Evangelicals and the nonreligious do differ in science literacy

Earlier this week, I quoted this from an op-ed in the LA Times:

I recently conducted survey research comparing the most conservative of Protestants â those who identify with a conservative Protestant denomination, attend church regularly and take the Bible literally, or about 11% of the population in my analysis â with those who do not participate in any religion. The conservative Protestants are equally likely to understand scientific methods, to know scientific facts and to claim knowledge of science. They are as likely as the nonreligious to have majored in science or to have a scientific occupation.

The paper reporting this research is not in press yet, but I thought I'd try replicating the results. After all, the basic information needed to evaluate these claims is available through the General Social Survey, a massive survey that has been conducted by the National Opinion Research Center since 1972. The GSS has a lot of questions about people's religious views and background, and since 2006, has included a science literacy survey on behalf of the National Science Foundation (previously discussed here).

The data are there to evaluate the claims about science literacy, about science methods, about majoring in science, and scientific occupations, but let's start with the claim about science literacy I made in the previous post. To test Evans's claim about science literacy, I used the NSF's science literacy questions. I defined the "nonreligious" as those who either rarely attend church (less than once a year) or who answer "none" when asked what religion they belong to. It should be noted that these "nones" are a growing group in the US, and are not necessarily atheists: they just don't associate themselves with traditional religion. There's a question where folks are asked if they consider themselves fundamentalists; that tends to overestimate the fundamentalist fraction of the country, so I instead tried to replicate Evans's protocol: I separated out people who believe the bible is the word of God (not inspired by God or a book of fables), and who attend church at least once a month. I didn't try to isolate fundamentalist denominations, because those denominational ties are getting weaker in many cases. In any event, this screen is closer to the standard set of screens used to identify evangelicals, rather than fundamentalists per se. If anything, these samples are over-inclusive, which would tend to diminish any differences, biasing us toward Evans's result.

The graph above shows what percentage of members of each group answered a given number of science literacy questions right. As you can see, the curve for self-described fundamentalists matches the curve for evangelicals nicely: many people treat the terms as synonymous, so it's not unreasonable that these might reflect very similar populations. Note also that these groups both have median science literacy substantially lower than that of the general US, and much lower than that of the self-describedly nonreligious. The median score for the nonreligious was 8, for evangelicals and self-described fundamentalists it was 5, and for the general US, it was 6.

This standard set of questions does include a question about evolution and another about the Big Bang, but those two questions cannot account for a difference of 3 right answers between evangelicals and the nonreligious!

Of course, religious attitudes are hardly the most important factors driving science literacy, and some of those other factors correlate with religious views. For instance, there's a correlation of -0.5 between science literacy and evangelical belief and correlation of 0.25 between science literacy and nonreligion. Science literacy has correlation of 0.48 with educational attainment (highest degree achieved), and 0.47 with ability on a vocabulary test. Both of those measures of academic ability and achievement correlate with religious attitudes as well. They also correlate with economic class, both personal income and parental income, as well as race and other class markers.

So I ran a multiple regression, incorporating nonreligion, evangelical belief, the vocabulary test, educational attainment, race. In this model, nonreligion was associated with getting about an extra half a question right, being evangelical with getting an extra question and quarter wrong. Having a college degree versus just graduating high school correlates to an extra question or so right, being white is associated with about an extra right answer, and every additional vocabulary word someone got right was associated with an extra half question right on the science survey.

It's possible to get a better model by incorporating interactions between variables, but the basic gist doesn't change (though the result for particular people might be). Evangelical identity and nonreligious identity interact strongly with vocabulary score and with race. You can see the large effects of verbal knowledge, of education, and of religious views in the graph above.
There's a lot going on in that graph, so let's pick it apart. On the top and right side, there's a histogram of answers on the science literacy and vocabulary tests, to help make sense of the scatterplot underneath the lines. Because it's impossible to get fractional answers right, I shifted the individual points around in the scatter plot, making all the data more visible, and showing the nice bivariate normal(ish) distribution of the data.

I plotted regression lines from each of four populations: evangelicals who stopped school after high school, the nonreligious who stopped after high school, evangelicals who stopped school after college, and the nonreligious who stopped after college. The bold lines are the regression models, and the lighter lines of the same color have their parameters drawn from the underlying distributions of the data, helping illustrate the internal variability of these individual models. To simplify the graph, I left out other variation in educational attainment (from not completing high school, not completing college, completing junior college, and completing graduate school); you should be able to eyeball where those lines would be. I didn't show variation related to race because evangelicals and the nonreligious both tend to be white, and factors affecting science literacy in different races are a whole different topic.

