There are three basic schools of thought on the Israel / Palestine thing.- Those evil Israelis are out to destroy Palestine, and the Palestinians are just fighting back.
- Those evil Palestinians are out to destroy Israel, and the Israelis are just fighting back.
- It's a cycle of violence, where both sides are fighting back against the other.
Wouldn't it be handy if science could provide an answer? According to the authors of a new paper in Proceedings of the National Academy of Science, the "cycle" school is right: both sides are fighting back against the other: Both sides retaliate in the Israeli-Palestinian conflict.
The authors (from Switzerland, Israel and the USA) took data on daily fatalities on both sides, and also of daily launches of Palestinian "Qassam" rockets at Israel. The data run from 2001, the start of the current round of unpleasantness, to late 2008, the Gaza War.
They looked to see whether the number of events that happened on a certain day predicted the number of events caused by the other side on the following days, i.e. whether a Palestinian death caused the Palestinians to retaliate by firing more rockets and killing more Israelis, and vice versa.
What happened? They found that both sides were more likely to launch attacks on the days following a death on their own side. The exception to this rule was that Israel did not noticeably retaliate against Qassam launches. This is perhaps because Qassams are so ineffective: out of 3,645 recorded launches, they killed 15 people.
These graphs show the number of "extra" actions on the days following a event, averaged over the whole 8 years, according to a statistical method called the Impulse Response Function. Note that the absolute size of the effects is larger for the Israeli retaliations (the Y axis is bigger); there were a total of 4,874 Palestinian fatalities and 1,062 Israeli fatalitiesThe authors then used another method called Vector Autoregression to discover more about the relationship. In theory, this method controls for the past history of actions by a given side, so that it reveals the number of actions independently caused by the opposing side.
the number of Qassams fired increases by 6% on the first day after a single killing of a Palestinian by Israel; the probability of any Qassams being fired increases by 11%; and the probability of any Israelis being killed by Palestinians increases by 10%. Conversely, 1 day after the killing of a single Israeli by Palestinians, the number of Palestinians killed by Israel increases by 9%, and the probability of any Palestinians being killed increases by 20%What are we to make of this? This is a good paper as far as it goes, and it casts doubt on earlier analyses finding that Israel is retaliating against Palestinians but not vice versa. However, the inherent problem with all of this research (beyond the fact that it's all based on correlations and can only indirectly imply causation), is that it focuses on individual acts of violence. The authors say, citing surveys, that
....retaliation accounts for a larger fraction of Palestinian compared with Israeli aggression: in the levels specification, 10% of all Qassam rockets can be attributed to prior Israeli attacks on Palestinians, but only 4% of killings of Palestinians by Israel can be attributed to prior Palestinian attacks on Israel.... 6% of all days on which Palestinians attack Israel with rockets, and 5% of all days on which they attack by killing Israelis, can be attributed to retaliation; in contrast, this is true for only 2% of all days on which Israel kills Palestinians.
Over one half of Israelis and three quarters of Palestinians think the other side seeks to take over their land. When accounting for their own acts of aggression, Israelis often claim to be merely responding to Palestinian violence, and Palestinians often see themselves as simply reacting to Israeli violence.But I don't think many Israelis would argue that the IDF only kills individual Palestinians as a reflex reaction to a particular attack. They're claiming that the whole conflict is a defensive one, that the Palestinians are the aggressors, but that doesn't rule out their taking the initiative on a tactical level e.g. in destroying Palestinian military capabilities before they have a chance to attack. And vice versa on the other side.
WW2 was a war of aggression by the Axis powers, but that doesn't mean that the Allies only killed Axis soldiers after they'd attacked a certain place. The Allies were on the offensive for the second half of the war, and eventually invaded the Axis's own homelands, but it was still a defensive war, because the Axis were responsible for it.
For Israel and for Palestine, the other guys are to blame for the whole thing. Who's right, if anyone, is fundamentally a historical, political and ethical question, that can't be answered by looking at day-to-day variations in who's shooting when.
Comment Policy: Please only comment if you've got something to say about this paper, or related research. Comments that are just making the case for or against Israel will get deleted.
Haushofer J, Biletzki A, & Kanwisher N (2010). Both sides retaliate in the Israeli-Palestinian conflict. Proceedings of the National Academy of Sciences of the United States of America PMID: 20921415
37 comments:
> perhaps because doing so would require a study of the last 2,500 years of history, religion and politics.
The Israeli-Palestine conflict is about 100 years old and it is essentially a political matter, not a religious one.
But both sides make reference to older history and to religious concepts, so you need to understand those to be able to critically evaluate such claims.
They do, political science.
