The Economist (warning: their site has lots of scripts* and is likely to crash your browser) writes,
Harvard’s lawyers hired David Card, a prominent labour economist at the University of California, Berkeley. His model includes factors like the quality of a candidate’s high school, parents’ occupations and the disputed personal rating. Under these controls, Mr Card claims that Asian-American applicants are not disadvantaged compared with whites. But given that these factors are themselves correlated with race, Mr Card’s argument is statistically rather like saying that once you correct for racial bias, Harvard is not racially biased.
Pointer from Tyler Cowen. The previous day, Tyler simply said Card is wrong.
I know of three works by Card. One is his paper, with Krueger, claiming that a higher minimum wage raised employment in an area. The criticisms of that paper are persuasive. The second is a paper claiming that college attendance helps people from poor families, controlling for ability. As I wrote in this paper (see the appendix), what he claimed was an instrumental variable (meaning it should have no independent correlation with the dependent variable) was anything but. The third is this latest piece of arrogantly-expressed unpersuasive analysis.
Card was awarded the Clark Medal, which is on par with a Nobel Prize. His body of work is enormous, and perhaps I have encountered the only three times he has been wrong. Perhaps he is only untrustworthy when he wades into a politically sensitive topic. But if you are looking for an economist’s work to examine to see how well it replicates, I have a name for you.
*All media sites do this, but The Economist really goes over the top. Just once I would like to see a major media site that does not invite you to “get notifications” and such. They are all apparently listening to the same Internet consultant, who is an idiot. If they want to listen to someone, they should listen to me. I proposed a better model almost twenty years ago. I knew they would resist it for a while, but I never thought it would be for this long.
Arnold, Like you, I am not a fan of Card/Krueger. He did write a pretty good piece, though, one of his first, I believe, on the Mariel boat lift’s effect on wages in Florida.
Isn’t the point of being an economist to lie with statistics. I’m not just talking sensitive political topics. My dealings with economists in industry have almost all involved people trying to tell bald faced lies on behalf of their clients interests. They are lawyers with additional stats courses.
When I asked an economist of my acquaintance about how to judge techniques for inferring causation in non-RCT data, she mentioned Card as a “randomista” known for extolling the virtues of doing RCTs rather than trying to tease out causation from nonrandomized data. I was surprised and mentioned the minimum wage study, and she said that was an unusually strong natural experiment because of (if I recall correctly) the similar situations of restaurants immediately to each side of a state border. If that’s unusually strong I wonder if I should discount non-RCT claims even more broadly.
By the way, it seems to me that views of Card/Krueger are depressingly like views of the Kavanaugh allegations, in that there is a fact of the matter which it should be possible to look at dispassionately, but the strength of the correlation between views on the issue and political preferences suggests a lot of motivated reasoning on both sides.
What motivated reasoning do you see on the plaintiffs side? Seems to me the statistics go one way here. To the point where it’s an open secret everyone knows on both sides, regardless of public pronouncements.
Card’s absurd claims in the trial wouldn’t even make it through the low standard of review for those bogus articles.
There is a larger question of whether anybody but professional mathematicians can really be trusted with any complicated statistical analysis.
But after some experience and reflection and exposure to more trials, I’ve gradually become completely disenchanted with the use (and abuse) of expert-based statistical evidence in adversarial legal proceedings. The attempt to police these abuses up with more rigorous procedures for laying the foundations and post-Daubert standards has, in my judgment, simply failed.
Card has also written articles claiming that increasing the size of the workforce with immigrants has a de minimis effect on wages. This seems to be a fairly common notion among economists though. When it is politically expedient to do so, econ 101 notions go out the window. It is not at all uncommon for see articles that try to come up with reasons for you to pretend that wages don’t tend to fall when demand is stable and labor supply increases, or that wages don’t tend to increase when demand for labor increases and supply remains the same. Not sure that this is necessarily a bad thing at the social level. It just makes one feel vindicated in ignoring experts and trusting common sense.
A lot of public intellectuals are complaining about other commentators or politicians diminishing trust in institutions, as a grave social evil, because having trusted institutions are important. But that criticism is only valid when the institutions are trustworthy.
When they are rotten, corrupt, and abusive, the equal and opposite crime is to prop them up.
We didn’t have sufficiently robust mechanisms to maintain the reliable trustworthiness of our most important institutions from the currently widespread Social Failure Mode. The true criticism of enlightenment liberalism is that it is like disabling the social antivirus software protecting against this particular danger. Yeah, that antivirus software is really annoying sometimes, but the alternative is eventually every vital social organ succumbing to infection, and a rough and unpleasant transition to a low-trust society.
I gotta say, I just don’t understand the first two sentences of the last paragraph. What does “enlightenment liberalism” have to do with “[not] maintain[ing] the reliable trustworthiness of our most important institutions”?
What would “maintain the reliable trustworthiness of our most important institutions”?
I’m afraid the inferential distance between a mainstream point of view and the proposition is too large to explicate four levels down in a comment thread on a different topic. The theory of anacyclosis has been regarded respectably by political scholars for millennia now, including by Plato with his ‘kyklos’ in The Republic), and by Chinese writers in their analysis of dynasties and succession of regimes.
You don’t have to go full Turchin with ‘cliodynamics’ to agree with the notion that many historical cultural systems – regardless of their other merits and accomplishments – nevertheless carried within themselves the seeds of their own demise, or, at least, the ingredients that led to a transition to a different kind of state of affairs, cultural pattern, and scheme of social organization. On occasion there was a strong incentive to lie and deny the lack of genuine continuity and conservation of old forms, as with a post-Julian Roman ‘Republic’.
And the point is that the enlightenment-era political philosophy of individualistic and secular classical liberalism was born with a fundamental weakness. It intentionally disabled the social technologies behind what you might call ‘Chesterton’s Antivirus’ and thus left itself no adequate mechanism to prevent and defend itself certain kinds of Social Failure Modes, such as ideological ‘singularities’ (moral runaway reactions powered by lack of logically limiting principles) driven by competitive sanctimony and a self-perpetuating consensus among influential elites enabled by using ideological tests as a condition of entering and maintaining one’s membership in the club.
I completely “agree with the notion that many historical cultural systems – regardless of their other merits and accomplishments – nevertheless carried within themselves the seeds of their own demise, or, at least, the ingredients that led to a transition to a different kind of state of affairs, cultural pattern, and scheme of social organization.”
Yet I don’t see how “enlightenment liberalism” is particularly bad in that regard or what would be better.
Card’s point about so many applicants with perfect scores is well made