Depends on what you're dealing with and the significance of it. 70 million is not a significant number of atoms of most things, for instance.

The important question is the second one I posed. What is the acceptable level of error in an approximation?

> Depends on what you're dealing with

We're dealing with people.

> and the significance of it.

People are generally pretty significant. 70 million people is very significant.

> 70 million is not a significant number of atoms of most things, for instance.

70 million people is a pretty significant number of people.

> The important question is the second one I posed. What is the acceptable level of error in an approximation?

Since you seem to have mistaken my rhetorical question for an actual question, I'll restate it as a statement: 70 million people is not an acceptable level of error in an approximation.

OK. Then what is an acceptable level of error in approximation when dealing with people? 7 million, at a scale of billions? 7? 1?

I think an acceptable level of error in approximation in legal and scientific terms would be one where the error is not likely to affect people negatively according to shared values (a shared value might be: people shouldn't get beaten to death or commit suicide--both problems among atypically sexed people). That number is difficult to come up with: it would have to look at impact analysis to see whether programs can effectively help people. And as technology, social programs, and medicine become better, the margin of error would likely go down. I can't come up with that number exactly, but given that we've seen significant benefits from social programs for atypically sexed people in the last few decades, I think it's pretty clear that we can still provide more benefit.

In other words, your answer to the question is to say it's a problem when an approximation is confused for the reality and policy formed accordingly.

Is that a correct assessment of your position?

> Is that a correct assessment of your position?

That's a reasonable approximation, yes. :P