[alpos]

In case you were looking for the literal answer, it is because a society is the aggregate of every individual in it. It follows that a society's preferences are defined by the aggregate of the preferences of each individual in it.

Therefore if one wants to know what the society's preferences are, that data must be collected in such a way the each individual's preferences are included in the data set.

We might still apply a weighting function at aggregation time, depending on why we're looking at society's preferences in that instance. There are certain group's of opinions that we don't value as much for answers to some questions, but we should like to be deliberate about those choices instead of applying an implicit weighting function during collection.

[adventured]

I worked at a chain retail store once upon a time. There was a self-proclaimed nazi (he was proud of it of course) that came in daily and stole a lot of merchandise. Company policy was to do nothing, essentially.

The nazi also had a criminal record for raping a young boy.

The parent put it this way:

> If a person is disabled or uneducated and can’t hold a job, that person should still get an equal say in society’s preferences even though they have no money with which to express their preferences

Morally and rationally the nazi child rapist should not have equal say in society's preferences. Any society - any culture - that is stupid enough to place that person's preferences at a level of equality, deserves and will have earned its inevitable collapse. And there are far worse monsters roaming about than the nazi child rapist. While my specific example is an outlier (on purpose), the premise is far more expansive in principle.

It's also why a Constitutional Republic is vastly superior to actual Democracy.

[alpos]

Collection of data is not equivalent to a judgement on the value of the results.

For example, when considering questions of crime and punishment / deterrence we probably wouldn't value the preferences of say, psychopathic murderers, on the same level as those of the families of their victims. That would be silly.

However, it would still be good to collect their preferences on conditions within the prison system and host of other topics. And we might still intentionally devalue those preferences too, but that would be a deliberate choice of values as society.

When we fail to collect the data effectively in the first place, we deprive ourselves of the opportunity to determine what we value and what we don't. So we're still going to want to know these things, that way we don't just discover later that we were ignoring a big thing that, turns out, we would have liked to value once we knew about it.

In your particular example, add in the one stipulation that both the Nazi and the family of his victims don't have much discretionary income to spend and yet they live in a world where society's preferences are collected primarily or singularly by way of observing what people spend money on. In that scenario, all of those people's preferences share the equal status of being completely ignored. That seems unfair to victims as well.

Modify the scenario a little differently and say the Nazi is very wealthy while the victims are poor. So then because of the data collection method alone, as a society, we would then be elevating the Nazi's preferences to 100% greater value than the victims.

That outcome seems vastly contrary to what we would actually value if we just fairly and accurately collected the data first and then applied our judgements about what should be valued.

Further notice that in just asking the questions, we can also easily discover at the macro level that this guy is a Nazi, after all, given the chance, he'll tell us so. This means deliberately ignoring his preferences is easy at time of aggregation and now we can do that on purpose, because as a society, we don't value the opinions of Nazis when it comes to questions of how to run things in a democracy.