Discover more from The Causal Fallacy
America is Only Going to Get More Divided on Israel
On Israel Polling and Age Polarization
Many Americans have been surprised by the ferocity of the campus protests against Israel’s invasion of the Gaza Strip. Students calling for the destruction of the state of Israel may seem like an extreme position, out of step with what most Americans want. To a point, that is an accurate perception. In recent polling, the plurality of American adults view Israel as an ally, while 84 percent cumulatively view it as either an ally or a “partner” (as opposed to a “rival” or “adversary”). A similar plurality say that their sympathies are with the Israelis in the conflict; in the same poll, Israel sympathists outweigh those whose sympathies are with the Palestinians by nearly 4 to 1.
But this general tilt towards Israel belies some important cleavages by age. More specifically, older Americans are almost uniformly supportive of Israel. Younger ones, especially those under 30, look different. They are not unanimously pro-Hamas, as some protests might seem to imply. But they are deeply divided, and ambivalent, over the conflict in a way that older generations are not.
Here, for example, are crosstabs from an Economist/YouGov poll asking respondents whether their sympathies are more with the Israelis or the Palestinians. Across the whole sample, The Israelis are +28 over the Palestinians. But in the 18-29 subsample, the Israelis are just +5. And more 18-29 year-olds say that they are “about equal” than either; in every other age group, the Israelis are the most popular choice.
There’s at least some evidence that this age effect is independent of composition. The crosstabs above break out support for Israel vs. the Palestinians by age and a variety of other variables. Across all the sub-groups that Echelon breaks out, support for the Israelis among under 50s is roughly half what it is among over 50s. (The two age bins, presumably a function of limited sub-sample size, elide what looks like a steady trend across age groups in the Economist crosstab.) Less than half of all groups actively support the Palestinians, but rates of support are much higher in the under-50 subsample than in the over-50s.
It seems like the decline in support for Israel, at least relative to Palestine, is a recent phenomenon. Millennials (who are currently between 23 and 43) are by far the least relatively sympathetic to the Israelis (i.e. net of their sympathy to the Palestinians). Notably, the Millennial curve used to track the Gen X curve, but it’s been bending downward since the mid-2010s.1
So at first pass, we can say three things with limited confidence:
young people, especially under-30s, are much more divided on the Israel-Palestine conflict than older people.
That age effect seems to exist independent of, but is attenuated by, compositional differences between the age groups. That is: young people are more educated, less white, more liberal, and less Christian than older people. But each of those differences individually does not explain the difference among age groups.
Age polarization on Israel is a relatively recent phenomenon. (Age and cohort are distinct, albeit related, so inferring this from the Gallup data involves a logical leap. But I suspect it’s true.)
Unfortunately, none of these data sets is public. But another data source is! The Pew Research Center has asked two waves of its American Trends Panel survey about whether they support or oppose separately the Israeli goverment, the Palestinian government, the Israeli people, and the Palestinian people. (Their write-ups of the data are here and here.)
These are large data sets. The 2019 wave has 10,517 respondents, roughly half of which answered questions about the peoples and roughly half about the governments. And in the 2022 wave, most of the 10,398 respondents answered all four questions. These data won’t let us get into the trends, but they will tell us something about covariates.2
First, here’s a depiction of how each age group measured by Pew responded to each question. In general, the 2022 wave was slightly more inclined to say they were favorable across all four measures, so I’m not sure any inter-year trend is meaningful at the individual response level.
Zooming in on the responses, we see a couple of interesting things. One is that there’s clear age polarization in views of the Palestinian and Israeli governments. Over-30s are progressively more likely to be favorable towards the Israeli government, and all about the same level of unfavorable towards the Palestinian government. Under-30s, by contrast, are equally unfavorable towards the Palestinian and Israeli governments. They don’t like Hamas or Fatah or the PA, but they also don’t like the government of Israel generically.
We see the same story in the responses to questions about the peoples, although the difference in intensity is far less pronounced. Respondents under 30 are equally favorable towards the Israelis and Palestinians, while those over 30 are (increasing in age) less favorable towards the Palestinians and more favorable towards the Israelis.
It’s worth noting that this age variation could be explained by a couple of different patterns. One is that all else equal, young people are more supportive of the Palestinians and less supportive of the Israelis. But another is that they are only less supportive of the Israelis, such that support is about equal, while older people are more supportive of the Israelis. In other words, it could be that young people are getting more sympathetic towards Palestinians, Hamas, etc. But it could also be that older people were unusually sympathetic towards the Israelis, and younger people are reverting to the mean.
What we can do is group respondents as falling into one of four categories: favorable towards the Israelis but not the Palestinians, favorable towards the Palestinians but not the Israelis, favorable towards both, and favorable towards neither. We can then look at patterns in that favorability to see which way young people are going.
This chart shows fraction of each age group going into each of our categories for both the government and people question. (I use the 2022 data for simplicity.) What you see in both cases is a stark decline in the fraction who are “lean Israel” (i.e. support Israel and oppose the Palestinians). The share who are equally favorable is roughly flat across age groups. In the government responses, most of the young respondents are equally disfavorable, with a few outright leaning towards the Palestinians. The same is true, albeit less pronounced, in the people questions. In short: it seems like young people are leaking out of the lean Israel category, and moving mostly to equal disavor, with a smaller share becoming lean Palestinian.
