[sudosysgen]

Standardized national exams are normally very highly correlated to academic ability as represented by grades by the same teacher. Actually, where I live, this is essentially the explicit goal of the national exams - they then generate a score that normalizes the difficulty of each teacher and use it both for selection and to evaluate the accuracy of the exam and update it next year.

The data from which I learnt this isn't public AFAIK, but I'm sure similar studies exist for other countries.

[thaumasiotes]

The published literature doesn't support your claims, but that's OK because you have private data which you find more palatable? This is a fun combination with your posture of

> If you disagree, you should provide a source.

[sudosysgen]

Well, I did five minutes of googling, and found a public article on the subject : https://onlinelibrary.wiley.com/doi/pdf/10.1002/j.2333-8504....

>One index of that result was an augmentation in the correlation between the NELS Test and high school grade average from .62 to a multiple correlation of .90 based on the test plus 31 additional variables and corrections for unreliability and grading variations (see p. 72) [p. 105]

Here you go : test scores have a correlation of .9 with grades once you account for the grading difficulty of individual teachers and their unreliability (which is done by multiple national exams in the final synthetic grade). This means that they are excellent as a proxy for academic achievement, which can be restated as the following hypothetical : "what would be the grade of the students if they were to all be graded by the same, reliable teacher?"

Generally I think that the burden of proof on the statement "are exams better at predicting grades than IQ tests" lies on the negative, for the obvious reason that grades are generally composed of exams.

Also, the really cool thing with this is that you can also the reverse - after the teacher has had all of his students graded on the exam, you have all of the variables and can predict exam scores with a correlation of .9 (and thus general academic achievement with a correlation of .81), which means that you can avoid dispensing national exams at every level of schooling and simply use grades. You can even use this data later in college to evaluate how harshly a college professor is grading, and use that to generate much better grades in college!