Sorry but I’m going to be a few days late and a few dollars short to all the interesting discussions over the next few weeks – some life changes ongoing. So I missed the big discussion on genes and education that broke out over the weekend, starting at Greg Mankiw’s blog: “The NY Times Economix blog offers us the above graph, showing that kids from higher income families get higher average SAT scores…Smart parents make more money and pass those good genes on to their offspring.”
I like a lot of the things that economists do. But there are other things they do that I’m not a huge fan of. A good friend, who deals with ethnography, law, music and international copyright has a rule that whenever an economist uses an interview with one person and tries to pass it off as a piece of ethnography (“I had an interesting conversation with the cab driver who picked me up from the airport…From this conversation, I think it is safe to say that the local population is behind rapid capital market deregulation”) her mind wanders away from the argument. For me, the same thing happens whenever economists start talking about genetics.
There’s nothing in the education of economists, which is training in a subset of engineering math techniques known as ‘convex optimization’, that prepares one for the biological sciences. Is there an economics class lab that compares to organic chemistry lab? I know economics people and I know bio/chem/biophysics people, and the techniques and research are way different. Now if economists approached their research into genetics in a manner that was “here’s an interesting thing…” as opposed to “here we have clearly solved for genetic influence…”, humbler conclusions, that would be one thing. But it tends to be of the “fighting out the door” manner.
Inheritance
Here is a good writeup of the last round of follow-up with Mankiw at Marginal Revolutions, along with this provocative graph:
The graph does not say that adopted children necessarily have low income. On the contrary, some have high and some have low income and the same is true of biological children. What the graph says is that higher parental income predicts higher child income but only for biological children and not for adoptees.
Since the children in question were randomly assigned to families in the United States from Korea the researcher, Bruce Sacerdote in his paper “What Happens When We Randomly Assign Children to Families?”, uses the data to find many interesting results; adoptees education is correlated with parent’s education, but not as much as non-adoptees. I’m more interested in the way it has been characterized in the current debate. Can that graph really be true?
Age
The data ‘Public Use Data Set of Adoptees’ is available on Professor Sacerdote’s webpage. Taking a peek at it, the adoptees average age is 28, while non-adoptees is 32. Well that’s a problem for the graph. The red line in question is 4 years old, and an important 4 years – 28 year olds may still be in graduate school, volunteering, traveling, etc., while people in their 30s are entering what is a more mature (and important for us, consistent) earning potential years.
Gender
How do the genders match up? To be an economist about it, the frustrating thing is that the study assumes that the ‘treatment effect for nurture’ is consistent across the entire cross-panel within families, but girls and boys are nurtured in different ways in our society. And 70% of the adoptees are female. Only 39% of the non-adoptees are female. Well that’s a problem for the graph. Women make less than men. Also, it isn’t clear at all that high income status transfers privilege the same way to daughters as it does to sons. Perhaps it is easier for male children to inherit social and cultural capital that is deployable into economic capital than it is for female children.
Nepotism has a gendered connotation, coming from the greek word usually reserved for nephew or grandson. Though Liz Cheney, Megan McCain and now Jenna Bush are making headways into breaking the glass ceiling of nepotism, and I’m rooting for them to help achieve a world in which the powerful can transmit their privilege across merit-less generations regardless of color or gender, I’m not sure if it applies for this sample study.
Anyway, the paper acknowledges this, noting “The survey measure of family income is much higher for the non-adoptees than for the adoptees: $61,000 per year versus $42,000 per year [as in the graph above]. But this huge difference narrows to $1,600 when I control for age, education, and gender.”
Income
So $1,600 is very narrow. My next question is how do they measure income? Three-year averages? Five-year averages? How do they account for investment income? What’s the minimum value they report income in – $1,000? $100? Well, checking the data and research the income is a box in a questioneer the mother fills out about what she thinks her children make. The box are in $20,000 increments (she checks $10K, $25K, $40K, $60K, $80K, $125K, $175K, $200K). Does your mom know what box your income goes in? Mine wouldn’t.
So this strikes me as a major problem for the graph. Your Mom guessing, in $20,000 increments, what your income is not the best proxy for actual income, and it seems like a rather blunt sword to use to declare the knot of “Nature/Nuture” cut. I think the lack of granularity among the categories alone could easily noise out that $1,600, no?
Just for kicks, I re-estimated the graph just using people in their 30s to cancel out the “people hanging out in their 20s” effect:
The x-axis isn’t draw to proportion like the top graph (I can’t get excel or google documents to do it); however there is a correlation, a small but significant one; beta of .1, t-stat of 3.
