Spend enough time around people who carefully read social science papers, and you’ll notice folks who notice that most policy effects are null, most individual behavioral effects are null, most not-null findings break down with principled replication attempts, and that most remaining studies should be treated with a high degree of skepticism, as their standing more than likely reflects folks not bothering to take them seriously and take them down.
Typically, I observe two general dispositional outcomes following these observations. One: social science is fake and scam and only engineering is real. Two: nothing really matters, your and my personal decisions aren’t actually big deals, policy fights don’t actually matter.
I think that Emily Oster represents the best possible outcome of viewpoint two. She wrote a book, Cribsheet, and my lay understanding of its message is the following: nothing really matters and you’re fine. You can have a glass of wine while pregnant. Your kids can watch some tv and be fine.
Here is this viewpoint in video form
There’s something seductive to this viewpoint. But I think it is incorrect. I’ll start with individual behaviors as an example.
I’ve noticed these arguments are applied to specific individual behaviors. “Exercise doesn’t lower weight.” “Diet doesn’t improve health.” “Eating chocolate doesn’t lower weight.” “Fasting doesn’t work.” “Posting your values only works for a day.” You’ve seen this too, I am guessing.
I’ve gone through periods of losing, for me, significant amounts of weight (10-15 pounds) and I suspect a study with a bounded timeframe (a month or so) would find similar results, if they focused on a specific decision of mine. You can probably isolate times in my life where I altered my eating habits, or my exercise, etc., and find a null association with my overall health.
Genetics Logic Applied to Personal Behaviors and Policy
My theory is that most of these “nothing really matters” conclusions and studies have an incorrect conceptualization of human behavior and society. I’ve been thinking of this after having read several popular-facing books on population genetics. Here is how Paige Harden, in her book The Genetic Lottery, describes how modern geneticists think about how our DNA matters:
“Thinking about combinations of many genetic variants, which can be concentrated to varying degrees in a single animal, might be unintuitive. If, like me, you first encountered genetics in high school biology, your introduction to genetics was Gregor Mendel and his pea plants. The pea plant characteristics that Mendel worked with (tall versus short, wrinkly versus smooth, green versus yellow) were determined by a single genetic variant. In contrast, the human characteristics we care most about—things like personality and mental disease, sexual behavior and longevity, intelligence test scores and educational attainment—are influenced by many (very, very, very many) genetic variants, each of which contributes only a tiny drop of water to the swimming pool of genes that make a difference. There is no single gene “for” being smart or outgoing or depressed. These outcomes are polygenic.”
If I understand this correctly, the bad old days involved researchers standing in front of a pile of a trillion pieces of genetic information and searching for the “gay gene” or the “education gene” or the “is more aggressive than the mean” gene. People would initially find a spurious correlation, declare that they found the “is more aggressive than the mean” gene, popular presses would glorify the study, TED talks would turn into popular books, and then a few years later other researchers would learn that this “is more aggressive than the mean” gene in fact doesn’t cause what it was purported to.
The good new days apparently involve researchers aggregating from the trillions of pieces of genetic information to determine overall traits and tendencies, or the multiple different genetic components that in combination push up the probability of being “more aggressive than the mean.”
Apply this to personal health. Exercising 25 minutes more today won’t affect my overall health. But overall health isn’t about one modular and manipulable piece of behavior, or firing synapse, or whatever. I think what that 25 minute piece of extra exercising is often used to proxy for is an accumulation of a whole bunch of different behaviors, mindsets, goals, dedications, etc. that people bundle together to become healthy. I will exercise more AND I’ll wake up at a consistent time AND I’ll not eat that one extra sun chip AND I’ll avoid my phone before bed to improve my sleep AND I’ll work on my stress AND I’ll walk on the treadmill when I read AND … etc. Any one of those can be swapped out with minor to no effect. But it’s the process of bundling dozens of behavioral and mental and emotional and routine behaviors together that can improve or worsen your health.
Similar to policy. 15$ versus 12$ minimum wage doesn’t affect wellbeing of lower income folks, or whatever. Or increase of medicare coverage by 2% doesn’t do much, or whatever. But I suspect that broadly shared wellbeing and thriving societies are created through the stitching together of multiple kinds of policies and organizational behaviors and activities that make for a denser web of support. In fact, focusing on just ONE policy to flip on and off reinforces the harsh logic of efficiency and minimal investment that probably itself is a driver of a bunch of negative social outcomes.
Similar with pregnancy. Yes, a glass of wine every now and then doesn’t matter. Or parenting: tv every now and then doesn’t matter. But you’re operating to shift the bundle of the whole of behaviors. Any one could go away, but the overall logic is to develop the thick, redundant, mutually reinforcing constellation of behaviors, dispositions, emotion management, etc., that accumulate to overall improvement and betterment.
So the analogy to me: studies of the stuff above are probably limited because they’re in the “education gene” mindset when educational outcomes are probably genome-wide, or whatever. Genetic inputs into human outcomes are the result of lots and lots of weird little minor correlations. Our behavioral and policy outcomes are similarly probably the outcome of thick, mutually reinforcing, redundant activities, any one of which could be swapped in or out without changing anything. Not exercising doesn’t really matter. But it also really does.