I like this idea. When we have to stretch too far to look for an explanation of a trait based only on that trait's effect on differential reproduction, it may be because there is no such explanation. Plenty of traits are the result of side effects that did not affect reproduction, and others may be cultural.

This idea has just what we need: it fits the experience, doesn't seem to affect reproduction, and is a side effect of sexually selected traits. When you add in a cultural component that may amplify or suppress this feeling of sympathy, you have what looks like a good explanation with no "just so"s necessary.

Children's social worlds aren't as large as adults', so one prediction this model makes is that children raised in small social worlds (homeschooling or other small communities) should have much less of an underdog bias than adults or children who interact with many strangers.

Intuitively, I'd say that's probably not the case; but it bears testing.

I like this idea too. One prediction from it seems to be that those who feel less like underdogs (such as a Saudi Prince) will support underdogs less. One might find those who feel less like underdogs viageneral socieconomic status too, but since we have a fairly egalitarian society high income people might actually be more likely to have considered themselves an underdog during their formative years.