Mankiw discusses why it's ok for him to advocate a gas tax increase even though it's surely politically infeasable. He quote Friedman as saying that normative economics should be analyzed separately from politics.
That to me makes no sense. Normative economics by its very definition is subjective. The so called "analysis" in normative economics stems from assumptions that in many cases may be largely informed by ones political beliefs. If I say that we should increase taxes because the government will do better with that revenue that individuals, then that is my economic opinion, perhaps based on an economic assumption of large market failures. However, another analyst might say, we should NOT increase taxes because taxes are inefficient - and he would be basing his analysis on the idea that markets are usually pretty efficient and governments are large bureacraies (in that particular situation).
So, you see, normative economics is dependent on political bent - small vs. large government, the belief in redistribution vs. efficiency.... Normative economics, in a word is a sham - and is a shame to the science. It is the single largest contribution made by Friedman that should NEVER have been made. It turned economics into a slumgullion of pseudeo-science mixed with politics. At a time when econmics could have incorporated other sciences like psychology and sociology, Miton turned to politics and turned the jobs of economists from informers to ideological idea men. Normative economics is fine in the sense that everyone needs to have a framework to make opinions - the danger is that normative economics has become a 'field' of study and has been artificially elevated and propagated as a part of science. That can become dangerous if policy makers feel influenced soley or almost entirely by their economic advisors.
When Mankiw says that he now can point to Friedman and say, 'see, it's not the job of the economist to take into account whether or not the government will cut other taxes if they raise the gas tax, he could not be further from the truth. Point to Friedman in this regard, and you only prove the conservative's point. And this is the ironic tragedy of normative economics. Normative economics says study a subject from a narrow lens of economics and advocate a policy. The problem is that economics is a study of behavior. So why stop at analyzing the effect of a gas tax hike? Why not analyze whether or not the government should cut income taxes? Why not take into account the probability that the government would cut other taxes? What makes that any lesser of a question than the direct effect of a gas tax hike? It's all about behavior - why ignore the likely behavior of the government? And so you see this is the slippery slope of normative economics - it's not a science - it's a study of whatever, based on whatever - guided by a few key economic assumptions. It's hard to say what people "should" do from a so-called scientific perspective when the intricicies of the real world make analysis so difficult. The best economics can and should do is positively study behavior and incorporate other fields of studies in order to try and explain and predict that behavior. Leave the "coulds" and "shoulds" to politics entirely.