It isn't wrong according to the definition I've found,
No, it's not. The definition you found is correct. When two variables are directly correlated, an equation can be developed to express this correlation as a direct proportional relationship between the two variables. The procedure is called linear regression, and is described in the following link.
stat.yale.edu/Courses/1997-98/101/linreg.htm
Before attempting to fit a linear model to observed data, a modeler should first determine whether or not there is a relationship between the variables of interest. This does not necessarily imply that one variable causes the other (for example, higher SAT scores do not cause higher college grades), but that there is some significant association between the two variables.It would also mean that every unit increase in intelligence corresponds with a unit increase in education. And the research doesn't show this, and I don't believe you really believe this--every year of school adds the same amount to intelligence as the previous year? Really?
Yes it does, that's what correlation implies. This is how it works, you use the data to find the correlation coefficient. You check if the coefficient is significant and if its square is high enough to proceed to modeling. The model yields a linear equation, where one variable is made to be directly proportional with the other variable (y=mx + b, where m and b are the constants derived from the regression analysis). If this model is strong enough, it would be possible to reasonably determine how many units of education is increased by every unit of intelligence found in the sample.
One can statistically predict a higher intelligence level when selecting from a pool of Ph.D. holders than a pool of BAs, sure.
That's precisely what I am talking about.
Using this to try to predict the intelligence of a particular individual based purely on a knowledge of their educational attainment represents a gross misuse of statistics...which I'm sure you know.
No, it's not. Doctors use this logic all the time when constructing screening tools. Your risk to acquire certain diseases can be predicted accurately by different factors (age, gender, etc.). In developing a screening instrument to say, replace/supplement an invasive procedure (such as a biopsy), what we do is we try to find which factors are the best predictors (variables from blood chem tests, urinalysis, etc.), and then use regression statistics to assign scores to them. In this analysis, we are actually able to estimate how large each point of a certain variable adds to your risk of getting the disease.
Do employers use this same logic in determining how smart applicants are? Why should they when they have an HR department equipped with instruments that can measure IQ? An exact measurement is always better than a predicted one. Still, the difference between the salaries of undergraduates and graduates in the United States implies much about how employers value higher degrees.
bls.gov/emp/ep_chart_001.htm