Nudging at the ATM: a field experiment
Study on the impact of removing high predefined withdrawal amounts on withdrawal behaviour

In 2012, CentERdata conducted the study “Consumer incentives for efficient payments” on behalf of the Foundation for Efficient Payments (SBEB). This study investigated how retailers and banks can reduce cash withdrawals and promote debit card payments, to reduce the social costs of payment transactions. One of the incentives studied was the lowering or removal of the withdrawal amounts that are predefined on the ATM display (see Figure 1).

This adjustment appeared effective in reducing cash withdrawal amounts in lab situations in which cash withdrawals were mimicked as well as in a small-scale pilot that was run in collaboration with Rabobank. The current study is a large-scale field experiment in which 100 ATMs of four Dutch major banks have been adjusted for three months. 

Figure 1. Original display (left) and test display (right)

Predefined amounts and en withdrawal behaviour
The underlying idea is inspired on research on scale range effects in the survey methodology literature, which shows how responses to questions in surveys can be influenced by the predefined response options. More specifically, the response options may be used as a reference point, as a starting point for further reasoning.

Based on this idea, Thaler and Sunstein (2008) predicted, for example, that people will donate more to charity when the options “€100, €250, €1000, €5000 (or other amount)” are pre-printed, compared to when the options “€50, €75, €100, €150 (or other amount)” are pre-printed. In the current study, we applied the same principle in the context of cash withdrawals, by removing the highest amounts on the screen (€150 en €250). Consumers still had the option to choose a different amount after the adjustment (see Figure 1). Thus, the adjustment did not limit consumers’ freedom of choice. 

Analysis and results
Via a “difference-in-difference” analysis, data on withdrawals and withdrawal amounts of the adjusted ATMs were compared to transaction data of the same ATMs three months prior to the adjustment and with transaction data in the same period of comparable machines (e.g. in the same region) that were not adapted.

The analysis revealed that the removal of the highest predefined amounts reduced the average withdrawal amount. There were no indications that consumers started withdrawing cash more frequently (at the same ATMs) as a result of the adjustment. The adjustment thus seemed to have the intended effect. 

In the first half of 2018, all ATMs of the Dutch banks were adjusted. The adjustment is an important first step towards uniform ATMs that are easy to operate for everyone.

Related projects