So, I was reading this article, “Robert F. Kennedy Jr. Has a New Plan to Paralyze the Vote Count in a Critical Swing State,” and I came to this bit:
After striving to get his name on the ballot in as many places as possible, Kennedy suspended his third-party run last month, endorsing Trump over Kamala Harris. The former candidate is now trying to take his name off the ballot in battleground states for fear that he might siphon off votes from Trump. Yet he is still fighting to keep his name on the ballot in other states, including New York. It is not entirely clear why Kennedy cares about staying on the ballot anywhere; his overall strategy, however, reflects an intent to destabilize the election and abridge the early voting period, as he did in North Carolina.
Kennedy was booted off the ballot in New York in August after a judge found that he had falsely claimed residence in the state on his nominating petition. (Smoking-gun evidence proved that he lied about living as a tenant at the listed address.) Every lower court affirmed that decision, agreeing that the false residence invalidated his candidacy. Yet Kennedy is now asking the U.S. Supreme Court to restore his name to the ballot in an emergency order. New York has already begun printing and mailing ballots, so this eleventh-hour change would massively disrupt early voting—but again, as in North Carolina, that seems to be the point. Kennedy’s lead counsel, Jed Rubenfeld, claimed that his client’s exclusion from the ballot condemns New Yorkers to be “irrevocably deprived” of their First Amendment right to support his campaign.
That name “Jed Rubenfeld” was vaguely familiar so I googled. He’s the Yale law school professor who one of his coworkers has described as “the most gifted constitutional theorist (not to mention the most elegant legal writer) of his generation” (yeah, the standards are low in legal academia) and I guess is more famous now for sexual misconduct allegations and being married to the mentor of J. D. Vance. All of that rang a bell, but I hadn’t realized before reading his wikipedia page that he’d represented an anti-vax organization in federal court and that he cowrote an article with Trump’s former Attorney General.
There are lots of legal academics on the other side, who are working in some way or another for the Democrats, so fair enough that Rubenfeld is offering his specious arguments on behalf of the Republican campaign. Sill, the anti-vax thing: that, along with election denial, that seems like it’s a bit over the top. I’d guess that Rubenfeld, like his Harvard colleague Adrian Vermeule, sees himself as more of a charming imp challenging preconceived notions, rather than a chaos agent who’d like to destroy the separation of powers.
To put it another way: My guess is that Rubenfeld thinks it will be the 1980s forever, and he’d like to continue hobnobbing with the rich and powerful and, just for fun, giving a poke in the eye to the sanctimonious liberals who he has to encounter every damn day at Yale.
This is now, that was then
Nowadays, anti-vax attitudes are highly tied to the political far right—Kennedy and Trump are part of the story, but it’s not just that. Indeed, back in 2020 when the government was working on the vaccine, Trump was criticized from the right on that.
Sometimes you hear people say that anti-vax sentiment used to be coming from the left and now comes from the right. This is related to another occasionally-stated but pretty much false claim that opposition to genetically modified foods was coming from the left.
Regarding vaccines, I happened to be cleaning out my inbox and noticed an email from 2015 from political scientist Eric Oliver:
I saw your recent Monkey Cage post about ideology, vaccines, autism [Is it really true that conservatives are more likely to believe that vaccines cause autism? Yes.] and thought you might be interested in the following.
Tom Wood and I have some recent survey data (SSI survey in January with 1281 respondents) on this and the relationship depends a lot on question wording, response distribution, and magical thinking.
We asked respondents if they believed vaccines caused autism and gave them the option of don’t believe (54 percent), not sure (37 percent), or believe (9 percent).
If you just compare believers to rest you get a breakdown by ideology that looks like this:
Believe Vaccines cause Autism
Very liberal 6
liberal 8
Moderate 8
Conservative 9
Very Conserv. 18However, if you break this down by a magical belief scale (believe in ESP, past lives, alternative medicine, power of positive thinking, manichean world view) you get distributions that look like this:
Magical Thinking
Low
High
Very liberal
2 15
liberal
3 15
Moderate
5 12
Conservative
6 14
Very Conserv.
7 29Still some small effects of ideology, but they pale in comparison to effects of magical thinking. If you run all of this in a regression equation with the standard demographic controls, the effects of ideology basically disappear once you take magical thinking into account.
In short, strong conservatives are more likely to think vaccines cause autism because a larger percentage of conservatives have magical beliefs (due partly to the fact that so many religious fundamentalists identify as strong conservatives).
This is all part of a book we’re working on called Enchanted America. It’s about how magical thinking organizes public opinion.
