Yesterday, Haaretz ran as its top headline a story by Gideon Levy that "Most Israeli Jews would Support Apartheid Regime in Israel." The actual poll, commissioned by Dialog, was released in Hebrew today by Avi Mayer of the Jewish Agency.
One of the few useful skills Ph.D students acquire during our graduate studies is the ability to analyze research designs. With regards to research on controversial subjects, a good design is all the more important. Previous research by Tel Aviv University, ACRI, the Center for Racism, and Dahaf have shown that racism is in fact an issue in contemporary Israel society. Especially given that Israel was founded as a democracy, that any Israelis would support banning Arabs from voting for Knesset is concerning and worth researching. The same holds for similar attitudes in Europe or the United States.
Thus, the academics who set out to conduct a poll of Israeli public opinion are right in their desire to better assess such attitudes in Israeli society. What their poll also gets right is that it interviews a random sample of 503 Israelis. 503 is a slightly small sample size but carries a margin of error of plus or minus 4.4%. Statisticians like about a 3% margin of error, but getting enough respondents (about 1,000) is expensive, especially for a small firm like Dialog.
However, the claim that this poll shows Israelis support apartheid or accept that Israel is currently an apartheid is unsubstantiated by this poll. Yesterday, the Daily Beast's Noam Shelef was able to show, without the full poll, that the poll was not saying what Levy claimed it was. Now that the poll itself is released, here are three more problems:
1) The poll has unclear coding.
The poll reports respondents in five groups: Russian, Haredi, Religious, Traditional (Masorti), and Secular. There are three problems with these groupings:
a) It's unclear whether each respondent was put into each category according to a set of objective criteria, or self-selected.
b) It's unclear whether the size of each sample is representative of the general population. A group which is 10% of the population might have been 20% of the sample size.
c) There's overlap between categories. Russian Jews could easily be coded as secular, for example.
2) The poll results show large in-group variation.
An engineer, physicist, and statistician go hunting for deer. The engineer shoots too far to the left. The physicist shoots too far to the right. The statistician says "We got him!"
The problem with the reported data is that it over-generalizes the results of the sample. In reality, the poll data demonstrate large variation across each group. For example:
"#6: In your opinion, is it desirable to pass a law that would ban Arabs from voting for Knesset?"
Russians Haredi Religious Traditional Secular Total
Desirable: 7% 70% 52% 43% 18% 33%
To generalize about "Israeli society" given such huge variation across sub-groups is tricky. The 33% "total" is probably a mean, but a median would also be informative in this regard. It would likely reveal a skewed (i.e. hard to generalize) distribution given the strong preferences of Haredim. A more accurate claim than "Israeli Jews are racist" would be "Religious Jewish Israelis tend to support restricting Arab rights in Israel."
2) The poll does not control for ethnicity.
The poll does not break down the response of the Mizrahi population of Israel, who are Arab Jews. Based on previous research, this would likely bolster the findings of the poll as Mizrahim generally tend to be socially conservative. However, If the intent of the poll is to demonstrate a convincing correlation between religion and racism against Arabs in Israel, it needs to show that controlling for ethnicity does not harm the significance of the results. Translation: How do we know its Israeli Jews, and not Israeli Ashkenazim (or any other sub-group) that support discrimination? If we don't, the poll isn't saying what people are claiming it says.
As a side note, the poll doesn't ask the responses of Israeli Muslim Arabs. This isn't a problem with the poll itself, but it is a point which isn't explicit in the original Haaretz article and most other reporting on the poll. Arab Israelis represent about 20% of the state's population. That means the polls results cannot be generalized to the Israeli public at large, but only to Israel's Jewish population. While Israel's population is itself a majority, polling Jews to draw inferences about Israelis is over-generalization.
3) The poll is based on perceptions, not legal definitions of apartheid.
The poll supports the claim that Israeli Jews, as a generalized population, support specific policies, which when combined with other policies, hypothetically in the future, could reasonably constitute apartheid. Most of the assessments in the poll which respondents are asked to make about apartheid are about the respondent's own perceptions, not legal definitions of apartheid. Besides the fact that the questions could be understood in a biased way (a "transfer" to PA control versus a physical transfer of Arab citizens of Israel), apartheid is not whatever a majority of Israelis think it is. The poll also does not ask whether Israelis think Israel is currently an apartheid as a specific question.
With regards to respondents' support for discriminatory policies, this is a problem of necessity but not sufficiency. In other words, support for institutionalized discrimination is a part of apartheid, but is not the all-inclusive definition. All elephants have four legs, but that doesn't mean all four-legged animals are elephants. In a manner of speaking, the poll does not demonstrate the existence of a trunk and big floppy ears.
In conclusion, the poll raises questions about Israelis attitudes towards Arabs which merit further research and the kind of self-reflection in which all democracies should engage. However, it has not conclusively demonstrated the Israel is an apartheid, nor that Israelis support apartheid. Given the importance of Israel's democratic values, Arab rights, and international public opinion, research on questions of racism and apartheid must be done often and done well. Good research is the key to legitimizing the push for equality and the strengthening of public debate in Israel (and for that matter, here in Washington).
Analysts should not ignore the findings of this poll, some of which raise serious questions about the level of equality in Israeli society. However, they also should not accept an interpretation of the poll's results which are demonstrably unsubstantiated. If Israel is indeed an apartheid, this poll does not show it.