follow CCP

Recent blog entries
popular papers

What Is the "Science of Science Communication"?

Climate-Science Communication and the Measurement Problem

Ideology, Motivated Cognition, and Cognitive Reflection: An Experimental Study

Motivated Numeracy and Enlightened Self-Government

Ideology, Motivated Cognition, and Cognitive Reflection: An Experimental Study

Making Climate Science Communication Evidence-based—All the Way Down 

Neutral Principles, Motivated Cognition, and Some Problems for Constitutional Law 

Cultural Cognition of Scientific Consensus

The Tragedy of the Risk-Perception Commons: Science Literacy and Climate Change

"They Saw a Protest": Cognitive Illiberalism and the Speech-Conduct Distinction 

Geoengineering and the Science Communication Environment: a Cross-Cultural Experiment

Fixing the Communications Failure

Why We Are Poles Apart on Climate Change

The Cognitively Illiberal State 

Who Fears the HPV Vaccine, Who Doesn't, and Why? An Experimental Study

Cultural Cognition of the Risks and Benefits of Nanotechnology

Whose Eyes Are You Going to Believe? An Empirical Examination of Scott v. Harris

Cultural Cognition and Public Policy

Culture, Cognition, and Consent: Who Perceives What, and Why, in "Acquaintance Rape" Cases

Culture and Identity-Protective Cognition: Explaining the White Male Effect

Fear of Democracy: A Cultural Evaluation of Sunstein on Risk

Cultural Cognition as a Conception of the Cultural Theory of Risk

« Who fears childhood vaccines and why? Research report & project | Main | MAPKIA! Episode 31: what is the relationship between "environmental risk perception" predispositions, science comprehension & perceptions of the risks of (a) fracking & (b) GM foods?! »

MAPKIA! Episode 31 "Answer": culturally programmed risk predispositions alert to "fracking" but say "enh" (pretty much) to GM foods

Okay!  "Tomorrow" has arrived, which means it's time to real the "answer" "yesterday's" "MAPKIA!" episode.

As you no doubt recall, the question was ...

(i) What is the relationship between environmental-risk predispositions, as measured by ENVRISK_SCALE, and perceptions of GM food risks and fracking, respectively? And (ii), how, if at all, does science comprehension, as measured by SCICOMP, affect the relationship between people's environmental-risk predispositions and their perceptions of the dangers posed by GM food and fracking, respectively?

What made this an interesting question, I thought, was that both "fracking" and GM foods are novel risk sources.

If you read this blog ... Hmmm...

I was going to say if you read this blog this might surprise you, because in that case you have a weridly off-the-scale degree of interest in political debates over environmental risks and thus are grossly over-exposed to people discussing and arguing about fracking and GM food risks and what "the public" thinks about the same.  

But if you do regularly read this blog, then you, unlike most of the other weird people who fit that description, actually know that most Americans haven't heard of fracking and aren't too sure what GM foods are either.

Indeed, if you regularly read this blog (why do you? weird!), then you know that the claim "GM foods are to liberals what climate change is to conservatives!!" is an internet meme with no genuine empirical substance.  I've reported data multiple times showing that GM foods do not meaningfully divide ordinary members of the public along partisan or cultural lines.  The idea that they do is not a fact but a "rule" that one must accept to play a parlor game (one much less interesting than "MAPKIA!") that consists in coming up with just-so explanations for non-existent trends in public opinion.

But I thought, hey, let's give the claim that GM foods are politically polarized etc. as sympathetic a trial as possible. Let's take a look after turning up the resolution of our "cultural risk predisposition" microscope and see if there's anything going on. 

To make what I mean by that a bit clearer, let's step back and talk about different ways to measure latent risk predispositions.

"Cultural cognition" is one framework a person genuinely interested in facts about risk perceptions can use to operationalize the hypothesis that motivated reasoning shapes individuals' perceptions of culturally or politically contested risks.

What's distinctive about cultural cognition -- or at least most distinctive about it -- is how it specifies the latent motivating disposition.  Building on Douglas and Wildvasky's "cultural theory of risk," the cultural cognition framework posits that individuals will assess evidence (all kinds, from the inferences they draw from empirical data to the impressions they form with their own senses) in a manner that reinforces their connection to affinity groups, whose members share values or cultural worldviews that can be characterized along two dimensions--"hierarchy-egalitarianism" and "individualism-communitarianism."  Attitudinal scales, consisting of individual survey items, are used to measure the unobservable or latent risk predispositions that "motivate" this style of assessing information.

