To raise a given amount of revenue, is it always better to reduce or eliminate deductions than to raise the top marginal rate?
In a word, no.
Assuming the deductions are undesirable subsidies (such as the home mortgage interest deduction, which I certainly would so characterize), rather than, say, costs of earning income, reducing or eliminating them is likely to be preferable to raising the top marginal rate from the standpoint of efficiency.
But tax policy involves a tradeoff between efficiency and distributional concerns. Otherwise, we would simply charge a "lump sum" tax that taxpayers' decisions could not alter - for example, a uniform head tax.
Suppose we are trying to raise a fixed amount of revenue from the top 1 percent of the income distribution. Imposing some sort of a deduction cap is likely to be better in efficiency terms than raising the top rate. But it is also likely to have a less progressive incidence - to hit the people at the very top somewhat less than raising the marginal rate. This reflects that their previously allowable deductions are likely to decline as a percentage of income as their income rises. For example, if your salary goes up from $10 million to $100 million, you probably won't buy a house that is ten times larger (and has a mortgage that is ten times larger). And even if you did, under present law the home mortgage interest deductions would be limited to those arising on $1.1 million of home mortgage loan principal.
I am certainly glad to see that cutting preferential deductions that used to seem politically sacrosanct is now on the table politically. Perhaps a good policy change will result from this. (And I do like the idea from the Romney campaign of capping specified deductions and inclusions, so long as we don't attach fantasy revenue projections to doing so.) But that does not mean that the top rate shouldn't also go up, as a way of reaching the very top of the top more than we would otherwise.