Thursday, June 9, 2011

Why bother with IQ anyway?

A great deal of ink is spilled in the reactionary sphere and the HBD sphere regarding IQ. One might ask why is this so? Is everyone still obsessed with the almighty SAT that they took in their late teens?

The answer to this question is that it's probably the most useful metric compared to the work necessary to obtain it in the social sciences. In the social sciences, you very rarely get correlation coefficients anywhere near that of the physical sciences. People are far messier than even organic chemicals. To get good predictive results out of correlations at the level found in the social sciences, you've got to be looking at large groups of people, at least if you're spoiled by the precision of harder sciences.

There are other metrics that are almost as useful, like executive capacity and conscientiousness, but the big problem with them is that they're generally awfully easy to game. Whenever there are significant stakes involved, the participants WILL try to game whatever test or screening/interviewing procedure you establish. I've been on both sides of such processes, although I've long since interviewed more people than have interviewed me. Every candidate will attempt to 'put his best foot forward'. Your fancy questions intended to reveal the candidate's true nature are discussed in detail in interviewing classes and books. Such is life. IQ at least has the merit that it's damned difficult to game.

There are two major types of IQ tests out there. The first type tests things that are commonly done and studied in the culture of interest. For instance, the old SAT is a good example of this, testing things that have a wide cultural relevance for prospective college students---no math beyond algebra II, for instance. You use the test to rank order the participants and assign scores based on the percentile of each test taker. In practice this corresponds pretty well with what we informally call intelligence in everyday speech. The test is of a form familiar to most of its takers and is difficult to prepare for specifically---even the test prep industry can't normally promise more than 50 points or so. Tests of this form have been affected much less by the Flynn effect, and in some cases, not at all. Sometimes these tests are referred to as tests of 'crystallized' intelligence.

The second sort of IQ test attempts to be 'culture fair' and thus tries to test things that nobody's culture does much of---ideally it's a total novelty, like your first encounter with a Sudoku puzzle. Lots of things like fill in the next picture in this sequence, patterns in matrices or mathematics, and the like show up on tests like these. Such tests are often called tests of 'fluid' intelligence and have been heavily affected by the Flynn effect. I'll advance my explanation of why:

These types of tests USED to be much more of a novelty than they are now to their takers. Look at the toys SWPLs give their children these days. Look also at the sorts of puzzles that crop up in ordinary newspapers and magazines now compared to in previous generations. They retain their correlation with tests of the other sort only because they are continually renormed. Their content, indeed, their type of content has become progressively less alien and more embedded into our culture as a whole. This observation is also consistent with the observation that the Flynn effect has largely stopped in most 1st world nations. This is my explanation of my observation that when I meet a healthy senior citizen described as bright, or very bright back when he was young (and often, in fact usually, tested by some military apparatus to stick a number to that ordinal description), he generally strikes me and others around at roughly the same ordinal level as he was described in his younger years. One could do a study if one had the military IQ test score results from a large cohort from say, 40 years ago with 55-60 year olds today. Would you see their results depressed by 40 years of Flynn effect? I seriously doubt it.

This isn't to say that improved nutrition (although I'll grant, a mixed bag here, but outright nutritional deficiencies on things like iodine, iron, and folate have ceased to be going concerns for most people) and reduction of lead in the environment haven't had some impact---they have, and a small fraction of the Flynn effect shows up on 'crystallized' tests of intelligence. But relatively speaking compared to the artifact above, I think they're small fry.

5 comments:

RS said...

I didn't know the Flynn effect was not the same for fluid and crystal.

The main problem you have in considering the FE unreal is dysgenesis. It seems that considering both differential fertility and differential generation time, you should probably observe a pretty serious decrement by now. But perhaps the 'real' fraction of the Flynn effect is just about enough to keep things on a pretty even keel (up to 1980, anyway).

Contrariwise, in light of the enormous correlation of IQ and GDP/capita across nations - even though not nearly all of the correlation is actually caused by IQ - it is somewhat problematic to consider the Flynn effect mostly real. Because growth rate of GDP/capita did not change much in the West round 1980, when the Flynn effect stopped. Long term growth in the US going all the way back to ~1810 is 2%, and it only deviates much in the Depression and the Civil War:

http://seekingalpha.com/article/224600-2-real-per-capita-gdp-growth-a-long-term-trend-we-can-count-on

(That post ends amusingly: "There's nothing happening right now that will change that long-term [2%] trend, and for that we can be thankful.")

Jehu said...

Yes, you see the Flynn effect strongly on things like Progressive Matrices but pretty much not at all on things like the SAT, which was renormed in the other direction (renormed retroactively, I might add, something you might notice if you have a copy of your transcript---you'll note that the score will have been revised upward). There's an experiment which could be performed to falsify this hypothesis that I've alluded to in the post (retest the same people @30-50 years later, removing from consideration anyone suffering more than normally from aging/senility).

RS said...

There is a bit of a problem for you in that Lynn claims the Flynn effect already is almost completely manifest in young children of, I don't know, something like five or six. That is Lynn's chief or sole reason for adhering to nutrition and infection as the cause of the Flynn effect, which he takes to be real. Presumably these facts refer to crystalline tests only.

Jehu said...

Not sure which tests Lynn has given, but most tests for IQ that little kids get are fluid g, partly because they've not had time to crystalize a whole lot and even 6 months to a year spans of age make such a big difference at that age. So my guess is most of those tests are the fluid type with shapes, patterns, and matrices.

RS said...

I'm not sure why I said "Presumably these facts refer to crystalline tests only."

I may have meant fluid not crystal.