But there are other ways to operationalize the "motivated reasoning" explanation for conflict over risk.  E.g., one could treat conventional left-right political outlooks as the "motivator," and measure the predispositions that they generate with valid indicators, such as party identification and self-reported liberal-conservative ideology.

Do that, and in my view you aren't offering a different explanation for public controversy over risk and like facts. Rather you are just applying a different measurement scheme.

And for the most part, that scheme is inferior to the one associated with cultural cognition. By that, I mean (others might have other criteria for assessment, but to me these are the only ones that are worth any thoughtful person's time) that the cultural worldview measures of latent risk predispositions have more utility in explainining, predicting, and fashioning prescriptions than does any founded on left-right ideology.

I've illustrated this before by showing how much left-right measures understate the degree of cultural polarization that exists among ordinary, relatively nonpartisan members of the public (the vast number who are watching America's Funniest Pet Videos when tiny audiencies tune in to either Madow or O'Reily) on certain issues, including climate change.

Cultural worldviews are more discerning if one is trying to measure the unobserved or latent group affinities at work in this setting. 

But certainly it should be possible to come up with even more discerning measures still. In fact, in between blog posts, that's all I spend my time on (that and listening to Freddie Mecury albums).

In a previous blog post, I referred to an alternative measurement strategy that I identified with Leiserowtiz's notion of "interpretive communities."  In this approach, one measures the latent, shared risk predisposition of the different affinity groups' members by assessing their risk perceptions directly.  The risk perceptions are the indicators for the scale one forms to explore variance and test hypotheses about its sources and impact.

I formed a set of "interpretive community" measures by running factor analysis on a battery of risk perceptions assessed with the "industrial strength" measure.  The analysis identified two orthogonal latent "factors," which, based on their respective indicators, I labeled the "public safety" and "social-deviancy" risk predispositions.

How useful is this strategy for explaining, predicting, and forming prescriptions relating to contested risk?

The answer is "not at all" if one is interested in explaining etc. any of the risk perceptions that are the indicators of the "interpretive community" scale.  If one goes about things that way, then the explanans -- the interpregtive community (IC) scale-- has been analytically derived from the explanandum-- i.e., the risk one is trying to explain. This approach is obviously circular, and can yield no meaningful insight.

But if one is trying to make sense of perceptions of a novel or in any case not yet well understood risk perception, then a latent-measurement strategy like the IC one could well be quite helpful.

In that case, because the risk perception that one is interested in examining is not an indicator of the IC scale, there won't be the circularity that I just described.

In addition, the IC risk measure is likely to be more discerning with respect to that risk than the cultural cognition worldview scales.  

That's because individual risk perceptions are necessarily even more proximate, measurementwise, to the latent risk-perception predisposition that generates them than are latent-variable indicators relating to values and other individual characteristics.

Accordingly, if we think the relationship between a motivating predisposition and a risk perception might be weak -- or if we just aren't sure what the relationship might be -- then it might be quite sensible to use an IC method to measure the predisposition.

The inferences we'll be able to draw about why any relationship exists will be less suggestive of the operative social and psychological influences than ones we could have drawn if we measured the predisposition with indicators more remote ("distal") from individual risk perceptions. But if a valid IC scale picks up a relationshp that is too weak to have registered otherwise, then we'll know at least a bit more than we would have.  And if nothing shows up, we can be even more confident that the risk perception in question just isn't one that originates in the sort of dynamics that generate cultural cognition & like forms of motivated reasoning. . . . 

So I thought I'd try an IC apparoach for genetically modified foods rather than just repeat for the billionth time that there isn't any reason for characterizing them as a source of meaningful public conflict, much less one that pits "anti-science scared liberals" against conservatives. 

I formed a simple aggregate Likert scale by normalizing the sum of the (normalized) scores on responses to the industrial-strength risk perception measure as applied to global warming, nuclear power, toxic waste disposal, and air pollution.  I confirmed not only that the resulting scale was highly reliable (Cronbach's α = 0.82) but also that it generated a sharp division among individuals whose cultural outlooks-- "egalitarian communitarian" and "hierarch individualist," respectively--tend to divide over environmental and technological risks.

I confirmed too that the degree of cultural division associated with these risks increases as people with these outlooks score higher on a science-comprehension measure -- as one would expect if cultural cognition is motivating individuals to use their critical reasoning abilities to form identity-congruent rather than truth-congruent beliefs.

That gives me confidence that ENVRISK_SCALE, the aggregate Likert measure, supplies the high-resolution instrument I was after to examine GM food risk perceptions, and fracking ones, too, just for fun.

