NOW THERE'S A CHARGE FOR WHAT THEY USED TO GIVE FOR FREE. (Apologies to Tom Lehrer.) John Palmer asks, "What Is a Reasonable Fee for ATM Usage?" The context: populist politicians in the People's Republic of Canada would prefer to ban the fees. His lead, however, illustrates the real problem.

When ATMs were first introduced over 30 years ago, financial institutions encouraged their own customers to use them, hoping to reduce teller costs. And for the most part this change has been successful.

That, in fact, was the motivation of Wisconsin's Take Your Money Everywhere campaign, circa 1978, and the selling point was that an account holder at Affiliated Bank could withdraw money from a machine maintained by Anchor Savings & Loan or First Wisconsin (not the current incarnation), or, once the board was sold on the scheme, the University of Wisconsin's employee credit union.

Professor Palmer goes on to note,

At the time I liked it, but I figured there must be legislation requiring these kind people to treat us so generously. It just didn't seem to me that zero transaction charges for ATM use would emerge in a competitive, profit-maximizing equilibrium.

With bundling and nonconvexities, why not? The cost to the member bank of serving another bank's depositor is lower through the network than it would be to go through all the rigamarole of cashing a check for someone (not necessarily an out-of-towner) who is not a depositor. Remember those days? "Do you have an account with us?" (No.) "Can you produce five forms of identification and a lien on your first child?" (May I speak with a manager?)

The ill will (and regulatory reaction) the bankers have engendered is partially self-inflicted. It is easier to introduce a service at a premium price and lower the price along the learning curve, than it is to introduce a service at no extra charge and then introduce charges. (That the latter behavior looks a lot like the popular perception of predatory pricing doesn't help.)