3 ms·
Family income shows different tends in the marshmallow test. Kids from rich families often do much better. To my knowledge, no one thinks this is because rich k
by godelski 7y ago
Family income shows different tends in the marshmallow test. Kids from rich families often do much better. To my knowledge, no one thinks this is because rich kids come from better genetic stock (you'd be amazed how many experiments and studies change when you control for income levels). This supports the gp's hypothesis of the causality stack being flipped upside-down. Having more money leads to less stress, because you have less want. If you're a child of wealth you may not directly have the money, but you sure have the benefits of it.
- gwern 7y ago> To my knowledge, no one thinks this is because rich kids come from better genetic stock (you'd be amazed how many experiments and studies change when you control for income levels). They do, they're just polite about it because no one wants to put it that way. However, this is just Herrnstein's syllogism in action: if intelligence/education are heritable, education leads to higher incomes, then higher income families will have higher genetic potential which is passed onto kids (especially with the existing high levels of assortative mating on education). And this is precisely what we see with polygenic scores: there certainly are class differences in means.
- Bartweiss 7y agoI'm not surprised if it's true, but is there data showing that kids from rich families do better on the marshmallow test? I can't find it, and I'd be pretty interested to see the size of the effect. The 2018 replication just added a single control for a whole battery of environmental factors. Family income was included (and correlates like parent's education), but it also lumped in biological factors (e.g. birth weight) and child-psychological factors (e.g. cognitive performance one year earlier). Given that, it's really hard to understand how much impact each specific element had.
- godelski 7y agoHere's a study [0]. From my understanding every behavioral economist would not be surprised by the results. You'll notice the mention of the 2/3 due to "the presence of controls for family background, early cognitive ability, and the home environment." That's family income. [0] https://journals.sagepub.com/doi/abs/10.1177/0956797618761661 https://journals.sagepub.com/doi/abs/10.1177/095679761876166...
- Bartweiss 7y agoThat's the same study I was referring to, though. They didn't provide the impact of any individual control variable, just an aggregate of about eight different factors that are both social and biological. Also, they didn't show that their control corresponds to marshmallow wait time specifically, but instead that it's at least as good a predictor of academic outcomes. (Which means the control and wait time must be broadly correlated, but not necessarily that their 'errors' are aligned.) What I'm hoping for is something that specifically breaks out the relationship of family income to "marshmallow score", even without a long-term followup. I certainly believe that they correlate, but I'd like to see how strongly, and how it varies across what range of incomes. Even among people who agree that wealth drives marshmallow patience, explanations range from "wealthier families can consistently fulfill long-term commitments" to "food insecurity makes kids take food now". An obvious way to differentiate those theories is to see whether a wealth/time correlation remains strong past the point of food insecurity. (Since the original study was limited to relatively wealthy families, I expect it does.) I agree that behavioral economists wouldn't be surprised by this result, but I'm not sure they're sold on "it's just wealth", either. While news stories are calling this a failed replication, I'm seeing economists argue that this was a successful replication of the test's predictive power, and only indicts some specific casual theories proposed to explain it. In particular, several of the control variables in the replication are measures of child academic performance, so without an individual-factor breakdown, there's a risk that this is effectively controlling for the variable being tested. I think the "patience makes you successful in life" narratives are laughably wrong, but I wish we had some clearer data on what the actual correlations are.