Andrew Leigh points to recent work showing that preferences for redistribution correlate negatively with education, income, being male, and being white rather than black. The work pointed-to suggests that as folks move up the ladder (education, income, etc), they want to pull the ladder up behind them.
I'll suggest an alternative explanation. Each of the covariates mentioned correlate VERY STRONGLY with "economic thinking". Bryan Caplan's work (US sample) says that the gap in beliefs on positive economic matters between economists and the public is narrowed by education, being male, and income growth; my work (NZ sample) says that economic thinking on a more mixed set of positive and normative questions (though none relating to the desirability of redistribution) correlates positively with being male, having higher income, and having higher education.
What's going on then, in my account, is folks who are more likely to understand economics are more likely to understand the costs of redistribution in terms of economic growth foregone.