It looks like the book came out in 2018. I should read it!
This is NHST-based evidence right? In stats 101 we learn the correct interpretation of “lack of significance” is “not sure”.*
I’d like to learn about who got the right answer.
* If there was a significant difference/correlation then the correct answer would still be “not sure”, but that is a more advanced topic.
Anon:
This is related to the one-way-street fallacy. Yes, the lack of good evidence that vaccines cause autism does not make it impossible that vaccines cause autism; the lack of evidence is also consistent with the hypothesis that vaccines reduce autism, or that they have small positive effects on autism in some settings and negative in others. No effect is completely zero. The problem is when the lack of evidence is taken to support a particular “not sure” position.
To put it crudely: some people are saying the effect is 0, or effectively 0; other people are saying, on the basis of bad evidence, that the effect is +1 (on some scale); it would be a mistake to split the difference and say that, because of uncertainty, we should act as if the effect is +0.5 or +0.1 or whatever, or that, just to be safe we should assume a positive effect. In the absence of good theory or good data, we should just as soon be comfortable saying that the effect could be -0.5 or -0.1 or whatever.
It’s bad to assume certainty; it’s also bad to use discredited evidence to impute the sign or consistency of any purported effect.
I’ve never looked into but would assume (since this is standard for medicine) these beliefs are based on (some authority figure talking about) one or more studies that found lack of significance.
So there is no direct connection to the magnitude (we need to know sample size and statistical + systematic noise).
Maybe there is an upper bound of like “vaccines cause autism at a rate less than 1 in 10k injections” or something. Really there are so many problems with the framing (assumes all vaccines are equal, etc), that I don’t see how these beliefs can be based on anything but magical thinking.
Ie, there is a difference (so it was caused by vaccines) or we didn’t find a difference (so vaccines don’t cause autism). Both are logical fallacies.
I’m not sure whether vaccines can cause monkeys to fly out of RFK’s butt. And no one can prove that they don’t.
Good example. That is just as plausible as “exactly zero correlation between vaccinations and autism”, which I am 99.99% sure was the hypothesis that got tested.
You could just as well test the “zero RFK butt monkeys” hypothesis, then conclude vaccines do/don’t cause autism based on the result. That would be an equally (in)valid process.
If you would like to move beyond wild speculation: Instead come up with a model that explains the rate of observed monkeys, then see if it can predict new data.
Here are the basic scientific steps:
1) Plot autism cases over time
2) Come up with a model that explains it (setting autism effect to zero/negligible)
3) See if the model predicts new data
4) If it can, then we have some evidence vaccines don’t cause autism.
Did anyone ever do this?
Anoneuoid –
What’s your model for how vaccines cause autism? Have you tested it?
Anoneuoid
You are being silly. Yes, the NHST tests whether the correlation is exactly zero, but the data used provides more information than the obvious fact that the correlation is not exactly zero. For example, a confidence interval for the size of the effect is more informative than your portrayal. I do agree with your basic scientific steps, but they are more complex than you suggest. All you can find from those steps is whether there is a correlation. Myriad other factors might cause an increase in autism, as you well know. So, step 2 is far more complex than you suggest and cannot be ignored by appealing to step 3 (good predictions should be required, but are hardly conclusive, depending on the details of step 2).
I don’t put much stock in a survey question about beliefs telling us anything about whether vaccines and autism are really related. At best, it tells us something about who holds such beliefs and may be interesting to compare with other beliefs they hold (as the original post attempts to do). The important question of whether vaccines cause autism should use real data on both, not survey data on beliefs. I think you would agree. But your comment seems aimed at a straw man. It is true that too many studies simply test the hypothesis that there is zero effect and seem satisfied with a binary conclusion. But most studies are not that superficial. Are you actually trying to say that vaccines do cause autism – without having to say that. Because that is what your comment says to me. It suggests that there is no evidence to the contrary by ridiculing the worst NHST study you can imagine.
Not at all. In fact, the only way to establish causality is to built it into a model with predictive skill. Other than that, all you have is correlation.
I bet no one who believes “vaccines don’t cause autism” has bothered to even attempt such a thing. But would love to be proved wrong. It should be easy, a single example will suffice.
From the ever-changing CDC website, they list these 3 references:
Institute of Medicine. 2012. Adverse Effects of Vaccines: Evidence and Causality. Washington, DC: The National Academies Press. https://doi.org/10.17226/13164.
