"nobody has really considered what they might look like to an outsider"

March 20, 2024 9:23 AM

More in my series curating work by finance expert Daniel Davies, this time focusing on academia and on the cultures and norms of research in general. In "why i am (still after all) an economist" (2023), he asks, "Is there anything that is actually definitional, something that you have to believe or you’re not an economist?" and offers his answer, which is that (I'd summarize) economics treats historical facts as descriptive but not necessarily prescriptive.

(As background on Davies's own educational history: he did the Philosophy, Politics and Economics course at Oxford University, and then London Business School's MSc in Finance since it "was substantially cheaper" than an MBA.)

"Creative arts and investing in systems", 2019: "one of the important outputs of a thriving fringe theatre scene is an educated and critical audience, which is the hidden crucial competitive advantage that the British media industry enjoys over most of its competitors."

"This year's A-level results are a fiasco – but the system was already broken" (August 2020): "when we see a big gap between the teachers’ predicted grades, and the 'adjusted' versions meant to bring those grades into line with previous years’ exam results, why would we necessarily believe that it was the teachers who were wrong and the exam results that were right?" (Related: "Grade inflation: It looks as though A-levels have got easier over the last 20 years, but perhaps it's a case of teachers increasing their productivity." (2007)

"LIBOR for the universities?" (2015) which asks: in academia, what scandals are waiting for public scrutiny of the type LIBOR got? (Among them: "Senior academics falsely claiming authorship".)

This isn’t a specific issue about financial sector corruption. It’s a general trend, one of gradual social re-assessment of whether the fiddles and skeletons of the past are going to be tolerated in the future.....

The things that all these kind of scandals have in common are a) common knowledge – they tend to be practices that everyone in the sector in question knows about, but nobody has really considered what they might look like to an outsider, and b) because of this, ubiquity – there is always a mass of evidence, because nobody in the past has seen it as something worth covering up.

A two-part analysis of adjunct pay problems, as of 2003.

"Republicans just don't like science" (2006): "...it's fundamentally wrong to assume that the deleterious effects on scientific research of the religious-mystical lobby are an unfortunate consequence of their ethical and religious beliefs. I think it's at least as plausible to assume that they are actually the purpose..."

"on the reading of difficult books (a guide to the bored)" (2023), on how to read regulatory documents, but also books by scholars: "a lot of the time, the author is carrying on an argument with other authors on the same subject...."

"Faking the physics" (2006): "If a sociologist can convince a jury of physicists that he is an expert on gravity waves, doesn't that mean he is one?" (quite a lot of comments in his Crooked Timber thread on the piece)

On the importance of proper data gathering: "ground truth" (2023). Relatedly, "From my email outbox, on behavioural economics ..." (2013): "it really lends itself to small-topic research of the sort that will get your PhD completed on schedule and/or produce a stream of publishable papers to demonstrate one's value to a department research assessment exercise."

Davies wrote a long-form review of Freakonomics between 2005 and 2009: Part 1, Part 2, "The Heterodox Theory of the Criminal Firm", Part 3, "Natural experiments ain't so Freaking Natural" (and addenda), Part 4, and Part 5, "How Freaked is Economics?". I never read Freakonomics and the most engaging part of the series, to me, is Part 4, which includes:

I've often joked that neoclassical economics and evolutionary psychology are unusual among social science fields because they're the only areas where you can gravely insult someone by agreeing with their theory (by telling an economist that he's only saying what he's saying because he's well paid for doing so, or telling an EP type that his latest paper is simply an attempt to draw attention to himself and gain status).....

There are a hell of a lot of "you scratch my back and I'll scratch yours" arrangements in the world. Some are illegal, some are disreputable, some are more-or-less standard and some are pretty much constitutive of what it is to behave honourably.

Humor and satire

on the Ayn Rand & charity beat in 2008, Davies notes:

A large charitable foundation attached to a bank has given the University of North Carolina Charlotte, among others a donation in return for making “Atlas Shrugged” compulsory reading.... This is essentially a bribe to the university and as far as I can see, taking a bribe to compromise your own vision of an undergraduate reading list doesn’t fit in particularly well with Rand’s particular ethics of self-reliance and independent thought.

I'm not always the best at detecting dry humor, but I think "A Solution To The Adjunct Problem" (2003) was satire, suggesting that "Universities should sell their professorships, at whatever price the market will bear" -- just as one used to be able to buy or sell British Army officer commissions.

And, again somewhat satirically, I think, "Religious education in schools is the cornerstone of a secular society, because it puts children off religion for life." (2006)

posted by brainwane (5 comments total)

He's a good writer but when it comes to topics that I know quite a bit about I find him often wrong. For example, on the 'natural experiments ain't so freaking natural' piece on Instrumental Variables, he makes a number of mistakes. The instrument needs to be as-if random, which is one of the only real requirements, and not just correlated with the IV but causal as well.

He writes that after IVs are done right you get a result where a variable 'it isn't correlated with the variance in the error term for finals marks'. But that's not it, its not correlated with the variance of the error term, its that its not correlated with the error term at all (correlated with a variance of an error term makes no sense outside a multilevel model).

He also writes here: ...."(because Welshness, closeted gayness, novelty hat purchase and FT readership are none of them correlated with beer drinking)" but again, they will be correlated with beer drinking, they just have no causal relationship. If they're not correlated AT ALL you the model makes no sense.

This is almost at the level of nonsense:
"As long as the left hand variables of your preliminary regression aren't themselves correlated with the variance in the original equation (in other words, as long as Welshness isn't itself correlated with drunkenness), and as long as the fit of the equation estimating Gamma is reasonably good, it will be OK."
posted by MisantropicPainforest at 9:35 AM on March 20 [2 favorites]

Welshness, closeted gayness, novelty hat purchase, and FT readership are correlated with Rugby playing in an effort to eliminate the confounding factor of beer consumption when attempting to measure the "thickness" of the populations. They are selected because they are known to have no correlation with beer drinking.
posted by betaray at 10:56 AM on March 20 [1 favorite]

And I guess my point is that the critique that the model makes no sense is the point.

While theoretically, yes, "as long as the fit of the equation estimating Gamma is reasonably good, it will be OK", but "In the real world (a place I have often visited), you aren't allowed to randomly pluck series out of the air and say that they are strongly correlated with the variable you want them to be instruments for."

That's why controlling for beer drinking using this approach doesn't work.

"So the take-away here is that weak instruments in IV estimation are really bad news, much much worse than poorly correlated regressors in normal regression analysis. It can actually be better from a mean-squared error point of view to just ignore the bias and do the ordinary regression, if the only instruments you can find are weak. This is important."
posted by betaray at 11:13 AM on March 20

Ha!

I am wholly suspicious of this outpouring of creativity on the part of economists, rather as I would suspect and fear a sudden outbreak of interest in stochastic calculus among teachers of modern dance.
posted by sixswitch at 2:14 PM on March 20

They are selected because they are known to have no correlation with beer drinking.

This is not even wrong. In the social world everything is correlated with everything else, but the assumptions of instrumental variables are CONDITIONAL independence, which is not even mentioned.

And I guess my point is that the critique that the model makes no sense is the point.

Its not that the model makes no sense, its that the core assumption is that there is no causal pathway from your instrument to your DV other than through the effect of the instrument on your IV.

This is all very technical but of course it has to be.
posted by MisantropicPainforest at 8:33 AM on March 21