RDP 2012-03: ATM Fees, Pricing and Consumer Behaviour: An Analysis of ATM Network Reform in Australia 5. Consumer Behaviour Post-reform

A key feature of the Australian ATM market post-reform has been consumers' sharp and sustained shift away from using foreign ATMs (Figures 5 and 6). Transactions at foreign ATMs accounted for almost half of all ATM withdrawals in the four years prior to reform, but fell to around 40 per cent post-reform despite there being no change in price.^[31],^[32] This behaviour cannot be accounted for by the model of ATM fees presented in this or any other existing paper on ATM fees. There are at least two possible explanations of this change in behaviour.

The first explanation comes from the literature on prompting and suggestive selling that shows consumers' decisions can be influenced by point-of-sale promotions (see DiClemente and Hantula (2003) for a review of relevant studies). In particular, Wansink, Kent and Hoch (1998) show that, consistent with a simple anchoring and adjustment model of consumer behaviour, point-of-sale promotions that convey no new information can change a consumer's purchase decision.

This is relevant for the Australian ATM market as reform saw the introduction of a new point-of-sale prompt for foreign ATM transactions. Since the reforms, the following message has been displayed prior to the completion of a foreign ATM transaction:

If you continue with this transaction you will be charged [amount of direct fee]. To continue press here. To cancel press here. (APCA 2009)

That is, consumers are now prompted to consider cancelling foreign ATM transactions. Given the results of Wansink et al (1998), this could contribute to a reduction in foreign ATM use.

A second, complementary, explanation for the decline in foreign ATM use when the price of foreign transactions was unchanged is incomplete information. This explanation rests on the observations that in Australia after the reforms: (i) the price of foreign ATM transactions shifted from being entirely comprised of the foreign fee to being entirely comprised of the direct fee; and (ii) foreign and direct fees differ in how and when they are disclosed to consumers. If charged, direct fees are displayed on the ATM screen prior to a transaction being completed, whereas foreign fees are disclosed to bank customers on regular bank statements or on request.

To the extent that consumers did not know the exact price of a foreign transaction when making the decision to travel to a foreign or own-bank ATM, the information available post-reform is news to the consumer.^[33] Moreover, once the consumer travels to a foreign ATM and learns the price, he can now choose to either complete the transaction if the cost is deemed acceptable, or cancel the transaction and travel to an own-bank ATM or look for a cheaper foreign ATM.

To explore this possibility, it is assumed that consumers know with certainty the price of own-bank ATM transactions but assign a probability distribution to the price of foreign transactions.^[34] It is shown that if consumers have incomplete information over fees pre-reform, the additional disclosure post-reform can lead to decreased foreign ATM use, even if the expected price of foreign ATM transactions remains unchanged.^[35]

Within this framework, the consumer's ATM problem from Section 4.2 can be adapted to have the price of a foreign ATM transaction defined as a random variable, Φ, instead of a fixed number.^[36]

Pre-reform, as no new information on price is gained from visiting a foreign ATM, consumers only make one choice: do they travel from their initial location to a foreign ATM or do they travel to an own-bank ATM? The consumer at distance d₁ from an own-bank ATM makes this decision by comparing the utility he would receive from using an own-bank ATM, u^o = x − d₁, and the utility he expects to receive from using the foreign ATM, E(u^F) = x − (l − d₁) − E(Φ). The proportion of customers using own-bank ATMs in this set-up is given by:

Post-reform, as new information on price is gained from visiting a foreign ATM, consumers that travel to a foreign ATM now face a second choice: do they stay and use the foreign machine or do they travel back to an own-bank ATM? For simplicity and consistency with the model set-up in Section 4.2, the possibility of looking for a cheaper foreign ATM nearby is ignored.^[37] Consumers stay and use the foreign ATM if they learn Φ ≤ l, but travel back to an own-bank ATM if Φ > l.^[38] The expected utility from initially travelling to a foreign ATM is therefore

The first term in Equation (20) is the expected utility of using the foreign ATM conditional on Φ being less than or equal to l, since the consumer only uses the foreign ATM if this is the case. The second term is the expected utility of returning to an own-bank ATM taking into account that by travelling back to the own-bank ATM the consumer avoids the cost of the foreign ATM but travels an extra distance l. Equating the utility derived from initially travelling to an own-bank ATM, u^o = x − d₁, and the utility derived from initially travelling to a foreign-bank ATM, u^F, and solving, the marginal consumer is located at

The proportion of consumers using own-bank ATMs now consists of those that travel to an own-bank ATM first, d₁^*/l, plus those that travel to a foreign ATM but decide not to use it and return to an own-bank ATM, (1 − d₁^*/l) Pr(Φ > l). The proportion of consumers using own-bank ATMs is therefore

There will be a rise in own-bank ATM use post-reform if the difference between ‘Equations’ (22) and (19) is positive. That is

This will hold under a variety of specifications for Φ. A simple example is when the random variable Φ takes values φ^* − 1 < l and φ^* + 1 > l, both with probability 1/2. In this case, the increase in customers using own-bank ATMs post-reform is given by:

which is positive if l > φ^* + 1/3.

Therefore, under incomplete information, foreign ATM use may fall relative to own-bank ATM use post-reform even if there is no change in the expected price.

Footnotes

As the price of an own-bank ATM transaction was also unchanged, this shift in ATM usage cannot be explained by a change in relative prices (Table 2). A shift in the composition of ATM ownership and/or rates of deployment are other possible explanations for decreased foreign ATM use, but given that these are generally slow moving factors and that the shift in consumer behaviour was abrupt, they are unlikely to explain this episode. [31]

There was also a modest decline in the foreign ATM share over the five months prior to the reform; this may have reflected public awareness campaigns ahead of the reforms. [32]

Anecdotal evidence suggest that some customers were unaware that a fee for foreign ATM use was applied at all (particularly in the case of non-cash ATM services such as balance inquiries). [33]

These assumptions are reasonable. It is widely known that own-bank ATM transactions are, in most cases, free in Australia. Conversely, direct fees can vary by ATM and, prior to the reforms, consumers may not have always known the exact level of foreign fees (foreign fee information is available on bank statements, but it may not be rational to keep track of this information if the costs of doing so exceed the expected benefits). Given the publicity surrounding the reform and widespread scrapping of foreign fees by banks, consumers are likely to have become aware of the fact that foreign fees were permanently set to zero after the 2009 reform. [34]

In the more general case, where direct fees are charged pre-reform and foreign fees are charged post-reform, the argument is still valid. So long as the direct fee constitutes a greater proportion of the total price of foreign transactions post-reform, then visits to a foreign ATM will reveal more information and so potentially lead to greater own-ATM use. [35]

Recall the price of a foreign ATM transaction is composed f_i and σ_j (Φ = f_i + σ_j). In this extension, pre-reform the foreign fee, f_i, is a random variable and the direct fee, σ_j, is a fixed number (set at zero, as it was widely known that banks did not charge direct fees at this time). Post-reform, f_i is a fixed variable (set at zero, as banks no longer charge foreign fees) and σ_j is a random variable. [36]

In the model set-up described in Section 4.2, if the consumer is already located at a foreign ATM he will not choose to look for a cheaper foreign ATM. This is because the consumer knows that an own-bank ATM (offering free transactions) is located at least as close as the nearest alternative foreign ATM. [37]

In our two-bank model, the disutility of travelling from a foreign to an own-bank ATM is l, so consumers will only do this if the price of the foreign transaction is greater than l. [38]