whether an interchange fee is charged or not. Salop informally shows this interchange fee neutrality by demonstrating that any transfers between a cardholder, his bank and the ATM owner brought about by a given set of interchange, foreign and direct fees can be replicated by the combination of just a foreign fee and a direct fee.^[9] Salop does not, however, formally examine banks' equilibrium strategies in setting fees or the determinants of the equilibrium level of fees.

With the widespread introduction of direct fees across the United States in 1996, theoretical models began examining the use of direct fees as a strategic tool to attract deposit customers. Massoud and Bernhardt (2002) show that banks have an incentive to charge a high direct fee to non-deposit customers, as this induces more consumers to open deposit accounts with them, in order to avoid the fee.^[10]

Massoud and Bernhardt (2004) endogenise the ATM deployment decision and find that when direct fees are charged, banks have an incentive to over-provide ATMs. All else equal, this raises the likelihood that consumers will open an account with a bank deploying many ATMs in order to avoid high direct fees.

The introduction of direct fees in the United States also provided a ‘natural experiment’ to test model inferences empirically. Massoud, Saunders and Scholnick (2006) find evidence that in the presence of direct fees, deposit customers will switch from banks with smaller ATM networks to those with larger networks. In addition, they find evidence that banks with smaller networks can increase their market share by deploying more ATMs. In contrast, Prager (2001) finds no evidence that the introduction of direct fees affects the market share of small banks.

A significant drawback of Massoud and Bernhardt's (2002, 2004) approach is that they exclude interchange fees and foreign fees from their analysis. As shown in a number of more recent papers, the presence of these fees critically influences banks' strategic behaviour.^[11]

Croft and Spencer (2004) incorporate interchange fees as well as endogenously determined foreign fees. Different degrees of customer lock-in are also allowed (where lock-in captures the ability of customers to switch their deposit account to a different bank). Although Croft and Spencer do not solve for the explicit profit-maximising level of fees, they show that the total price faced by consumers for foreign ATM withdrawals is higher when direct fees are charged. In addition, they also prove the intuitive result that banks' foreign fees will be higher when customers are locked-in at the time (non-interchange) ATM fees are set, because deposit customers are not able to move to banks offering lower foreign fees. They also show that joint-profits are lower when banks use direct fees, providing an explanation for why US ATM network operators had previously tried to ban direct fees.

Donze and Dubec (2008) use a model where only interchange fees are charged to make predictions regarding the adjustment of the ATM market in the United Kingdom when an interchange fee-only regime was implemented in 2000.

They predict a decrease in ATM deployment by banks, but an increase in deployment by non-bank ATM providers, which retained the option to charge direct fees. They then show that this prediction is qualitatively consistent with the UK experience to date.

Donze and Dubec (2009) also analyse the effect the Australian ATM reforms may have had on the entry of non-bank ATM providers. They predict that the elimination of interchange fees will lead to more non-bank ATM providers entering the market and a fall in banks' share of ATMs. There is some tentative evidence of this occurring following the Australian reforms, with the share of ATMs owned by independent operators (that is, those not owned by banks, credit unions or building societies) rising to 51 per cent in 2010 from 48 per cent in 2008 (Edgar, Dunn & Company 2010).

A drawback of existing models is their failure to accurately reflect the costs faced by banks and ATM owners. The common assumption is that banks bear the same marginal cost of a transaction irrespective of who uses their ATM, and that banks bear no costs when their customers use a foreign ATM. In reality, banks face an own-bank processing cost each time their deposit account customers use an ATM, irrespective of whether the ATM is owned by the bank or is ‘foreign’. This processing cost is significant, and is estimated to be 14 per cent of the average total cost of an ATM transaction (to banks; Table 1). It may, therefore, affect the incentives of banks and modify equilibrium fee strategies and outcomes.

More generally, the usually innocuous practice of normalising costs to zero to simplify the analysis – which occurs to some extent in most recent models of ATM fees – might be undesirable in a market where multiple costs and fees are present. A normalisation of costs to zero in this setting precludes analysis of the relationship between each of the fees charged by banks and each of the costs incurred. It also prevents prediction of the equilibrium level of fees in absolute terms, and the comparison of the relative magnitude of each of the fees. This paper addresses these issues by explicitly including marginal ATM usage costs, including the own-bank processing cost, in the model.

Footnotes

This result is shown more formally in a number of later papers with varying market set-ups including Croft and Spencer (2004). This result is also demonstrated in this paper. [9]

This strategy has become known as the depositor-stealing motive for direct fees. It is employed by banks because fixed account-keeping fees lend themselves to rent extraction more readily than per transaction direct fees. [10]

See, for example, Croft and Spencer (2004), Donze and Dubec (2006, 2008, 2009). [11]