DeStefano, F., Price, C. S., & Weintraub, E. S. (2013). Increasing exposure to antibody-stimulating proteins and polysaccharides in vaccines is not associated with risk of autism. The Journal of pediatrics, 163(2), 561–567. https://doi.org/10.1016/j.jpeds.2013.02.001
DeStefano F, Karapurkar Bhasin T, Thompson WW, Yeargin-Allsopp M, Boyle C. Age at First Measles-Mumps-Rubella Vaccination in Children with Autism and School-Matched Control Subjects: A Population-Based Study in Metropolitan Atlanta.Pediatrics 2004 Feb;113(2):259-66.
I only looked at the abstracts so I won’t vouch for any of these. But they do appear to ask the right questions. They lack the predictive evaluation you ask for – as do almost all published studies in social science disciplines when regression modeling (in this case logistic regression) is used. I’ve seen discussion regarding the lack of validation in regression studies – while it has been defended a number of times, I don’t agree with omitting validation. So, I’d agree if you object to them not reporting how well these models predict new data. But I don’t think the lack of validation renders these studies meaningless.
So, perhaps your comment concerns a different issue. Can a statistical study be of any value without an evaluation of its predictive performance?
Here’s a comedic (sometimes perhaps a bit too much) video about the first and faked “vaccines cause autism” study, with references to various articles, like the replication ones that show no link (linked in a document under the video):
https://www.youtube.com/watch?v=8BIcAZxFfrc
Doesn’t look like any of those are relevant.
There should just be a plot of the number of autism cases over time. Then a line showing a model fit to past data followed by predicted values of future data.
Any model with predictive skill will do.
Anything goes for the abduction (guessing, hypothesis generating) step. But then you need to check that the predictions of your model actually do occur in the real world.
Here’s an overview article, with references, discussing counterevidence against the faked original “vaccines causes autism” n = 12 study:
https://doi.org/10.4103/0019-5545.82529
On a crude superficial data, MMR vaccination rates for children have been fairly stable over time (https://www.statista.com/statistics/385577/mmr-vaccination-rate-among-us-children-aged-19-35-months/) while autism rates have increased markedly over the same time period (https://www.cdc.gov/ncbddd/autism/data/index.html – with anywhere from a doubling to a 5 fold increase, depending on which measurement you use). So, the superficial picture is not encouraging for finding a link. Of course, better studies would need to include more relevant factors, and then there is the issue of whether the autism diagnosis has been stably applied over time. I don’t think we disagree about this, and I will agree that the literature (that I’ve been able to see) doesn’t provide what you are asking for. At the same time, there is a lot of information about potential mechanisms for a link, with overwhelming failure to find any link between MMR vaccines and autism.
As you commonly do, your comment asks for studies that don’t exist or meet your required standards, but stop short of declaring what you believe. Are you willing to say you think there is a causative relationship? Or are you willing to say you are “very” uncertain (say a 50-50 chance), or quantify how likely you think a causative link is? Just as your often-stated concerns about the net effects of COVID vaccines, you somehow avoid stating your beliefs – beyond stating that you think all the research is badly done.
Anon:
I don’t see any plot like I asked for in that paper, and I am certain it is not in the Wakefield paper either.
Dale:
I still don’t see any plot like I asked for in those links. Also, I do not care about why the discussion now “mmr” instead of “vaccines”, and why the data only goes back to 2000.
Here is the simple fact. No one is able to provide an example of (even an attempt at) actual science being applied to this question.
And that is asking to meet an extremely low bar. How many of these regression models have been produced on the topic while never checking the predictions of a single one?
So if you believe anything but “not sure” then its based on something other than science.
Plotting the data and seeing if one can explain all the trend seems like an absolute crap way of evaluating whether vaccines cause autism, or anything else for that matter. As long as there is some trend that is not perfectly lined up with the model, as it never is (for various reasons, such as there being a non-linear trend that is hard to predict), then this way seems to lead to the conclusions that “vaccines cause autism”, or, “butt-monkeys cause autism”, or “gender-equality causes autism”, or “the prevalence of Trump as a political figure causes autism”, or “aliens did it”, or whatever one would like to imagine as the cause. It seems to me to be the “vaccines causes autism” side that is making a specific claim. That side started out with a fraud doctor paid to help win a lawsuit, and with none of the predictions from the suggested mechanism panning out. The only way that explanation has survived is by proponents making their predictions vaguer – as for example evidenced by them now talking about some unspecified “vaccines”, when its was once about “mmr vaccines”, which allows them to creep back into the safe assertion that all possible ways to evaluate every vaccine has not yet been conducted, as it can never be, in an ever self-serving “vaccine of the gaps” explanation to autism.
