I am generally in agreement with what Miles and Noah have to say, but would add or alter a few bits:
1. By all means do not just focus on the top five or 10 or 15 programs. The poster child here is probably Amitabh Chandra (now at the Kennedy School), whose doctorate is from Kentucky. On my very first visit to Kentucky back in my assistant professor days, I was assigned to meet with Amitabh and told to talk him out of staying at Kentucky for his doctorate. I failed, but his career seems to have turned out fine anyway. The reason it turned out fine is that the faculty at Kentucky, who already knew Amitabh as a stellar undergraduate, treated him like a colleague throughout his doctoral studies. He got a lot more attention and opportunity than he would have at a top program. Example number two is my colleague Martha Bailey, whose doctorate is from Vanderbilt. What Amitabh and Martha have in common is a lot of internal drive, which you need to make this strategy work because there is less external pressure from peers and faculty outside the top departments.
2. You should take more than just one statistics course. If your college offers an upper econometrics track for undergraduates (many do, and Michigan will soon) take the whole thing. If an upper track is not available, be sure you do what you can and think about taking some courses in the statistics department as well.
3. Learn to program if you do not already know. Pretty much any reasonable programming language will do. Once you have learned one, others (including statistical packages like Stata or Matlab) are much easier to learn.
4. I think there can be more value in doing an MA first than Miles and Noah. This is particularly true if your undergraduate record is a bit weak and/or if you are unsure you really want to do a doctorate. The trick is then picking the correct MA program. Many are aimed at mid-career people adding a credential and not at people thinking about a doctorate. I recommend in particular the programs at UBC and Toronto. They have the added bonus of plugging you into a somewhat different network and letting you experience life in another country (assuming you are not Canadian).
5. It is harder to get someone to hire you as a research assistant, even at a zero money wage, than Miles and Noah suggest. The time cost to the professor is really large of having a research assistant. Paying that time cost for someone who turns out not to produce - it happens! - is something faculty really try hard to avoid. So if you want to do this, it is probably best to first make a good impression in a class, or in someone else's class who is willing to write an email of introduction for you to the person you want to work for.
6. Think about taking, or at least auditing, first-year graduate courses at your undergraduate institution. I did this at Washington, taking Gene Silberberg's excellent first quarter of graduate micro.
7. Getting a doctorate at a biz school with an economics group is at least as good as a straight-up economics program. It is easier to get in and you will likely have more financial aid and a nicer place to work. The same holds for some policy school doctoral programs.
Who was my favorite student this term?
3 years ago