Also they head hunt the most promising scientists when they're still studying at uni and cart them off to military centres.. sometimes sending them around the world surveying bombs, collecting data, complete with police escorts, top secret stuff they're not allowed to discuss.
In that they might build regression models based on sensitive data, like from intelligence agencies but they don't publish it in journals. It gets sent to the relevant officials, policy makers, think tanks, private firms etc. CIA quarterly probably won't be a good idea.
Makes sense why they do that nowadays. It doesn't give time for the bright undergrad scientist to build an academic portfolio or a reputation. Rape their minds and discard them. The world may never even hear of these scientists. Monster of capitalism.
From my understanding the Israeli-Palestine conflict makes sense from a geopolitical perspective because that's the language policy makers use to set agendas there. It isn't based on philosophy, economics, science etc.
I mean you could interpret it from those perspectives but that would be like asking an artist to draw a picture of a war scene and making sense of it. It would be difficult for politicians to make real time decisions from that.
Both sides definitely fight back, but there's an obvious power disparity.
But, really, as in most things, oversimplifications make monsters out of us. Just as in the 90s (and still) we talked about how those tribes in Africa just couldn't get along with each other and were killing everyone and we reduced what was a complicated genocide of retaliation on top of retaliation on top of colonially-induced stratification to racist words of savagery. And the worst part is that war isn't over and it's still spilling over into all the countries that used to make up the Belgian Congo.
It's offensive and we should be better than that.
The authors are biased! Biased, I tell you! I don't know what their bias is, but I know they're biased!
On a slightly more serious note, one concern I have is the impulse response functions don't return to baseline. Setting an arbitrary cutoff of 22-or-so days keeps the math simple, but it could also mean that if responses last longer than that, it might non-trivially alter the initial response and their core results. (I've only skimmed the paper, but that's what jumped out at me)
Sorry, comment deleted, as per the comment policy. Other posts have no restrictions but I really don't want this to turn into a debate on the legitimacy of Israel, there are plenty of other places to talk about that.
God, I wish Nancy Kanwisher focused on real science, something she's actually very good at.
Well, he methods used in this study are not a million miles away from fMRI data analysis... maybe that's why she was on board?
Why hasn't Big Pharma created a pill to "cure" this "madness"? In this era of biological psychiatry hegemony, surely these behaviors are the result of BRAIN illnesses, en masse? Why think of the potential market for "anti-hate" or "anti-prejudice" drugs!!! Big Pharma could dispense pills to Israel and all of the Arab nations! Imagine the profits!!
Quote:
"....perhaps because doing so would require a study of the last 2,500 years of history, religion and politics." Unquote.
It is merely a seven year data that was analysed - a blink in the 2500 yrs history which would surely be more interesting!
anonymous: Well, we've heard about how lithium in the water supply prevents suicides. Maybe it would prevent suicide bombers too. There's only one way to find out...;)
The "halitosis bomb" and "gay bomb" are informal names for two theoretical non-lethal chemical weapons, which a United States Air Force research laboratory speculated about producing, which involved discharging female sex pheromones over enemy forces in order to make them sexually attracted to each other.
http://en.wikipedia.org/wiki/Gay_bomb
Calm down, veri, calm down, the devil is watching...
(and waiting!)
JC you scared me! What are you still doing here? This is insane, really. Look, I read Neo's blog because I used to play with plastic brains in the bath. It isn't something you'd understand. So quit fretting like a maniac.
Why isn't there a fourth school that believes both sides are evil and out to destroy each other? (My guess is that it's nearly indistinguishable from "both fight back".)
*Feel free to delete this if it's too Israel-Palistiniany!*
This perplexes me..
Prior studies using these methods have argued that Israel reacts strongly to attacks by Palestinians, whereas Palestinian violence is random (i.e., not predicted by prior Israeli attacks)...
That needn't have meant that attacks were "random". There is, after all, a continual blockade.
Israel launches pre-emptive strikes to assert the balance of power in the Middle East. Palestine engages in guerilla warfare i.e. random to weaken the world's economy. They're just strategies not religious or evil when it comes to the politics.
Both sides are retaliating, yes obviously. Therefore the retaliations are equal and equivalent? It doesn't follow, and obviously they're not.
The conflict is expressed in religious terms, therefore it's a religious issue? Is that really what you're saying?!
Kapitano: I never said it was a religious issue, I said you need to understand the religious issues to understand it.
For example, while it is not surprising that the Jews wanted a state of their own (especially after WW2), I don't think you can understand why they wanted a state there unless you've read the Old Testament.
The Hebrew context should be understood but doesn't mean Jews are religious because they're kind of not. Let me just say Herzl's vision of Zion or a homeland to unite the diaspora was not tied to a historical space. The Jews were initially offered a place in Africa not Palestine and they were happy to settle there.
