Greg Mankiw goes into a number of reasons a prospective PhD student in Economics needs to take a boat-load of math class:
4. Your math courses are one long IQ test. We use math courses to figure out who is really smart.
5. Economics graduate programs are more oriented to training students for academic research than for policy jobs. Although many econ PhDs go on to policy work, all of us teaching in graduate programs are, by definition, academics. Some academics take a few years off to experience the policy world, as I did not long ago, but many academics have no idea what that world is like. When we enter the classroom, we teach what we know. (I am not claiming this is optimal, just reality.) So math plays a larger role in graduate classes than it does in many jobs that PhD economists hold.
Is it possible that admissions committees for econ PhD programs are excessively fond of mathematics on student transcripts? Perhaps. That is something I might argue with my colleagues about if I were ever put on the admissions committee. But a student cannot change that. The fact is, if you are thinking about a PhD program in economics, you are advised to take math courses until it hurts.
But, he forgot one: because math is fun!