To appreciate how cool what one can see with ENVRISK_SCALE is, consider first the blurry, boring view one gets with a right-left political-outlook scale, which as I indicated supplies only a low-resolution measurement of the relevant motivating dispositions.

These scatterplots array members of the 1800-or-so-member, nationally representative sample with respect to their right-left political outlooks, measured with a composite scale formed by aggregating their responses to a party-identification measure and to a liberal-conservative ideology measure, and their perceptions of global warming, fracking, and GM food risks, all of which are assessed with the industrial-strength measure.

The visible diagonal pattern formed by the observations, which are colored "warm," red & orange for "high" risk concern" and "cold" green/blue for "low," shows that there is a strong right-left political influence on climate-change risk perceptions.

By the same token, the absence of much of a diagonal pattern for GM food risk perceptions illustrates how trivially political outlooks influence them.

To quantify this, I plotted regression lines, and also reported the R^2's, which reflect the "percentage of variance" in the respective risk perceptions (models here) "explained" by the right-left political outlook measure.  In the case of global warming, left-right outlooks explain an "impressively large!" 42% of the variance.  For GM food risks, political outlooks explain a humiliatingly small 2%.... But hey, don't let facts get in the way if you want to keep "explaining" why liberals are so worried about GM food risks!

Now, interestingly, right-left political outlooks explain 30% of the variance in fracking risk perceptions.  That's also "impressively large!"  Seriously, it is, because as I said, most members of the public don't know much if anything about fracking; I suspect at least 50% had never heard of it before the study!

I could turn up the resolution with cultural outlook measures but I've done that a bunch of times in the past and not seen anything interesting on GM foods.

So now let's zoom in with the even higher-resolution ENVRISK_SCALE.

Here I've just plotted fitted regression lines for the sample as a whole, and lowess ones for those subjects in the bottom 50% & top 10% on the "science comprehension" scale. I've left out global warming, for as I indicated, it makes zero sense to use an attitudinal scale to explain variance in one of its indicators.

Clearly, ENVRISK_SCALE is more discerning than are right-left political outlooks.  The R^2s have gone up a lot!

Indeed, at this point, I'm willing to accept that something at least slightly interesting seems to be going on with GM foods.  There are no "hard and fast" rules in assessing when an R^2 is "impressively large!" (I think the main value of R^2 is in comparing the relative fit or explanatory power of models, in fact).  But my practical sense is that the "action" that ENVRISK_SCALE is indeed meaningful, and suggestive of at least a weak predisposition among individuals, mainly egalitarian communitarians, who are on the "risk concerned" side of issues like climate change and nuclear power to worry.

The impact of science comprehension is also quite revealing, however, and cuts the other way!

As one would (or ought to) expect for risk perceptions that genuinely trigger motivated reasoning, science comprehension magnifies the polarizing effect of the disposition measured by ENVRISK_SCALE for fracking.

But it doesn't for GM foods.  Science comprehension predicts less risk concern, but it does so pretty uniformly across the range of the disposition measured by ENVRISK_SCALE.  

That suggests positions on GM foods aren't particularly important to anyone's identity.  If they were, then we'd expect the most science-comprehending members of competing groups to be picking up the scent of incipient conflict & assuming their usual vanguard role.

So on balance, I'm a little more open to the idea that GM foods could be a source of meaningful societal conflict--but only a tiny bit more.  More importantly, I'm less sure of what I believed than before & anticipate that someone or something might well surprise me here -- that would be great.

I'm really excited, though, about fracking!

Fracking already seems to warrant being viewed as a matter of cultural dispute despite its relatively novelty.  There's something about it that jolts individuals into assimilating their impressions of it to the ones they have on cluster of very familiar contested risks (climate, nuclear, air pollution, chemical wastes) that are the focus of the ENVRISK_SCALE.  That the most science-comprehending individuals are even more polarized on fracking suggests that the future for fracking might well look a lot like that for climate change.

As I adverted to last time, it's possible -- likely even-- that the wording of the fracking item, by referring to to "natural gas" being "extracted" from the earth, helped to cue relatively unfamiliar or even completely unfamiliar respondents as to what position to form.  But I think the settings in which people are likely to encounter information about fracking are likely to be comparably rich in such cues.

So watch out fracking industry!  And everyone else, for that matter.

Well, who won the game this particular "MAPKIA!" contest?

I'm going to have to say no one.

There were literally thousands of entries, most sent in via postcards from around the globe.

But for the most part, people just assumed that GM food risks perceptions would behave like the other risk perceptions measured by the ENVRISK_SCALE, both in the nature and extent of variance and their interaction with science comprehension.

