There’s this idea called “domain specificity” that I learned from my sister many years ago: it’s the idea that a skill might work in some domains but not others. For example, knowing (some) Spanish helps a bit with my French but it doesn’t do anything for my Chinese. Traditionally in psychometrics, this sort of thing is studied using correlation–scores on various math tests are highly correlated with each other, and these are also correlated (but less so) with scores on English tests. Michael Jordan would probably be good at just about any sport he tried (even if he couldn’t quite hit the curveball). Top actors can typically sing, dance, do magic, etc.–they’re great all-around performers. And so on.
I was reminded of domain specificity after reading a blog by Felix Salmon arguing that Robert Rubin and Lawrence Summers made big mistakes when they served as Treasury Secretaries during the Clinton administration. Salmon writes:
The connection here is a direct one: the single biggest reason that Citi needed its US government bailout was its misadventures in the derivatives markets which Rubin played such a central role in deregulating, and which he then sought to exploit when he was at Citi.
Rubin was a huge cheerleader for Tommy Maheras, the fixed-income king at Citigroup; the two of them, together, encouraged Citi’s mortgage desk to take on ever-greater risks and to stuff Citi’s balance sheet full of the nuclear waste that soon became notorious as “liquidity puts” and “super-senior CDOs”. What both of those things have in common, of course, is that they were derivatives against which Citi didn’t need to put up any capital.
I have no comment whatsoever on the substance of this. Salmon seems persuasive to me, but I know nothing–nothing–about any of this stuff. (I do remember that, about 10 or 15 years ago, a former student told me he was working on Wall Street on “mortgage-backed securities,” and I was like, Huh?)
What interests me here is the connection to domain-specificity. Larry Summers is famous for being really, really smart in thinking about economics, and Robert Rubin is famous for becoming really, really rich on Wall Street. The implication has often been that success in either of these domains should transfer to economic policymaking.
I don’t really know how to think about this question. To start with, there are questions that go beyond expertise. There are conflict of interest issues–Rubin and Summers are both notorious for these, but perhaps conflict of interest is inevitable in the world of moneymaking–and also differences in priorities. For example, people like Rubin and Summers are probably more concerned about the congestion problems faced by helicopters and private jets than by long waiting times for city buses.
But let’s set aside such concerns and assume that Rubin and Summers, acting as policymakers, have the same general goals as 90% of Americans and want economic growth, low unemployment, low inflation, and all the other good things that go with prosperity. To what extent does their track record as rich smarties provide evidence that they’ll be good at the job?
It’s hard for me to imagine direct research in this area, so I’ll extrapolate from my own experience. I think that my expertise in data analysis in medical imaging, political science, and environmental science is transferrable to other fields. I can read an article or go to a talk on economics, public health. sociology, etc., and make useful comments. On the other hand, I can’t design an experiment to save my life. Also, I’ve spent a lot of time teaching, have written two textbooks and even a book on teaching methods–but I’m still not much of a teacher. And a few years back when I was the director of graduate studies in our department, I didn’t do a particularly good job of keeping our Ph.D. program organized.
To get closer to the Summers/Rubin thing: Suppose someone asked me to be head of the Bureau of Labor Statistics or the Census Bureau or maybe some smaller statistical agency. I don’t think I’d do a particularly good job. I think I’d be good as some sort of science advisor–someone with the job of judging the scientific/statistical evidence on important issues and making recommendations–but I don’t think my expertise would extend well to the domain of heading an agency. Conversely, I know some very smart statisticians who are, I imagine, as brilliant as Summers (not that I really know, having never met Summers) whom I’d never trust near an actual live social science data analysis. It’s just not what they do. I don’t know enough about Rubin’s and Summers’s areas of expertise, or about the duties of the Treasury Secretary, to have a sense of where they would fit in this picture.
I think what you really want in a lot of these jobs is someone who knows when to listen to whom, and a lot of whoms, with a variety of different kinds of expertise, surrounding them.
