Vaccine hesitancy, acupuncture mania, and the methodological challenge of making senses of "boutique risk-benefit perceptions" (BRBPs)
A thoughtful correspondent drew my attention to evidence of the persistence of enthusiasm for acupuncture despite evidence that it doesn’t have any actual benefit.
He was struck by the contrast with the mirror image resistance to evidence that the benefits of childhood vaccines far outweigh their risks.
What sorts of cultural outlook might there be, he wondered, that predisposes some people to believe that sticking needles into their bodies promotes health and others that doing so will compromise it?!
Maybe it’s a continuum with vaccine-hesitant people at one end and acupuncture devotees on the other?
Tongue-in-cheek on his part, but there’s an important point here about the role of fine-grained local influences on risk perception.
I am willing to bet that belief in the benefits of acupuncture will defy explanation by the sort of correlational, risk-predisposition profiling methods of which cultural cognition is an example.
Indeed, your comment actually highlights a research blind spot in the project to identify risk-perception propensities and to anticipate them through effective science communciation.
The counterproductive media din to contrary notwithstanding, vaccine hesitancy defies explanation by the sorts of cultural & like profiles that are so helpful in charting conflict over various other risks.
Same w/, oh, concern about pasteurized milk (and belief in the benefits of raw milk); fear of cell phone radiation; anxiety about drones; fluoridation of water etc.
There's some small segment of the US general population that believes in the effectiveness of acupuncture and its advantages over conventional medical treatments, which presumably those same people view as nonbeneficial or overly risky. I bet their views are unshared by the vast majority of people who share their cutlural commitments generally.
Let's call these outlier views "boutique risk-benefit perceptions" -- BRBPs.
But let’s agree with "fearless Dave" Ropeik’s consistent point that it is not satisfying to shrug off BRPBs as disconnected from any social context, as lacking any genuine social meaning, or as simply random patterns of risk perception, unamenable to systematic explanation ...
I think the problem in accounting for BRBPs has two related causes:
First, the sorts of characteristics that matter in BRBPs might be ones that are featured in schemes like cultural cognition but they always depend in addition on some local variable, one that makes those characteristics matter only in particular places, & indeed could make different sets of characteristics have different valences across space.
Second, the large-sample correlational studies that are used to examine such relationships in standard risk-profiling studies are unsuited for identifying the relevant indicators of BRPB because the local variable will resist being operationalized in such a study, and when it's omitted the remaining cultural characteristics will always lack any systematic relationship to the risk perception in question.
For an example of a closely related research problem where this dynamic is present and researchers just don't seem to get its significance, consider studies that purport to corroborate the trope that "rich, white, liberal, suburbanite parents" are anti-vax militants.
The most recent highly publicized study (or most recent highly publicized one I noticed) that purported to support this conclusion used a form of analysis that identifies "clusters" of school districts in which parents requested personal-belief exemptions in Calif.
The clusters, as hypothesized, were in particular highly affluent, white, suburban school districts in Marin county (bay area) and in certain demographically comparable suburban school districts in the vicinity of LA.
Taking the cue from the authors' own characterization of their results, the media widely reported the studying as confirming that “[t]he parents most likely to opt out of vaccines” are “typically white and well-to-do" etc.
One doctor, who has no training in or familiarity with the empirical study of risk perception and science communication, & who apparently no familiarity with the empirical methods used in this particular study either, excitedly proclaimed that "[w]hile the study looked only at California, ... similar patterns of demographics on parents would show up in other states as well."
Well, if so, then the conclusion will be that personal-exemption rates are not correlated with being "affluent, white, and suburban."
In a state-wide regression analysis, this same study showed that suburban schools (which are affluent and mainly white in California) had substantially lower personal-exemption rates.
There's no contradiction or even paradox here.
"Cluster" analysis is a statistical technique designed, in effect, to find outliers: concentrated patterns of results that defy the sort of distribution one would expect in a statistical model in which one variable or set of variables is treated as the "cause" of another generally.
If one can find such a cluster (i.e., one that can't be explained by a simple linear model that includes appropraite predictors), and can confidently rule out its appearance by chance, then necessarily one can infer that there is some other unobserved influence at work that is causing this unexpected concentration of whatever one is observing.
Generally speaking, cluster analysis isn't designed to identify causes of diseases or other like conditions. It is a form of analysis that tells you that there's some anomaly in need of explanation, almost certainly by other forms of empirical methods.
Strangely, the authors of the study apparently didn't get this.
They noted, with evident surprise, that "[s]suburban location had a negative relationship with PBEs [personal belief exemptions], opposite of what was anticipated given the maps of cluster assignments” -- & trot out a series of post hoc explanations for this supposed anomaly.
But there was no anomaly to explain.
If there are genuinely high-personal-exemption-rate clusters in certain white, affluent, suburban schools, that implies that that there isn't an association between those characteristics and high personal-exemption rates generally--indeed, that there is more likely a negative association between them (if the association weren't negative outside the clusters, the high concentration in the clusters would be more likely to generate a positive linear correlation overall, albeit a weak one).
Thus, the researchers, if it made sense for them to resort to spatial cluster analysis in the first place, should have anticipated the finding that "affluent, white, and suburban" school districts don’t have high personal-exemption rates generally.
Instead of announcing that their results had corroborated a common but incorrect stereotype, they should have recognized and advised readers that their study shows that in fact the influence that accounts for higher personal exemption rates in these schools is not that they are “affluent, white, and suburban” -- and is necessarily still unaccounted for!
They should also have called attention to the surplus of personal-exemption rate requests in school districts that are non-suburban-- in fact, among students in charter schools, whose attendees are more likely to be poor and minority.
I don't know why there would be higher exemption rates in students attending those schools. I seriously doubt that parents of these children are teeming with anti-vax sentiment. More likely, there’s a hole in the universal-vaccination net that should be identified and repaired here.
But the point is, researchers (at least those looking for the truth and not for the attention they can get for confirming a congenial misconception) aren't going to find out what influences, cultural or otherwise, explain vaccine hesitancy or ambivalence using general-population correlational studies. The influences are too local, too fine grained, to be picked up by such means.
Indeed, the "cluster" analysis methodology used in this and other studies is proof that something else-- something still not observed -- is causing such behavior in these areas.
It's something that necessarily evades the sorts of profiles one can identify using the sorts of attitudes and characteristics one can measure with a general-population survey.
That's exactly what sets BRBPs apart from other types of risk perceptions.
BRBPs fall into a blind spot in the study of risk perception and science of science communication.
We need valid empirical methods to remedy that.