Why are Americans just eating up all of these “tax the top” policy proposals? Easily, because it’s “not us.” It’s a monetary transfer to anyone who isn’t being taxed.
Romer (1975), and later, Meltzer and Richards (1981) came up with an elegant model which demonstrates how societies with majoritarian democracy will continue to vote for more redistribution. Taxation is proportional, which means everyone pays the same percentage of their income, and the tax revenue is redistributed evenly (total tax revenue/population). In this scheme, the average (mean) earner is taxed and receives about the same amount, taking into account the inefficiency of taxation — otherwise he would vote for a tax rate of 100% .
As long as the amount an individual is taxed is less than the amount he would get back from redistribution, that voter will vote for more redistribution. Since in all real world distributions, the median income is less than the mean income, the median voter — political economy’s favorite son — will always vote for more redistribution.
Just to do a quick overview: Let voters in a society be uniformly distributed on the interval [0,1], where 0 is the farthest left one could be and 1 is the farthest right. If two policies are competing for votes, and voters vote for the policy closest to their own position, then the Nash equilibrium position for both policies is at 0.5, where the median voter is.
Why? Well, suppose a policy deviates from the median — it will surely lose votes.
The Romer/Meltzer-Richards model predicts that since the median voter prefers to get back more than he pays, he would vote for policies that increase the tax rate. Everyone earning less than the mean gets a positive transfer, and everyone earning more than the mean gets a negative transfer.
In the US, the difference in income between the mean and median household is about $7,000. For the median voter, voting for redistribution in the RMR model is like voting to increase taxes for your slightly wealthier neighbor, the guy you aspire to be in few years. Moreover, the median voter might not even be the median earner (and almost certainly isn’t — voter turnout drops dramatically as income decreases).
Well, this certainly wouldn’t result in more redistribution. The solution? Devise taxing schemes that only tax the top x% — these are not your neighbors, nor people you think you’ll ever be, and nowhere near the mean earner (at ~65th percentile) or median voter. This will certainly win over the median voter, wherever he or she turns out to be. In fact, Karl Marx had a similar idea, really.
From the foregoing, it is evident that Marx deduces the inevitability of the transformation of capitalist society into socialist society and wholly and exclusively from the economic law of the development of contemporary society. – Vladimir Lenin, 1914
Indeed, RMR corroborates in a very limited fashion Lenin’s sentiment, though Lenin himself seems to have interpreted “economic law of the development of contemporary society” to mean “bloody revolution.”
Marx believed society looked something like this, which was probably not far from the truth in feudal societies but not quite true in capitalist ones:
Since members of the proletariat are pretty much the same and basically have the same preferences, they will act as a bloc and are all equivalent to the median voter. The mean and above mean earners are all among the bourgeoisie, so tax, tax, tax away! (Rabble rabble rabble!) Likewise, when policies are created to tax the top 2%, the proposition creates two somewhat artificial classes: those who are in the top 2% and those who are not.
In reality, radical redistribution doesn’t happen in the US for a slew of mostly sociological and historical reasons. In the status quo of American culture, we don’t like to see these kinds of class divisions. Americans in general, and especially the more right-wing ones, tend to believe that we have a great deal of social mobility and tomorrow one could end up anywhere on the spectrum (for our Rawlsian friends, this a post for another day.) Along similar lines, Americans tend to see low income as a result of laziness and not bad luck, so relatively limited redistribution schemes are punishment/pity for the lazy (from George H. W. Bush’s inaugural speech, “[we should reach out and help] those who cannot free themselves of enslavement to whatever addiction — drugs, welfare, demoralization that rules the slums.”) For another game theoretic analysis of welfare, be sure to check out Jeff’s “Welfare for lazy bums” post.
 Also, the median voter doesn’t vote for any tax rate above the Laffer curve, since reducing the tax rate would increase revenue.
 An excellent overview: Alberto Alesina and Ed Glaeser’s lecture series “Fighting Poverty in the US and Europe”