The topmost line, the group with the highest level of science literacy, includes the most-educated nonreligious folks, but note that increased literacy has less impact on this group than it does for the other groups. The other three lines are basically parallel, with the regression lines for nonreligious groups always above those from evangelicals (though that'd be different if I had included higher and lower education levels on the graph).

There's a lot of scatter in the data, so the fact that the models follow this strict ordering by religious view should not be taken to imply that no evangelicals are science literate, or that the nonreligious have uniformly high science literacy. These are just general statistical tendencies, and shouldn't be taken to show causation, either. The forces that drive people into evangelical traditions are complex, as are the forces driving science literacy and educational attainment. Those causes are interwoven, and while there may be direct causality, it's far more likely that shared causation is a more important factor over all.

I look forward to seeing Evans's paper, since it looks like he used a different set of science literacy questions, and his sampling was surely different than the broad sampling available from the GSS. If, for instance, he was looking at a sample of college students, he may have missed the shift in science literacy that we see between high school grads and college grads, which could bring the science literacy differences to within a margin that might be explained by the handling of questions on evolution and the origin of the universe.

But this is also a cautionary tale. I quoted lots of his essay, with only modest caveats, because it tends to support by own suspicions. When I first tweeted the story, Daniel Loxton retweeted it, saying "Skeptics need to know this." He then expressed some caution, noting: "It's been pointed out to me that I don't know the research @JoshRosenau describes, and can't properly evaluate it. Fair enough," explaining, "It agrees with what I already think: that people of faith are no less bright or curious or imperfectly science literate than non-believers." That's my belief, too, and the analysis above need not shake us from that basic principle. It does complicate that principle, because while I firmly believe religious and nonreligious folks are equally capable of science literacy, for some reason they don't achieve science literacy at the same rate.

These data don't tell us enough to unearth the causes of the differences we see, but I can speculate based on what I've read elsewhere. Fundamentalism and evangelicalism (to a lesser degree) are associated with authoritarian mindsets that seek hierarchy, and (somewhat paradoxically) with a desire for individualism. That's not a mindset that meshes well with science, where authority can be overthrown relatively easily, which tends to resist imposed hierarchy, and which favors communal effort over a selfish individualism (though, of course, science contains plenty of hierarchy and plenty of selfish individualistic nonsense). This, I'd argue, explains the massive political divide between scientists and the general public. Only 6% of scientists are Republicans and 9% who called themselves conservatives (52% of scientists consider themselves liberal). Compare that to the 23% of the public who called themselves Republicans in a contemporaneous survey of the US public, 37% who called themselves conservatives, and 20% who called themselves liberals.

That same hierarchal/individualist complex of traits tends to associate itself with efforts to impose antiscience policies (whether or not one thinks antiscience beliefs show a partisan divide, it's hard to point to antiscience policies that the left has imposed or even made a concerted effort to impose), and also with religious beliefs that validate hierarchal/individualistic beliefs. Evangelical and especially fundamentalist churches do that in spades, and are thus likely to attract people with a predisposition against science. That predisposition is likely to be deepened by sermons and by interactions with other antiscience folks in those religious communities, and in some cases (evolution most prominently) that antiscience position becomes not just an individual assessment of the science, but a cultural marker of a person as a member of a particular tribe.

Evolution became such a marker of fundamentalist identity in the 1920s, and global warming denial seems like it may well be headed that way as a marker of political conservatism today in the US, just as global warming acceptance and concern have become markers of political liberalism in the US. The beliefs of one's parents and cultural heroes will have a serious impact from a young age, laying a groundwork of religious views, of views on science, of science knowledge, and of attitudes and affect. These issues are correlated on generational scales, making causality really hard to tease part.

To be clear, the folks I'm labeling "antiscience" above would not call themselves antiscience, and would insist that they like science a lot. But, as I said in the previous post on Evans's work, they'd mean something different when they talk about science. They tend to conflate science and technology, and praise practical, applied science that lets them do new things, and oppose it when it places limits on them, or challenges beliefs they hold dear. In some cases, that will give the appearance of being pro-science (as when Newt Gingrich stands up for NIH funding), but it isn't motivated by an understanding of what science is, let alone an appreciation of science for science's sake. As a reminder, then-Speaker Gingrich happily funded NASA and the NIH, but famously killed the Office of Technology Assessment, Congress's independent advisor on science and technology.