> The best way to assess the effects of alcohol would be to randomly assign people to drink it or not in childhood and then monitor their health and drinking over the rest of their lives. Since such studies cannot be done, researchers instead have to ask people about their drinking habits and follow them over much shorter periods of time.
Studying humans properly is nigh-on impossible.
Say you do the setup from the quote, then one group would also be in a special situation where researchers told them to do this thing, and they would tell their friends, or maybe they would keep it secret, or maybe it would make them a bit weird, or maybe it would pave the way for more addictions, or make them more conscious of it, and they would have another item on their usual shopping list, and they may meet their spouse in a bar, and so on — all of which influences outcome.
a) you break a number of laws and ethical rules
and b) you have a population of hundreds of billions of humans to study. Humans are very different from one another in important but subtle ways.
This is why I find the proclamation that "correlation != causation" so annoying. Of course it doesn't, but we cannot make a causative study of humans!
See Nazi twin studies, Japanese Unit 731, and who knows what is going on right now which will take years to be uncovered..