Given the hundreds of thousands of Macanese children who never miss a "MAPKIA!" episode and who undertandably view its players as role models, I can't in good conscience declare anyone the winner under these circumstances! 

As I've emphasized -- zillioins of times -- cultural polarization on risks is the exception and not the rule. Ignoring the denominator-- as commentators sadly do all too often -- makes cogent explanations of this dynamic impossible

No problem whatsoever, of course, to predict a polarized future for GM food risks. But we're not there yet, and any interesting prediction of why that's where we'll end up would have to reflect a decent theoretical account of why GM foods will emerge as one of the lucky few risk sources that get to travel down the polarization path when so many don't.

Feel free to file your appeals, however, in the comments section!

PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (10)

Months ago, I was disabused, by data, of the misconception that fear of GMOs follows party lines, so of course I'm now following all this with the zeal of a convert.

I do have one thought. We know that most people have no idea what GMOs really are, and I take it that most people also have no idea what fracking is. Does the graph you posted use the actual wording of the questionnaire? If so, there's an explanation of "fracking," but none of "genetically modified food." And, although people are probably no more likely to know the meaning of "hydraulic fracturing" than "fracking," the description does give a basic sense of what's going on -- we're extracting a natural resource from the bowels of the earth. That could set off the alarm bells of respondents who see themselves as earth-friendly, even if they have no clue about the process. If there were an equally brief explanation of GM foods (foods in which a gene from another organism has been inserted, or some such), would that change the results? It might set off some of those same alarm bells.

Just a thought.

January 22, 2014 | Unregistered CommenterTamar Haspel

@Tamar: I agree.
But if "descriptions" are needed, then the question is -- what is one measuring? No description needed for climate, nuclear, guns, etc. B/c people really *do* know about those issues and have identity-related positions. If they don't know what something is, then one can bounce them all over w/ question wording.

I think the fracking item wording is not so different from what people would likely hear in a context in which they learn about fracking. But that is, as I tried to indicate here & in last post, a conjecture. The evocative wording weakens strength of inference that there's a there there. Agreed!

"Genetically modified" is actually pretty darn descriptive & one might guess emotionally evocative; it's interesting that it *doesn't* pick up anything! Certainly if GM foods were today a contested issue of any significance, then this term would, by itself, get a rise.

In the real world, too, efforts like those in calif & also in Washington state to kindle cultural conflagration w/ framing calculated to excite "us vs them" has failed. That's really really interesting. Why is the issue inert here when in Europe it really does divide public? I have no idea!

But what's interesting is to try to explain the facts -- not explain them away or ignore as most popular commentators do on GM foods

January 22, 2014 | Registered CommenterDan Kahan

Do you think this has anything to do with people's sense of agency or perception of control/influence on a risk? I'm thinking about the descriptive vs predictive nature of this kind of work, asking myself why X and not Y, (like we all are after reading this I'm sure), and one thing that does seem to stick out to me is a sense of control or agency in some risks and not others.

This is totally off the cuff and not at all thought-out, but it seems to me that risks where we see more polarization tend to be risks that a lot of people see as beyond their control. Am I way off base on that? Are there some good studies on this? Not necessarily even a person's actual measure or sense of control over a situation or risk, but his perception of that control?

For example, its my experience that a lot of people see something climate change or nuclear power as beyond their control, while GMOs tend to be associated with food and even if there are a lot of bigger implications, at the end of the day, an individual can choose not to eat certain foods. Is it possible that polarization could be somehow associated with the public's perception of individual agency in increasing/decreasing/avoiding a particular risk?

Perhaps a risk that feels "beyond our control" is more worrisome to those who are inclined to worry (more risk sensitive), and meanwhile less worrisome to the less risk sensitive types who just throw up hands and shrug- pushing those two extremes apart and creating more polarization?

What constitutes a perception of agency is a big question mark of course. Climate change, nuclear power, fracking- these seem to be represented as risks we can't individually control very well, either due to the hugely complex nature of the risk itself (climate change) or because the organizations that make decisions about these risks (nuclear power, fracking) are essentially untouchable by individuals, or maybe even because the risk is largely one perpetuated by other people whose behavior we can't control (guns). GMOs, on the other hand, may be something most people just associate with crops and food, which I can avoid if I want. Do what you want to the food supply but at the end of the day I can choose not to eat it. (Unless we don't label it...? and then maybe the issue becomes a little more active?) Same for vaccination? Generally, even if the more complicated nature of vaccination and community immunity, (much like the more complicated nature of seed regulation and other Monsanto 'evils'), is known (and tends to be the part that people react to when prompted), at the end of the day I can choose whether to vaccinate or not.