This study introduces a lot of interesting questions for follow-up, especially with the way education levels transmit across generations. Personally I’d like to see work in how the habitus of female adoptees from foreign countries are constructed to deal with the fields of race and gender in our country, especially in regards to educational opportunities. But the idea that this study is the end of a discussion rather than the beginning of one isn’t giving it credit for what it does accomplish, and gives off a bad impression of what economists can do with subtly.
Rortybomb 
It is mind-boggling how economists are unable to process simple causal relationships like “I bet parents who have more income can afford expensive private schools and SAT prep courses.” Nope. They always have to leap straight into eugenics.
That statement in the opening paragraph is so transparently stupid that I have to wonder if the authors made it just to get attention. An undergraduate (let alone anyone with a PhD) would understand that “Well, this observed variance can only be explained by genetic differences” is an unwarranted conclusion.
Judging from my experience of getting a BA in econ a lot of what they teach is to believe whatever story sounds the cruelest and the shittiest. Economists get off on pissing people off. Mankiw is either a perfect example of this or a complete moron. Maybe both.
Great post once again Mike. I have to agree with you that it is annoying to see economists feel entitled to comment on just about anything just because they know how to run a regression in Stata or Eviews or something (dont know what is used these days). I dont know if this is one consequence of the popularity of pop economics/freakonomics, but it’s funny how a guy like Mankiw, who has done first-rate research in economics, feels like he is qualified to go peddle some half-assed analysis on something he probably knows very little about. Weird double-standards.
Pingback: Genetics and Income « Modeled Behavior
“I know bio/chem/biophysics people, and the techniques and research are way different”
As a bio/chem type trained person who is working in model system genetics, who teaches undergrad and grad level genetics, I gotta say bio/chem/biophys is not genetics. And genetics (like I use and teach) is really really different from the genetics you are talking about. Population genetics is a hard field. Human population genetics trips up even the really best geneticists. Add in the really loaded topics of psychology and achievement and what you get is most often a mess.
Man, this is a fantastic post.
Pingback: ginandtacos.com » Blog Archive » HERE IS A GRAPH. NOW LET’S DO SOME EUGENICS.
Excellent work as always, Mike!
I’d just like to augment the list of survey errors by pointing out that people who are making $250k are more likely to overestimate the income of others, including their own kid.
Also, where is the correlation between adopted children’s income and their foster parents’ income?
So exactly what part of Mankiw’s point are you disagreeing with? His post was about omitted variable bias. You’ve made no response to that and he’s the one “fighting out the door”.
Well, checking the data and research the income is a box in a questioneer the mother fills out about what she thinks her children make.
ISTM that an even bigger problem with this methodology is that your mom knows whether or not you were adopted. That could subconsciously influence her expectations for you. (Or, for that matter, her treatment of you during the whole time she was raising you – but if you admit *that* it imperils the whole idea of using adopted vs. non-adopted children as comparable.) If your *method of measuring the dependent variable* is itself dependent on the independent variable, then you can’t draw any conclusions about the *actual value* of the dependent variable being dependent on the independent variable, even if the measured value is.
The opposite comparison (i.e. if adopted children’s income is strongly predicted by their genetic parents’ income) would be much stronger evidence for genetic factors, if there was any evidence for it (since, if the adoption is early enough and there is no contact with the genetic parents, there would be almost no non-genetic explanation possible; and a study of “adopted children whose genetic parents were poor” vs. “adopted children whose genetic parents were rich” wouldn’t be troubled by the kinds of systematic differences between adopted and non-adopted children that would otherwise be a problem). Yet it never seems to come up in these arguments.
P.S. Income, of course, is a very very VERY bad proxy for intelligence. It is, however, a much better proxy for “desire to make a lot of money, as opposed to other things you could do with your life” (or, in short, greed). It would be interesting to know if THAT is strongly genetically influenced, but doesn’t have much bearing on questions of intelligence. But, unfortunately, Mankiw doesn’t consider this variable at all in his discussion of hidden variable bias (!).
Since income is an outcome of labor markets, and labor economics is a significant subfield of economics, I’m not sure how you write this:
“There’s nothing in the education of economists… that prepares one for the biological sciences.”
Income inequality, discrimination, returns to education… these are all topics with which economists have long studied and made large contributions to. Whether you agree with the rational agent model or not, Gary Becker’s contribution to these fields is undeniable and important. James Heckman and Daniel McFadden did win the nobel prize for their “development of theory and methods for analyzing selective samples” and “…development of theory and methods for analyzing discrete choice” respectively, two concepts which are sort of important in the field wouldn’t you say?
