The Diamond-Dybvig framework assumes that the bank cannot distinguish between short-term agents that withdraw for effective consumption needs and long-term agents withdrawing because they self-fulfillingly anticipate a run. Mixed with the sequential service constraint, even if the bank invokes a suspension clause there is a risk that short term agents would be at the end of the queue, and seemingly starve to death. How does this assumption hold up in 21st century banking where we have algorithms to instantly detect unusual withdrawals, to protect depositors from fraud? Is it science-fiction to think those algorithms could be calibrated to trigger a variant of the suspension clause on panicky depositors exclusively?

I’m asking because in his 1993 paper George Selgin criticizes the bank suspension as portrayed in Diamond-Dybvig. While so-called “bank holidays” fit the Diamond-Dybvig suspension, some better conceived bank suspension policies were more partial in the sense that depositors could still use their checkbooks and banknotes to consume. They didn’t starve. But bank suspensions might also be more partial in other ways; convertibility might be suspended only for depositors that seem to be in panic. In fact, even thought it was aimed at predatory redemption “duels” rather than Diamond-Dybvig “panic” runs, the option clause of the Scottish free-banking experience was not always used as a blanket measure, applying systematically to all banknotes. In some instances of duels, bona fide customers could still convert their notes while the clause was invoked against other banks’ agents. Granted, it might be easier to tell a regular customer from a competing bank’s employee, than it is to tell apart a customer withdrawing for real needs from a customer that’s panicking, if only because competitors would present a much bigger volume of notes for redemption than your regular customers ever would. But with  nowadays’ technology…?