[IPAC-List] differential validity?
Mark.Hammer at psc-cfp.gc.ca
Tue Mar 17 16:38:03 EDT 2009
I guess I didn't convey it right the first time.
1) An accomplishment
2) A qualification
3) A marker of socio-cognitive attributes
4) A marker of socio-historical factors (not quite as many folks in their 80's with Masters degrees even though they've had more time to get one than 30 year-olds)
5) A marker of cultural factors (compare post-secondary rates in Angola vs France)
6) A marker of access routes and all demographic and socio-economic factors leafing to access
If it was ONLY a marker of socio-cognitive attributes (and I'm one of those fools who thinks of intelligence as intrinsically social), and an accomplishment, then there should be no differential validity. The trouble is that education is all of those other things too, such that it CAN be telling a very different story about identifiable group A vs B.
Case in point: Some 20 years ago, working on my doctoral research in adult cognition, I sat back for a moment and realized that proper matching of my senior/retiree group and college group for education would involve much more than simply saying that the 20-30 year-olds had 18.3 yrs of education and the 60-80 year-olds had 18.7 (or whatever it was). After all, just exactly how many young women in the 1930's went to university and completed not one but two degrees? Heck, how many folks in general went that far? So, I figured I'd get some census data for the two birth cohorts, construct a regression slope, and calculate how similar or different each age group was from the other based on the residualized education scores. In other words, how much closer to the *typical attainment* for their birth cohort was group A than group B? If they both reflected their birth cohort equivalently, then they were "matched" on education.
Unfortunately, the analysis never came to fruition. The data provided me by Statistics Canada was simply not optimized for the analysis I wished to do (you would think they could have anticipated my needs in 1906! Harumph!!). Then there was the small matter of WWII during which data collection on men was poor because they were often somewhere else, and immediately after which the national average was altered by the huge influx of immigrants from Europe. But then that's pretty much the exception that illustrates the rule, isn't it? People *get* education and are *able* to get education for different reasons. I'm pretty confident that educational attainment within demographic group predicts the way you think it ought to, because of #3 above, but I am equally confident that the same attainment can easily mean something different at time A or for group A than it does at time B or for group B.
More information about the IPAC-List