Friday, February 18, 2005
Contagion or inhibition?
A timely article (for me, anyway) notes that the Israeli army has decided to stop demolishing the homes of Palestinian suicide bombers.
I was taken aback by the evidence-based approach of the Israeli army on this one. Too often in dealing with violence, governments and democratic publics decide that the only response that will reduce violence is forceful retaliation. While in the limit as retaliation goes to genocide, this is true, for any level of force that leaves some of your enemy alive, retaliation may be counterproductive, for the same reason it was appealing in the first place: when hit, people just want to hit back.
So props to the Israelis for looking at the evidence and trying to fashion a policy that will actually work---thus protecting their own citizens better.
The timeliness comes from the topic of my lecture today, the statistical analysis of event counts. Suppose we count the number of times an "event" happens in a period of time. The event might be the number of phone calls you receive in an hour, or the number of guerilla attacks made in a day, or the number of presidential vetos overriden in a year. A key issue is whether the occurence of an event has any effect on the likelihood of future events in the period. If not, we can say the events are independent.
But suppose we are counting phone calls by the hour, and usually we see 2 in the first 30 minutes, and 2 in the second thirty minutes. Then, something weird happens---in the first thirty minutes of one hour, we get 50 phone calls. Independence says we should drop back to the usual 2 or so in the next thirty minutes. Common sense says that we'll continue to get lots of calls---the calls are "contagious", and probably related.
Inhibition is the opposite of contagion. A classic example is challenges to presidential vetos. They are rare, and once challenged, the chastised president is more cautious with his veto. Having more challenges in the first half of a year should make them (and vetos) less likely in the latter half.
So the Israelis are looking at the past history of terrorist attacks and retaliation, and asking themselves the following question:
When we retaliate, does that lower the need for retaliation in the short-run? Or does it increase it? Does retaliation inhibit suicide bombing, or does it spread it like a disease, so that we must retaliate ever more in a spiral of violence?
This is a question that data can answer, and apparently, the Israelis have looked at the data and decided that while punishment might satisfy the needs of vengeance, may actually be harming the interests of security.
(An aside, reaching from statistics to economics: One might think from the above that a great way to test the effectiveness of deterrence logics is to look for inhibition following actual punishment. I suspect that in most case when actual punishment occurs frequently enough to be analyzed, you will find contagion, overdispersion, and failure of deterrence. Deterrence requires the capacity for rational and coordinated action---in which case, threats need only be credible, not actually carried out. The most effective punishments are those that are never used. If you have to ask whether a punishment is working, the answer is likely "no".)
Israel's defense minister ordered a halt Thursday to the controversial policy of demolishing the homes of Palestinian suicide bombers and gunmen after an internal army review concluded it has not deterred attackers but has inflamed hatred.
Since the 1967 Middle East War, Israel has razed more than 2,400 Palestinian homes -- leaving thousands of people homeless -- including 675 houses destroyed in the past four years of fighting, according to the Israeli human rights group B'tselem.
I was taken aback by the evidence-based approach of the Israeli army on this one. Too often in dealing with violence, governments and democratic publics decide that the only response that will reduce violence is forceful retaliation. While in the limit as retaliation goes to genocide, this is true, for any level of force that leaves some of your enemy alive, retaliation may be counterproductive, for the same reason it was appealing in the first place: when hit, people just want to hit back.
So props to the Israelis for looking at the evidence and trying to fashion a policy that will actually work---thus protecting their own citizens better.
The timeliness comes from the topic of my lecture today, the statistical analysis of event counts. Suppose we count the number of times an "event" happens in a period of time. The event might be the number of phone calls you receive in an hour, or the number of guerilla attacks made in a day, or the number of presidential vetos overriden in a year. A key issue is whether the occurence of an event has any effect on the likelihood of future events in the period. If not, we can say the events are independent.
But suppose we are counting phone calls by the hour, and usually we see 2 in the first 30 minutes, and 2 in the second thirty minutes. Then, something weird happens---in the first thirty minutes of one hour, we get 50 phone calls. Independence says we should drop back to the usual 2 or so in the next thirty minutes. Common sense says that we'll continue to get lots of calls---the calls are "contagious", and probably related.
Inhibition is the opposite of contagion. A classic example is challenges to presidential vetos. They are rare, and once challenged, the chastised president is more cautious with his veto. Having more challenges in the first half of a year should make them (and vetos) less likely in the latter half.
So the Israelis are looking at the past history of terrorist attacks and retaliation, and asking themselves the following question:
When we retaliate, does that lower the need for retaliation in the short-run? Or does it increase it? Does retaliation inhibit suicide bombing, or does it spread it like a disease, so that we must retaliate ever more in a spiral of violence?
This is a question that data can answer, and apparently, the Israelis have looked at the data and decided that while punishment might satisfy the needs of vengeance, may actually be harming the interests of security.
(An aside, reaching from statistics to economics: One might think from the above that a great way to test the effectiveness of deterrence logics is to look for inhibition following actual punishment. I suspect that in most case when actual punishment occurs frequently enough to be analyzed, you will find contagion, overdispersion, and failure of deterrence. Deterrence requires the capacity for rational and coordinated action---in which case, threats need only be credible, not actually carried out. The most effective punishments are those that are never used. If you have to ask whether a punishment is working, the answer is likely "no".)