(This starts to sound like an argument centered around individualist vs communitarian values but that's not what I'm thinking. I hope it makes sense. I could imagine someone with a sense of very little agency being more inclined to then polarize over a risk/topic and where they land on that polarized spectrum might then be determined by their cultural worldview/risk sensitivity/etc.)

Is there a straightforward way to determine/measure/analyze how people tend to assess their own agency in risk avoidance, (are there any questions to this end in your giant survey?) and if so, would we see any patterns for which topics become polarized based population averages on that information?

This also raises the previously discussed question of what we're really measuring (i.e. description vs issue itself) BUT If you were to describe/word risks in a way that really emphasizes the aspects that are beyond people's control vs the aspects that are, I wonder how the results would change, if at all.

January 22, 2014 | Unregistered CommenterJen

To tack on to Jen's "beyond my control," how about "doesn't affect me."

Not many people see themselves as directly impacted by fracking, and people do have some measure of control over consuming GMO (or at least think they have a large measure of control even if they aren't aware of how much GMO they are ingesting).

Perhaps there is some distinction there w/r/t issues such as gun control, climate change, and nuclear energy - all of which people think might affect them in some way that is completely beyond their control.

Downstairs we talked of whether views on fracking orient in line with part ID or political ideology in areas which are directly affected, and even somewhat differently in different areas affected in association with the predominating political affiliation in those areas.

I suspect that the results of this analysis would show quite different results if the respondents were from PA or NY State.

That might also help explain the differences in the polarization around GMOs in the US and Europe - if for some reason people in Europe were more aware of the prevalence of GMOs?

January 22, 2014 | Unregistered CommenterJoshua

How does envrisk vary along conserv_repub? Apparently they don't correlate as well as I thought. It blew my mind a little that the two GM risk perception relationships (vs envrisk and vs conserv_repub) were so different.

January 23, 2014 | Unregistered CommenterScott Johnson


Your answer is here. As you can see R^2 is 0.27. For cultural worldviews, R^2 w/r/t envrisk_scale is 0.43, which is one way to characterize/comprehend the greater discernment of the cultural worldivew scales w/r/t the latent motivating risk predisposition

January 24, 2014 | Unregistered Commenterdmk38


just to round this out. The R^2 for model that regresses GM food risk perceptions on the cultural worldview variables is 0.04. Also, science comprehension doesn't interact w/ cultural worldviews w/r/t GM food risk peceptions.

so you can see the increase in resolution going from left-right, to CW to IC.

January 24, 2014 | Registered CommenterDan Kahan

@Jen - I think you have a point. A global warming threat would be earth-wide, fracking much less, the size of a watershed, maybe, while GM food is individual. We don't know exactly what the effect of releasing GM organisms into the biosphere will have, so in that sense it could be Earth-wide, but I'm not sure if many of the respondents would know about that aspect.

@Dan - I'm thinking that the more dimensions we have, the less variance. The cultural world view is 2-dimensional while the left/right is one dimensional. All things being equal, we would expect the left/right variance to be greater because it has less dimensions. Has this been factored into things?

January 24, 2014 | Unregistered CommenterFrankL

Right. I'm thinking about (for example), about how in Europe there is greater concern and polarization over GMOs: Is it because in Europe, the issue has been framed (by who? media? hm) as a more of a far-reaching, this-will-be-everywhere-you-can't-avoid-it, you-are-powerless-to-do-anything kind of problem? Most of what I've read does seem a lot more like that... While here in the US, it has been publicized more often as a matter of informed consent. Could this help explain the disparity? Can we run some kind of experiment to test this? When we look at people's response to these kinds of surveys, are we really comparing apples to apples? Even if we use the word GMO- is what people in the US respond to the same as what people in Europe respond to? This, of course is just circling back to Tamar's point so I'll quit here.

January 24, 2014 | Unregistered CommenterJen

@Jen - interesting quote from "Frankenfood debate...":

"While one of the historic issues that have kept GMOs out of the European Union has been protectionist concerns about large American corporations like Monsanto over-running European markets, the de facto (and sometimes explicit) bans on GM crops have actually created an environment that has stifled Europeans from developing their own competitive versions of GM products. In addition to denying farmers use of potentially cost-saving crops, the previous anti-GMO climate has produced a “brain drain” of scientists working on biotechnology in Europe."

January 26, 2014 | Unregistered CommenterFrankL

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>