I want to try to understand some of the underlying determinants here. Why are younger people less likely to lean towards the Israeli government? I construct a variable that equals 1 if the respondent supports the Israeli government/people but not the Palestinian government/people, and a 0 otherwise. I then try to see what explains age disparities on this variable, using regression.
This table shows three successive models applied to the 2019 (the first three columns) and 2022 data (the second three columns) separately.3 These are linear probability models, such that each coefficient should be interpreted as a percentage point change in the probability of supporting only the Israeli government.
The first columns are just the effect of age groups compared to 18-29 (the reference category). As you can see, being in an older age group significantly increases the probability of leaning towards Israel. The second columns add in race and ethnicity controls, with “white” and “non-Hispanic” as the reference groups. These attenuate the effects of age more in the 2022 sample than in the 2019 sample, but age remains large and significant. (You also see that black respondents are much more likely to not support only Israel.) The third column adds in controls for ideology and partisanship. These reduce the coefficients on age by somewhat more than the demographic controls, but the age effects are still large and significant. You can also add in education and church attendance controls, and both have some effect on the DV, but do not really reduce the age gap appreciably.
Here’s the same regression, but using exclusive support for the Israeli people as the DV. The story is essentially the same: the age effect shrinks but does not disappear with demographic and political controls (and education and attendance controls, which I ran but did not include for readability). This is a confirmation of the Echelon finding: controlling for a bunch of factors, the age effect persists.
One other variable we can sort of mess around with is level of internet usage. In the 2019 wave, each half of the respondents were asked different versions of how much time they spent online: how much time they spend online generically, and how much time they spend online *not counting their email usage*. All respondents were also asked if they’d used five websites in the past month: Facebook, Instagram, YouTube, and two fake websites, FizzyPress and Doromojo. (The latter two are presumably to catch people just pressing agree.)
This table shows results from regressing all these measures of internet usage on exclusive support for the Israeli government and people. (It also includes our demographic and political view covariates, though they’re not shown). The first two columns use the two different time online questions, while the third collapses them into a single variable with a control for which question respondents got. The fourth column instead uses controls for use of each of the five websites. As you can see, there’s really no meaningful reduction, except in the second model for Israeli people among 30-to-49-year-olds. That’s a specific enough finding that I’m willing to attribute it to chance. (Although it’s kind of interesting!)
In short: whatever is causing age polarization on Israel, it is neither differences in political or demographic composition, nor amount of time spent online. Some other variable, beyond the reach of the survey data I’ve found thus far, may explain the difference. Or there may be a genuine age or cohort effect.
If the former, there are a number of possible culprits. Most obviously, none of these surveys tries to measure antisemitism, which is more common among younger people. There are a few good antisemitism batteries, including in the paper linked in the last sentence and here; it would be useful to see how they relate to views of Israel/Palestine. I suspect this is a big omitted variable, and would love to see it measured.
Another explanation might be some variation on issue polarization in the U.S. electorate. Here I am thinking primarily about former President Donald Trump, who a) is among the most polarizing figures in American history and b) put a lot of emphasis on his support for Israel. Maybe this association changed young Americans’ perceptions of Israel purely in reaction, an extension of TDS? There are a few questions about Trump in the 2019 wave of the ATP data, and they don’t seem to have an effect in regression. But there might be more to test there, and maybe Trump—or some other driver of polarization and issue salience—is an important omitted variable.
A third explanation is a secular decline in supportiveness by age. Maybe young people are just more ambivalent in general! To test that, you’d need control questions to measure views of, e.g., other countries.
Or maybe there really is a difference in how different cohorts view Israel. People over 65 were born in 1958 and earlier. They remember the Six Day War and the Yom Kippur War, when Israel was the underdog. Younger Americans, by contrast, have formed their opinions in the context of the BDS movement, a concerted domestic effort to raise the political salience of the Palestinians. Perhaps feelings about Israel are driven by age-correlated impressions of Israel per se.
If that’s the case, then the political implications of rising age polarization on the issue are particularly notable. Let’s assume this is a cohort effect, not an age effect—that respondents’ views about Israel are a function of when they were born, not how old they are. (We’d need repeated samples to differentiate.) If this is true, then as younger generations replace older ones, America’s unique support for Israel will necessarily become less of a consensus, and more of a polarizing issue. This could drive partisan realignment, as Jews shift towards a Republican party comparatively more sympathetic to Israel. But it could also affect U.S. policy commitments in the Middle East, both in terms of funding Israel and of our relations with other regional actors like Saudi Arabia and Iran. American voters and policymakers who do not regard Israel as any better than the Palestinians will be less inclined to provide military and financial support, and more inclined to side with the nations that criticize and attack Israel.
In other words, age polarization in America may have profound political effects in the Jewish state. To those who wish to preserve the American consensus in support of Israel, understanding why is vital to shifting the direction of the trend. In the absence of such knowledge (and the action that can follow it), then the campus protests are simply a vision of the future.
I don’t really talk about Zoomers in this piece because they are mostly still too young (23 and under means only a five-year age band are adults) to get picked up in large numbers surveys that don’t actively look for them.
I considered trying to combine them, but decided that creating weights for the combined sample was complicated enough, and I was likely enough to get it wrong, that it wasn’t worth the effort. Might be interesting for someone else to do, though.