Do you believe that “humans and other living things have evolved over time”?

The other day on the sister blog we discussed a recent Pew Research survey that seemed to show that Republicans are becoming more partisan about evolution (or, as Paul Krugman put it, “So what happened after 2009 that might be driving Republican views? . . . Republicans are being driven to identify in all ways with their tribe — and the tribal belief system is dominated by anti-science fundamentalists”).

We presented some discussion and evidence from Dan Kahan suggesting that the evidence for such a change was not so clear at all. Kahan drew his conclusions from a more detailed analysis of the much-discussed Pew data, along with a comparison to a recent Gallup poll.

Also following up on this is sociologist David Wealiem, who pulls some more data into the discussion:

Although the Pew report mentions only the 2009 survey, the question has been asked a number of times since 2005. Here are the results—the numbers represent the percent saying “evolved” minus the percent saying “existed in their present form.”


The 2009 survey is an outlier—everyone was more likely to express support for evolution than they were in the surrounding years. I [Weakliem] am not sure why the results from this survey should be so different from the others, but I noticed that it focused on science, while the others had a mix of topics. So perhaps people felt more inclined to go along with the scientific consensus after answering a lot of questions on science and technology. Another possibility is that people who didn’t have much knowledge of science or faith in science were less likely to agree to participate in a survey on science. In any case, any comparison involving the 2009 results should be taken with a grain of salt.

If you omit the 2009 survey, partisan differences have become a bit wider over the whole period, not because Republicans have become less likely to believe in evolution, but because Democrats and independents have become substantially more likely to believe in it.

But, as Kahan notes, only 36% of surveyed Democrats believed “that humans and other living things have evolved over time due to natural processes”—with another 22% believing in evolution that was guided by a “supreme being.” (My own take on the supreme-being-guidance thing is summarized by item 1 on this list.)

Punked by Pew?

Regarding Weakliem’s post, Kahan writes:

This is true. I pointed out that for all we know 2009 evoluton belief was “high,” in which case prior level of belief (in creatonism) is rebounding etc.

That’s why it’s useful to look, too,, at Gallup, which has been asking same q. for longer time. So has NSF in its science indicators.

No one doubts there is partisan divide, but the idea that it could change a lot in 4 yrs, while certainly *possible,* would be pretty amazing. There’d be more evidence.

But in any event, now that we have all the data, we can see plainly that the only thing that happened even in Pew data was a reallocation of small # of Repubs between “divine guided evolution” (somethin akin to “intelligent design”) & “creationism.” There was no meaningful shift in proportion of Repubs rejecting *real* evolution (natural selection kind).

In fairness, Pew baited the misunderstanding trap by failing to release the the partisan breakdown for *entire* question in its initial “report”/”press release.” Not only concealed that Repubs hadn’t shifted on natural-selection evolution, but that of course the vast majority of Dems don’t accept “natural selection” version either — making “monkeys,” as it were, of the commentators who cited this poll to heap ridicule on Repubs for being “the party that rejects Darwin.”

Basically, the commenators were “punked” by Pew. Very very very unlike Pew, which is the only “public opinion survey” operation that does real opinion studies rather than “issue du jour.” I really love Pew precisely because it isn’t in the ‘opinion polling’ business but in the ‘public opinion analysis’ business. It employs super smart researchers who understand what survey items do & don’t measure & who use that understanding to enlarge knowledge of lots of complicated things, particularly relating to public understanding of science.

P.S. I would label the lines in Weakliem’s graph directly rather than with a legend, also I’d use the conventional red/blue/purple colors rather than black/red/green, and I’d label the y-axis more descriptively. But, hey, I didn’t make the graph, and it’s waaaay better than a table. I make these sort of graphing comments not to discourage or intimidate the Weakliems of the world but rather as suggestions so that the future graphs they make can be even more useful.

19 thoughts on “Do you believe that “humans and other living things have evolved over time”?

  1. Pingback: Friday links: the grand gender convergence, statistics vs. scotch, SO BUSY, and more (UPDATED) | Dynamic Ecology

  2. Well done, Andrew!

    I count myself in the tribe of Prof. Kahan, and his acolyte: Andrew—rigorous analysis, skeptic of partisans, and finding a theory to fit the facts, instead of the reverse. These principles cut across belief as a matter of faith, whether in science or religiosity.

