..for some commonly agreed-upon, objective, quantitative, non-discriminatory intelligence criterion.
Before you rush to close this as a duplicate, yes, I am aware of this other question that treats this problem. However, the accepted answer fails to take into account other factors such as education level and cultural background, to name a few. I'm not interested in lumped "statistics" that compares apples to oranges, I'd rather see a scientific study that:
Uses a sufficiently large sample size to compensate for Law of Large Numbers. The sample may or may not be a representative of the population in any given country/region, it could be biased towards certain society groups (e.g. university students or kids) or genders, but that is fine, as long as it is able to demonstrate a strong correlation (or lack thereof) between the factors for that specific sample. If and how these findings can be projected to a larger population group is another question entirely.
Performs the study on a group with similar cultural and educational backgrounds, for example among students in an university, or even a whole country. Country-to-country comparison is mostly useless to draw any sort of meaningful conclusions due to the lack of a common objective performance criterion.
- Uses a quantitative performance criteria. Standard IQ tests qualify, but any other sufficiently objective criterion is okay, for example grade(s) received on a course or the overall grade achieved in an educational programme, results from standard tests or state maturity exams, etc.
It probably goes without saying, but "blondes" are defined here as "natural blondes" as opposed to dyed blond wannabes. It's another question entirely whether dying one's hair blond (or dark, for that matter) has any effect on intelligence.
Bonus points for studies performed on a single-gender group which avoids gender bias.