You’ll have to create the plot yourself (I can’t post graphs here, but the data I linked to makes it easy). I picked MMR since it is one of the supposed vaccines linked to autism – of course, other vaccines can be used as well as longer time periods. But I’m not your research assistant, so I’ll leave that to you. Your final answer “not sure” is a cop out. Of course, I’m not sure – of anything, literally anything. But there is a difference between 10% not sure and 90% not sure. You just like to take contrary positions and complain that “real science” is not being done. The first linked reference I provided is a 900 page book exploring all the evidence (at that time) concerning vaccines and their safety or lack thereof. There is some evidence for some vaccines of deleterious side effects. Overwhelmingly, the evidence fails to find such links for most vaccines. While none of that book provides the graphs you want, they do more “real science” than those graphs – they actually explore the potential mechanisms that might tie vaccines to these side effects. I’m not vouching for these studies, but I frankly have more faith in them than your conclusion that you are “not sure.”
I agree 100%. The links shared in this thread do exactly that.
They collect some data and fit some equation. Then never predict anything with the equation, in fact it is never spoken of again. I would bet even the thought of testing the predictions never crossed the minds of most involved.
This is not science.
No, actual science requires much more work. You don’t just do the easiest first step over and over.
Warning: Car Analogy
It is like thinking sitting in the drivers seat of a parked car 1000 times means you know how to drive on the highway. And if 100 different people all do that, give each and claim they know how to drive on the highway, it doesn’t get more convincing.
I mean this isn’t even failed predictions. It is failure to even bother.
Anon: You are deliberately avoiding the rest of that comment. Weak. Just like all the people too afraid to recognize that “Trump causes autism”, as evidence that any model of autism without this as an effect will not perfectly predict new data.
Damn, that should be “as evidenced by how any model…”
Plot autism cases over time
So this assumes that there is a stable diagnostic criteria for autism over time, which a brief internet search will show is a bad assumption. The apparent increase in autism diagnoses over time is a result of changing criteria and more and earlier detection of autism.
Vaccines are irrelevant to this issue. Other research has found that the overwhelming cause of autism is genetic and that kid are born with it.
The only reason vaccines, mainly the MMR, are an issue is that children are most commonly diagnosed with autism around the same time that the MMR vaccine is administered. The old ‘post hoc, ergo propter hoc’ bias kicks in and of course the vaccine caused the autism.
I never assumed this. In fact I have argued over and over on this blog that you need to model testing/diagnosis, and the failure to do so has lead to lots of misinformation.
I would love to see this done for autism, then maybe that can be used as an example of a huge confound that inspires another look at many other phenomena.
I would expect many sacred cows get sacrificed if it became standard to quantify the role of testing/diagnosis.
“Zero correlation” would be falsified if vaccines reduce the rate of autism, but that would also falsify “vaccines cause autism”.
This is a good discussion which on part describes the history behind the changes in diagnosistic criteria:
https://www.npr.org/2016/09/09/493148713/neurotribes-examines-the-history-and-myths-of-the-autism-spectrum
Another interseting aspect discussed is the increase in awareness of autism – with recognition of autism increasing significantly after the movie Rain Man came out.
Anoneuoid –
I’m still waiting to read about your mechanistic model to explain a causal link between vaccines and autism, and how you tested that model with predictions.
And while you’re at it, do you have a mechanistic model to prove that vaccines don’t make monkeys fly out of RFK’s butt?
Instead of asking, “Do vaccines cause autism?,” try a different question: “Do vaccines cause me to be still here?” Some of us are old enough to have heard our parents lament the loss of their (many) siblings before childhood vaccinations were common.
The following website is very informative regarding the history of recommended vaccines:
https://www.chop.edu/vaccine-education-center/science-history/vaccine-history/developments-by-year
A closer look at the above website makes me wonder how parents can possibly keep track of all the different recommended vaccinations. And a tangential issue, how to pay for all these medical visits given the cost of health care in the U.S.
The confounding variable here is trust in media/academia.
On “vaccines cause autism” being due to magical thinking, I can’t decide if this is insightful (an underlying cause transcending political ideology) or tautological (people who have irrational beliefs have other irrational beliefs). Anyway, it’s fascinating; thanks for posting.
That’s a good point Raghu. Also, some of the way regression is used in these instances kind of misses the point. If the ‘root’ explanation really is magical thinking, it might still make sense to refer to the specific attitudes as broadly being a conservative thing if conservatives are overwhelmingly more likely to engage in magical thinking. I get the point statistically, of course, but it can paper over important and real observations in a way.
