Models vs knowledge

The Age reports on Australian legislation that forced banks to make ATM transaction fees explicit to the customer:

In place of the indirect fees were direct fees in which the owner of each foreign ATM took the money directly from our accounts each time we made a foreign withdrawal. But the size of the charge, typically two dollars, didn’t change. All of the economic models – including the Reserve Bank’s own model – suggested we would use ATMs pretty much as we had before. The incentives were much as they had been.

Instead withdrawals from foreign machines dived from around half of all ATM withdrawals to just 40 per cent. …A Reserve Bank study released yesterday says it’s behaviour that “cannot be accounted for by the model of ATM fees presented in this or any other existing paper”. To work out why, it has turned to research on retailing and a finding that point-of-sale displays can change purchasing decisions even when they convey no new information… The RBA’s tentative conclusion is that it is not the fee that is frightening us, it is being continually told about it.

  1. Framing effects such as loss aversion are hardly new so I’d be staggered if the RBA didn’t know about them.
  2. Just because your model doesn’t include an effect that you know to exist, that doesn’t mean it disappears or has no effect. It also doesn’t mean that you don’t know about it. I think we all know that being prompted to pay money affects behaviour so it would be surprising if the legislation was expected to have no effect. Of course, since it isn’t normally a relevant effect for the RBA they may well not have included it in their models previously. That doesn’t mean they’re idiots or didn’t know about framing.
  • http://tvhe.co.nz/ Matt Nolan

    Good post.

    It’s interesting – I think if I was told what the fee was I would probably be more willing to use it, as the implicit fee I expect is really high. For some reason there is no pressure on me to update this fee (the information cost is relatively high compared to expected benefit), so I just sit there.

    Of course it could be the other way, given I’m trying to guess how my expectations differ from actuals!