That place in Africa was the present day Rift Vally in Kenya.
bsci said...
> Setting an arbitrary cutoff of 22-or-so days keeps the math simple, but it could also mean that if responses last longer than that, it might non-trivially alter the initial response and their core results.
Hi Bsci,
just to clarify, the lag orders we used for the autoregression weren't chosen arbitrarily, but based on the degree of autocorrelation in the timeseries as determined by the Bayesian information criterion (which is consistent for the true lag order). best regards, johannes
This is why I never comment on technical stuff. Authors know their baby.
@johannes
I'm not sure your answer fits with what the data is showing. If what you say is correct, then repeating the analysis with a lag of one more day would show the extra day's values being very close to zero (i.e. the impulse response doesn't have much predictive power one day later. One doesn't usually see such sharp drop offs using such a model.
Perhaps there was an issue with Baysean prior selection or the arbitrary cutoff threshold for significant autocorrelation wasn't best for this data. I don't know what is right or wrong, but seeing such shapes raises a red flag that something might not be properly modeled.
@veri, I might be right or I might be full of it, but science doesn't advance my assuming the author knows best. Raising questions and asking people to justify their assumption is how some of the more interesting discoveries are made. In fact, this paper seems to be a partial replication of an earlier paper which focused just on killings, while these authors noted the importance of tabulating non-lethal attacks.
Bsci.. just know they'll defend that baby no matter what. Not so much because of the academicness but.. you know why. "Both sides retaliate in the Israeli–Palestinian conflict." These sorts of analyses are done by game theorists.
For example something like this--
http://ideas.repec.org/p/zbw/csledp/200209.html
Anonymous: And then there was the notorious Jewish Autonomous Oblast aka "Stalin's Zion"...
Of course, any "objective" study of this kind of thing is going to have to make so many assumptions as to make it irrelevant to most people (given their own assumptions).
For instance, it sounds like the rocket attacks were counted as attacks but not house demolitions. If house demolitions were counted as attacks, (and they are international war crimes) then I imagine the final analysis would have looked very different.
Dear Neuroskeptic,
Thank you for an intriguing post. Your critical breakdown of the study highlights its strengths and its flaws and certainly got me thinking. The study seeks to understand the causal factors which perpetuate the Israeli-Palestinian conflict. They conclude that the inability of both sides to recognize the others violence as retaliatory is at least one causative factor. Although the statistical data seems to have merit, I feel the conclusions drawn from it assume too much and approach the issue from an outsider's perspective. To begin, the ideologies of the sides participating in the violence are incompatible. I'm not entirely sure that an end to violence on both sides would solve the territorial issue this conflict originates from. As outsiders with nothing at stake in the conflict we cannot possibly understand the importance of the past for those involved in the conflict today.
On a more concrete level I found it interesting that the data showed a lack of retaliation to Qassam attacks by Israel. Although Qassam attacks are not deadly they are politically salient in Israel because they cause a great deal of panic and uncertainty. Because the attacks are so frequent the IDF cannot address the problem directly, so most anti-qassam efforts have focused on eliminating the import of rockets using methods such as destroying tunnels used to smuggle equipment into Palestinian territory. Did the study take into account retaliation by the IDF that did not result in Palestinian deaths? It does seem unlikely that Israelis would argue that the IDF kills individual Palestinians as reflex reactions to particular attacks. And like you mention, there is a whole tactical level to this which the study does not address at all. It assumes complete reactivity on both sides when the reality of guerrilla warfare requires sides to take proactive steps towards limiting the enemy's ability to function. Would the results of the study change completely if this were taken into account? I feel like the study takes an almost romantic view of the nature of the conflict, hoping that once both sides realize the cycle of violence they are contributing to it will simply end. This completely ignores the ideological reality that fighters on both sides are willing to risk their lives for their cause, causes which are mutually exclusive, an Israeli state and a Palestinian state. Again, I would like to thank you for your sound analysis of the merits and limitations of this research.
bsci said...
@johannes
I'm not sure your answer fits with what the data is showing. If what you say is correct, then repeating the analysis with a lag of one more day would show the extra day's values being very close to zero (i.e. the impulse response doesn't have much predictive power one day later. One doesn't usually see such sharp drop offs using such a model.
Perhaps there was an issue with Baysean prior selection or the arbitrary cutoff threshold for significant autocorrelation wasn't best for this data. I don't know what is right or wrong, but seeing such shapes raises a red flag that something might not be properly modeled.
