Do we really know much about host-country effects of refugee immigration?

This recent wave of media hype on Tamil, Burmese and Afghan asylum seekers has stimulated some fairly spirited discussion lately. I fear though, much of the commentary involves people speaking beyond their competencies. This raises a question: in a debate in which most commentators have put "feelings in place of ideas", how much do we really know about the host-country effects of settling refugees?

I will ignore the (non) question of whether or not Australia should accept refugees, and focus on the discussion of what kind we should be allowing.

Harry Clarke disagrees with a few of his readers. They argue lines like:

Boat people have demonstrated their motivation and enterprise and are therefore differentially likely to make good citizens.

Against

Encouraging queue jumpers on the grounds that they show entrepreneurial zeal and enterprise would be about the most daft policy I could imagine.

Meanwhile, over at Core Economics, Josh Gans argues that refugee immigration is a case of revealed preference over institutions

When we do not have free migration, we stop people being able to match different rules to their preferences. Personally, if someone in Sri Lanka prefers our rules, why should they be forced into convincing everyone else around them to adopt them rather than just moving?

The problem with both sides of this argument is that there seems to have been a lack of objective research into the relative successes of a cross-sectional sample of refugees, rather than a cross-sectional sample of immigrants. We do not know which of our Afghans, our Vietnamese, our Tamils, have done better than the others---the boat-people? former store-owners looking for good institutions? sign-up-and-wait rule abiders?

With this gap in knowledge, it is simply impossible to develop policy that minimises the loss/maximises the gain of refugee immigration in Australia.