I was planning on blogging about another paper from the “NYU-CESS Experiments Conference”:http://cess.nyu.edu/conferences/3-2011/, but then I had the following fascinating real-life encounter with randomization.
I had just completed going through customs at a Mexican airport – which included having all of our bags go through a security scanner, much like the one we had gone though on the way out of Newark – when we were told to push a button. The button then lit up either a red or a green light. In our case, the green light lit up, but if the red light had gone on, our bags would have been searched.
Now I don’t know exactly what was going on when the button was pushed, but my best guess is that it was some kind of random number generator determining whether or not the light went on. Now maybe the process wasn’t totally random, but at the very least it clearly took the search/not search decision out of the hands of the individual customs agent.
The more I thought about this, the more impressed I was by the procedure. First, I considered this from the stand point of trying to prevent smuggling. There are some obvious parallels here to the whole debate about whether “TSA agents should or should not engage in profiling”:http://www.cbsnews.com/8301-503544_162-20023731-503544.html, but it struck me that at the very least, a truly random system ought to at least give serious pause to smugglers, as every border crossing would present a chance of being searched. Assuming that the standard alternative is to search “suspicious” people, then a random system takes away the tactic of trying not to look suspicious as a way of getting through customs. Now the downside of relying on only random searches is that you would lose the value of the agent’s intuition in spotting suspicious characters, so probably a system that combined random searches but also gave agents the discretion to conduct additional searches would be optimal in this regard, and for all I know that’s how this particular border actually worked.
But even more importantly, the red/green light system seems an incredibly powerful tool in the fight against _corruption_, a notorious problem at borders. For what the system really does is that it makes it impossible for an agent to ever _guarantee that someone will not be searched_. And if an agent can’t guarantee this, then there’s no reason for a smuggler to try to bribe an agent. And to the extent that there is less of an incentive for the briber to offer a bribe, that should make the whole system less corrupt.
Thus to the extent that we consider this type of corruption – smugglers buying off customs agents – to be a principal-agent problem (ie., the higher ups want the lower downs not to be corrupt, but the lower downs have incentive to be corrupt), the red/green light solution strikes me as a very valuable tool for reducing the prevalence of corruption in customs agents.
This leads to three questions that I want to throw out to readers of The Monkey Cage. First, has anyone ever seen this sort of red/green light decision mechanism at customs checks other than in Mexico? More generally, does anyone have evidence of a state ever implementing a randomization process like this not for the sake of study, but to take a potentially corrupt decision out of the hands of an agent in a similar principal-agent set up? Finally, is anyone aware of any research on whether these kids of measures have succeeded in reducing corruption?