Hi Bsci,
the method you're suggesting (looking at longer lags) can be used for selecting lag orders, but is inferior to the bayesian information criterion for reasons discussed in most standard econometrics textbooks (e.g. check Hayashi). I think you might be confused by the fact that the figures represent impulse response functions rather than regression coefficients, which don't get rid of autocorrelation as the coefficients do. We did this to make the figure comparable to those in earlier papers.
cheers
johannes
..and I proved my point. Johannes, in solidarity. :)
Wow, Hunter.. Hunter..
@johannes,
I'm familiar with what an impulse response is. I honestly don't know if my critiques would affect your results, but it seems like the paper you are responding to, Jaeger & Paserman, AER 2008, calculated impulse responses with a 60day lag and the mean values of the impulse responses only crossed baseline around day 40. If the goal was comparison, why did you use such a radically different lag?
Jaeger & Paserman also used a Granger causality based metric rather than VAR. I understand why you wanted to use VAR, but it doesn't help comparison to not replicate the core methods of the study you are trying to rebut.
bsci said...
@johannes,
I'm familiar with what an impulse response is. I honestly don't know if my critiques would affect your results, but it seems like the paper you are responding to, Jaeger & Paserman, AER 2008, calculated impulse responses with a 60day lag and the mean values of the impulse responses only crossed baseline around day 40. If the goal was comparison, why did you use such a radically different lag?
Jaeger & Paserman also used a Granger causality based metric rather than VAR. I understand why you wanted to use VAR, but it doesn't help comparison to not replicate the core methods of the study you are trying to rebut.
Hi Bsci,
we used the lags we used because they have the greatest probability of being the correct ones for this timeseries, as determined by BIC.
Regarding Granger causality, the results of that analysis can be found in Table 3.
johannes
The article doesn't go anywhere. It does a test based on a short period of time, during the more current history of the conflict. It also kicks back to generalizations of a conflict that has been going on "for thousands of years," which it hasn't.
They are also incorrect in assuming that your position on the conflict has to do with your location. It has to do with how well you know the history of the area and conflict and how attached you are to the "God gave to to the Jews because they are his chosen people" religious Ideology. However, one might argue that people living in the US know less about world history, and so, due to their geographic location they will lean a certain way...
Its a simple conflict. The cause of it is pretty straight forward. Usually the corporate news tries to over complicate it so that it doesn't offend anyone by addressing streight forward key issues.
Here it is in a nutshell:
A. An indigenous population called the Palestinians were living on a strip of land for over one thousand years with a common language. The Palestinians integrated into the land naturally over thousands of years. they did not commit any ethnic cleansing or dispose of other people to live there. During the early 1920s Most of them were Muslims, some of them were christians. Around 5% of them were Jewish.
B. A foreign population, based on religious ideology decided that their god promised them this land around 2000 years ago and it was their birth right to take the land of Palestine and turn it into the Land of Israel. The Jews lived in the land of Palestine during Roman times and naturally left the area due to poor economic conditions and some measure of persecution from the Romans that everyone endured including the ancestors of the Palestinians. Later during Islamic rule, some Jews returned as a result of Islamic tolerance. But the majority of Jews chose to stay in Europe and other places in the Middle East. All in all, the Jewish population in Palestine became a small minority for more than a thousand years.
C. Modern day Israel was created by forcing almost a million indigenous Palestinians from their natural homelands and replacing them with European Jews (1940 - 2010). The Jewish claim to the land was based on religious ideology and an ancestry that hasn't lived in that area for many hundreds of years. The rest of the Palestinians remained under occupation, living as second class citizens in their own ancestral lands.
In conclusion, this is not a complicated conflict that has been going on for hundreds of years. Its a pretty straight forward conflict that has been going on for a little over fifty years. Over the last sixty years or so, one foreign population has forced another native population out of their biological land. Its called ethnic cleansing. That is the foundation stone of the conflict.
Doesn't take a genius or complicated graphs and charts to figure this one out.
Cheers
Anonymous, your 'simple account' is full of historical fabrications, going from blatant lies to minor misrepresentations. I won't bother to refute every little lie, but just to name a few: 1) Half of Israel's Jewish populations came from Arab countries, not from Europe. They were forcefully driven out of their homes in Arab countries where they lived for hundreds of years 2) The name 'Palestinians' never appeared as a name for a people or a nation before the 1960' 3) you ignore the numerous attempts by the Israelis (and the UN resolutions) to create 2 state solutions: one for Arabs and one for Jews. The Israelis always agreed for a compromise, the Arabs always rejected it. 4) Many places in Israel e.g. Jerusalem, always had a Jewish majority, for thousands of years, until the Arabs forcefully expelled the Jews in 1948. I can go on and on, you get an F in history.
For further discussion of the statistical methods used in this study and their possible flaws, see this article:
http://www.pnas.org/content/108/15/E53.full
Post a Comment