+1
Re “you get distributions that look like this”, I’m having extreme difficulty what the following, which is what I see in my Web browser, is telling me:
Magical Thinking
Low
High
Very liberal
2 15
liberal
3 15
Moderate
5 12
Conservative
6 14
Very Conserv.
7 29
I think it means 2% of “Low Magical and Very Liberal” respondents believe autism is caused by vaccines, while 7% of “High Magical and Very Liberal” respondents believe autism is caused by vaccines, and so on.
I had trouble understanding it at first, too.
Sorry, a typo on the second %:
“I think it means 2% of “Low Magical and Very Liberal” respondents believe autism is caused by vaccines, while 15% of “High Magical and Very Liberal” respondents believe autism is caused by vaccines, and so on.
I interpreted it differently: with Low and High magical thinking as column heads, so 2% of very liberal with low magical thinking and 15% of very liberal with high magical thinking think autism is caused by vaccines; 3% liberal with low and 15% liberal with high; 5% moderate with low and 12% moderate with high, … 7% very conservative with low and 29% very conservative with high.
I was replying to your first comment–now I see we are interpreting it the same way.
If vaccines cause autism then it’s win-win: in a globalised world we need more autism and fewer pathogens. Sadly they almost certainly don’t, and autistics will have to compete for mates the old fashioned way.
Fraac:
I hope that people who wish to do so can find life partners (not “mates”) by meeting people with common interests, values, and temperaments (not by “competing”). We’re people here, not some sort of theoretical version of an animal from a sociobiology textbook.
Okay but we aren’t *not* animals. We haven’t transcended the animal kingdom. That’s what AI is for. :)
Anon:
Yes, we’re real animals, we’re not some sort of theoretical version of an animal from a sociobiology textbook. Distancing expressions such as “compete for mates” have some value but they can also be dehumanizing in a scientistic way.
Andrew, Anon: kind of a humorous interaction.
As a human, my physiology similar to all the beings in the group we call “mammals” and highly similar to that of other great apes. It’s extremely difficult to deny that I am an animal.
OTOH, all mammals and certainly all great apes do, in fact, compete for mates and the competition is extremely intense. I’m not sure how that can be “dehumanizing” as the competition for mates is a key element of the success or failure of any organism with sexual reproduction. We see daily evidence of the competition in the press: apparently Michelle is competing with Jennifer (and has competed with others in the past) as we speak! The competition is such an important part of our existence that a significant segement of our economy is devoted to the sales of recordings of humans vocalizing about their success and failure in that competition. In that sense, Rock and Roll kind of actually is analogous to wolves howing at the moon!
Hilarious, thanks for that.
Anon:
The expression “compete for mates” does not seem to me an accurate description of the process of meeting people with common interests, values, and temperaments and pairing off with the goal of being life partners. I say this first because there is not necessarily any competition involved, second because mating is only part of partnership. I agree that sometimes people compete for mates, but there are a lot of other ways that people get together that don’t involve competition and where mating is not the highest priority.
I think the expression “compete for mates” is abstracting and somewhat dehumanizing in the same way as if you were to describe your interaction with your boss as a “dominance ritual.” There is a value in social science of using these sorts of abstract and dehumanizing concepts, as they can detach us from the details of social interactions and help us see larger patterns. Indeed, my colleagues and I wrote a paper with mathematical calculations of the rationality of voting, even though I don’t think people actually decide whether to vote by making mathematical calculations and weighing utilities. Using a rational-choice model gives some insight about voting, even though it’s abstracting and dehumanizing. I similarly think the expression “compete for mates” can have value–as long as we recognize that it is an extreme abstraction of the actual processes by which we find life partners in non-competitive environments in which we meet people with common interests, values, and temperaments.
For me there are a large amount of unknowable parameters in a vax/autism discussion. What is the concordance rate of an autism diagnosis by different diagnosers? Surely different vaccines produce different responses in humans. Do natural infections cause adverse effects? I think that we can know the number of home runs by the Red Sox versus the Yankees very precisely. The number of autistic children is not as knowable. Our arguments are loud because our facts are elusive.
You can look up Jed Rubenfeld’s political donations at opensecrets.org (as well as Adrian Vermuele, who just donated once, to Republican April Farris of Texas). It’s strictly to Democrats, including a punching bag (or eye-poking bag) like Hillary Clinton. My guess is that he takes the lawyerly view that every client deserves representation.