Based on my impression of my classmates in a intensive intro summer course, I think your Chinese is probably better for knowing some other European language than it would be otherwise. That your Spanish (such as it is) helps your French more than your Chinese seems reasonable, but might depend on definitions of "help." Presumably the Defense Language Institute has plenty of data that could nail this down. (Doubt they'd give it to me, but they might give it to you.)
In my experience when I've asked, adult non-heritage Chinese language learners who have very good [tone] pronunciation (N ~ 5) have turned out to have had some music training. I think this allows them to listen to themselves critically and correct their sound production, and not because Chinese is a particularly "musical" language.
Here are some numbers from DLI indicating how hard to learn Chinese is compared to other languages. Presumably those data are not conditioned on previous language experience.
On a hobbyist scale, this is related to Engineer's Disease, where someone with a great deal of expertise in one field – say, engineering – assumes this also makes them an expert in some other field – say, theoretical physics – and they begin holding forth on their crackpot theories or patenting nonsensical devices. It's painful to watch, because the victim usually is genuinely good at their profession but comes off looking like a lunatic.
I'd suggest that you're talking about the difference between intelligence and <a>wisdom, but you'd probably just think I was joking.
(P.S. I'm kidding. But not joking.)
Of the last two Clinton Treasury secretaries, Citibank made Rubin their boss, and then Harvard made Larry Summers President. They were both pretty bad at their new jobs.
I think Citi and Harvard were thinking that their new bosses were part of the Inner Party, the clued-in insiders who really know what's going on, like O'Brien in 1984 who explains everything to Winston Smith.
The scary thing, as Greg Cochran has pointed out to me, is that in reality, there is no Inner Party. Nobody knows what's going on. Dick Cheney, for example, acted like he surely belonged to the Inner Party, but he proved clueless about Iraq.
Let's throw some made-up numbers at the problem. Say that you take the most all-around competent 1% of the adult population. That's 2 million people.
Then you take the 10% of the population that has the overweening confidence to think they just plain deserve to be a Treasury Secretary or Vice President or whatever. That's still 200,000 people..
How do you empirically find the best person out of those 200,000?
Well, you can't. Life isn't long enough to run enough real world tests to distinguish in a valid manner among that many people.
So, you end up appointing people based on X number of people you've heard of or people you've heard of based on their reputations that are based on their doing okay in Y number of situations. But X and Y are never big numbers.
So, you end up appointing somebody with a big ego who has mostly been lucky so far in his career. And then his luck runs out …
I think expertise in formulating and using mathematical models does generalize, more than most kinds of domain knowledge. But one of the best qualities of an experienced modeler is humility about the generality and realism of models.
"Expertise" in economics should only be trusted so far. The honest answer to many big economic questions is "we don't know". Egos being what they are, though, both sides of every policy debate have Nobel prize winners (and billionaires) eager to provide rationalizations.
My experience has been that the more specifically you define the domain, the task, and measurable attributes that correlate with task performance, the more interesting are the things you can say.
As an example of domain specificity, in manufacturing and production environments, generally speaking adding more resources to a task gets it done more quickly, whereas in design / engineering / knowledge environments, adding people to a team may *increase* the time to complete a task because of communication complexity. So transferring a great production manager to a research and development position may be counterproductive if the task is running a single project, but might be fine if the task is the overall productivity and organization of a group doing many projects (at which point the lab starts to look more like a factory again).
As an example of measurable attributes, somebody who has a relatively large short term memory stack is likely to be better than average at diagramming sentences in linguistics, but they might have an similar advantage in reading long legal clauses or memorizing lines from a play.
On a distantly related note, I believe this Australian psychologist (http://philpapers.org/autosense.pl?searchStr=Graeme%20S.%20Halford) has interesting things to say about a general capacity for inference (i.e. constructing and understanding models with multiple components)