I would take issue on the size of the population.
It's not the size of the population that's the issue, but the number of accidents. Fortunately, fatal or serious aviation accidents are rare enough that it can take decades to accumulate sufficient statistics to provide the safety case you seem to seek.
Imagine there's a rate of 10 fatal accidents per year. Is a safety measure that cuts the underlying rate by 10% worthwhile? It would save a few lives each year, so a regulator should probably consider it.
But how do we prove the effect of such a safety measure? Imagine we can compare 10 years with the measure (let's say there are 100 accidents) against 10 years without it (let's say there are 90 accidents). Is that a statistically significant reduction? Not even close. At those rates, you'd need something like 40 years with and without the measure to see a statistically significant difference. And that assumes a perfect control, i.e. that nothing else has changed in those 40 years to skew the results.
My impression is that there are very few accidents involving either instrument rated or IMC-rated pilots that can be clearly attributed to IF aptitude -- a lot fewer than 10 per year. That means that, unless we are prepared to write off lost lives simply as data, regulation has to be built to some extent on plausibility of cause and effect as well as statistics.