It's not clear from the article if those percentages include all of the society or just those that started out poor. If it's the former, it might be just correlation, not causation.
I'm pretty sure it's the former. I don't find it plausible that these three things wouldn't have a causal effect on income, but it's a fair point that it would be more useful to see those percentages broken down by starting income level.