    As a Tea Party sympathizer and believer in intelligent design, I pass Groucho Marx’s insightful test: “I wouldn’t want to join any club that would accept me as a member.” Dan rejects intelligent design as “goofy” and seems to regard the Tea Party as a fringe. Are Token Memberships available?

    Re list item 1 on supreme-being: yes, of course. It’s the only way to reconcile evolution (a very intelligently-designed process) with intelligent design.

    Where did the list come from? A lot of it is over my head, but I smell a learning opportunity. Are explanations available, or do I have to dig?

    • Anon: “and finding a theory to fit the facts, instead of the reverse”

      Is this not ex-post theorizing? I would have thought the right approach would be:

      “finding the right research design to [severely test? learn?, falsify? take your pick] a theory”.


      “find a theory to fit the facts, and then test it against new and different facts”.

  3. Thanks Andrews. I think this does a good job showing the danger of relying on two surveys to identify I trend. In my own work I usually try to work with a dozen or so surveys (where possible) when identifying temporal patterns. Too much less than this and you run the risk of having your assumptions confounded by outliers.

  4. I am a bit taken aback by Rep/Ind/Dem division. Is it relevant? We know that the parties (including independents) have different mixes of people with different education, upbringing, life experiences, interests etc. At first blush, I would suggest that those variables are more pertinent than the party affiliation itself. Or we are studying the political question of the type what happens with a school curriculum when the party control of something or other changes?

  5. > “… also I’d use the conventional red/blue/purple colors…”

    I’m a little embarrassed to ask but what are considered conventional colors for graphs? I usually go red/blue/black/green – or red/blue/green/black. I tend to avoid purple as I find I sometimes confuse it with blue (on other people’s graphs) if I’m not looking close. Never yellow. Never, ever yellow. Orange is pretty bad too. And if there are more than four plots on the graph then I try to rethink what I’m presenting because if I go with >4 then it’s probably going to be a very busy – and potentially confusing – graph.

    Related: What if any conventions are there for symbols and line styles on B&W graphs?

    • Chris:

      I just mean the now-standard coding of red for Republicans, blue for Democrats, and purple for independents, also best to label the lines directly rather than via a legend on the side.

      And I think that it’s usually better to avoid multiple lines on a graph if they aren’t clearly separated. Labeling lines with a legend seems like it’s almost always a bad idea, much better to put the labels right on the lines.

      • Is there a reason to use this “percent saying “evolved” minus the percent saying “existed in their present form.”” rather than simply “percent saying `evolved'”?

        They don’t allow both or neither to be ticked, right? I thought the ordinate metric was somewhat contrived.

      • > I just mean the now-standard coding of red for Republicans, blue for Democrats, and purple for independents

        Doh! Right. Got it.

        > Labeling lines with a legend seems like it’s almost always a bad idea, much better to put the labels right on the lines.

        Agreed. Thinking about it for a minute, graphs with labels on the lines generally are clearer. Unfortunately, legends are still standard practice amongst my peers and I. (Note to self: Start putting labels right on the lines.)

  6. On the graph:
    1. I agree that it would have been better to put labels on the lines rather than in a legend. I was just going with the default without thinking about it.
    2. The “evolved minus existed” was because there are some people (usually 5-10%) who say they don’t know. So I wanted a measure of sentiment among people who gave an opinion.

    On the larger issue of why look at party, party identity is fairly stable, so a big drop in the percentage of Republicans who say they believe in evolution over a four-year period would suggest that many Republicans had changed their minds. That sometimes happens with issues that are the subject of political debate, like the individual mandate: people who support a party follow their leaders. But evolution vs. creation has not been a top political issue over the last few years. So if there had been a big change in the distribution of opinions by party between 2009 and 2013, it would be an unusual event that needed an explanation (at least if you’re interested in politics).

  7. @Andrew, from erstwhile anon(ymous) (I hit submit before entering name–sorry.)

    On ex-post theorizing, yes, absolutely, once the earlier theory is found wanting. Hard not to see improvements in experimental design, then theory, once the results start to come in.

    I was skipping past the steps you identified as understood. No conflict here unless you don’t believe a theory should be revised after testing (experimentation), if some variant of “falsify” is the conclusion.

    The learning part is exactly what I had in mind. Speaking of which, “this list”–link right above “Punked by Pew?” header. I’d like to dig into this. Source? Annotations or commentary? Looks really interesting to me.

Comments are closed.