You ask whether economists have any training comparable to an organic chemistry lab? My first answer would be to, again, point to a nobelist. In this case, Vernon Smith, who won the prize for his utilization of lab work in economics. But more importantly, is there any training you get in an organic chemistry lab that prepares you to understand the determinants of labor market outcomes?
Pingback: Hot Links: World of Seinfeld « The Reformed Broker
the problem is that you are all seeing ones income as some kind of determinant of the quality of a human being…how would ghandi stack up on your charts?
Terrific post. A question: is the racial composition of the adoptee pool the same as the non-adoptee pool?
Suppose a a Korean child is adopted by a white family. Many of the soft cases of nepotism may not be available. Sure, the parents will work hard to make their connections/opportunities available. But all the helpful social moments – “hey, aren’t you Bill Smith’s kid”/”you really remind me of your old man”/”you’re going to love XYZ sorority, your mom was XYZ president” – don’t come naturally. OTHER people don’t acknowledge your legitimacy, even if your parents shower you with love.
If we are using small sample sets, look at the media’s treatment of Meghan and Bridget McCain. It could, of course, simply be that Bridget does not want attention, but I have my doubts that people turn to her for some sort of inside angle to the McCains the same way they look to Meghan. Says nothing about genetics or aptitude and everything about our assumptions about other families.
great post! Also, you can argue this from the other side, which you are actually doing as well.. That is, it’s not just that economists don’t (and have no reason to) understand genetics. But also that there are simply enormous factors BEFORE we get to genetics, such as age, gender, access to education/tutoring etc etc etc, that so clearly influence the outcomes being measured.
Even before getting to the sweeping generalizations about genes, you have to account for these other things that we know or could reasonably expect to affect that outcome. Since this is so rarely done, the focus on genes is pernicious because it deflects our attention from factors that are not only easier to observe but also likely easier to affect through policy and activism..
“The graph does not say that adopted children necessarily have low income. On the contrary, some have high and some have low income and the same is true of biological children”
Is that statement true? Doesn´t the fact that the curve for adoptees is flat precisely show that there´s no dispersion in the adoptee´s income? You see that all of them move around the 40k line, while the curve for non-adoptees ranges from less than 40k to almost 80k.
Pingback: Matthew Yglesias » Inequality Begets Inequality
You have to give Greg this: he is consistent.
As an economist, he is clueless. He had the “cojones” to write this after the housing bubble and the Great Recession:
“Despite the enormity of recent events, the principles of economics are largely unchanged. Students still need to learn about the gains from trade, supply and demand, the efficiency properties of market outcomes, and so on. These topics will remain the bread-and-butter of introductory courses.” (That Freshman Course Won’t Be Quite the Same, New York Times May 23 2009)
And now he shows he is equally capable as a biologist. I guess that’s why they call economics “the dismal science”.
Pingback: Do Smart, Hard-Working People Deserve to Make More Money? « The Baseline Scenario
A link to a Wall Street Journal (WSJ!!!) blog talking about economics. When one hears this kind of things from the WSJ, is a sign that something is happening:
“The pain of the financial crisis has economists striving to understand precisely why it happened and how to prevent a repeat”.
“The crisis exposed the inadequacy of economists’ traditional tool kit, forcing them to revisit questions many had long thought answered, such as how to tame disruptive boom-and-bust cycles”.
“‘We could be looking at a paradigm shift,’ says Frederic Mishkin, a former Federal Reserve governor now at Columbia University”.
Crisis Compels Economists To Reach for New Paradigm
http://online.wsj.com/article/SB125720159912223873.html
A pitty nobody told Greg about this.
PS: I don’t believe Geanakoplos have the correct answer, either. But it is a start.
Mike Konczal drastically understates the deeply repugnant thinking behind Mankiw’s assertions. Mankiw claims to be making a point about omitted variable bias. But the comparison of adopted and non-adopted children is riddled with potentially crucial omitted variables:
1. What about emotional factors among adoptees as opposed to non? Were there controls for children knowing that they were adopted?
2. Do wealthier families fret more about ‘preserving the family name’? Do they treat biological children preferentially?
3. Were there controls for families in which all children were adopted as opposed to families where some children were adopted and some were not?
All plausible concerns. It took _seconds_ for them to occur to me. But Mankiw is so driven by right-wing, dogmatic bigotry that it either didn’t occur to him at all — or he concealed the facts. That latter option would fit perfectly with right-wing practice of the past 30 years.