PhD Preparation Classes
A Berkeley-centric list of classes to take. I didn't take all of these, so some of this is hearsay.
Math
Multivariate Calculus (Math 53)- of course, this class is required to apply to any top graduate school, but I put it on the list because of how important it is to take it seriously. It was tough when I took it here at UC Berkeley- competitive. The physics students, and people who had taken Math 54 (see below) seemed to have an advantage. I would advise any students who are at all weak in mathematics, either because of lack of recent experience or perceived ability (confidence), to clear their schedule when taking this class. Any math class can be overcome with enough time, but this may take more than others.
Linear Algebra/Differential Equations (Math 54)- again, foundational. Teaches you how to do math on a scale that matters; really, you are learning the basis of both computer science and pure math. It is powerful material that you aren't expected to master the first time you see, (but if you do, you'll save yourself a lot of time later...) Don't get scared off by 53 and 54 (400 person lectures at Berkeley, ick)- in my mind, these are the the last really hard "computational" math class in the curriculum. The upper division classes tend to be a lot more fun than this.
Analysis (Math 104)- the backdoor to the world of undergrad math! Ever wonder what's going on when the math prof goes off "proving" something? I can't say enough. Take this class. It will make everything else a lot easier, in large part, because it will make you smarter. If I had to identify the one class which best combined usefulness and enjoyability along the path I took, this was the one. (Some would say I am perverse.) Prepare well for, and work hard in this class. I sat through the first few weeks of a logic class in the philosophy department, and it helped.
Analysis II (Math 105)- this is the perfect terminal pure math class for an economist, (unless you are going to do theory.) It will teach you pretty much everything you need to know about measure theory and analysis in Rn, without the abstraction and time commitment of Math 202. With this background you will understand all the maths behind your first-year classes, and give you the bare bones background for math stat, probability and graduate applied math courses.
Linear Algebra (Math 110)- You could take this class before 104; the math website recommends it. I did not take it because several advisors privately told me that "an ambitious student can learn this material as needed." That being said, I've kicked myself for not taking this class more than any other math class, by far. Linear algebra is useful in almost every application of mathematics to economics; you might as well be good at it.
Numerical Analysis (Math 128)- This class reviews and expands on Math 53 and 54, (the lower-division multivariable calculus, linear algebra, and differential equation sequence), and in the process teaches you to use computers to solve or approximate the solution to the problems you see in these classes. It may not be the most exciting material (there are definitely some cool ideas), but I am mostly struck at how useful the class is. This class is great preparation for the sorts of the problems you see in graduate econometrics classes, too.
Differential Metric Geometry (Math 140)- A good class to take after 104, Analysis at Berkeley, especially if it is taught by Prof. Evans. (In Fall '08 lots of bad reviews overwhelmed the initial flood of praise for Prof Evans and this course. I am now thinking of taking this off my list.)
Mathematical Methods for Optimization (Math 170)- You will learn more about MATLAB in Math 128, which should be taken first. I liked Prof. Gu a lot, but if you can catch the class when it is taught by Prof. Evans you should definitely take it. (Though I haven't taken a class with Prof. Evans, he has been recommended more often and in more glowing terms than any other math prof at Berkeley.)
Other math classes: most math classes will have the level of rigor and offer training in logic that will help you in graduate economics programs. Unfortunately, most of the other math classes, while useful in this general way, will not get used in graduate economics programs, except in passing. For instance, Abstract Algebra, or Complex Analysis are great classes, but not really applicable in econ. Ask your adviser what classes are right for you. Some other classes with specific applications in economics: Combinatorics, Set Theory, ODEs, PDEs. If you are interested in economic theory, or econometric theory, you should go as far as you can in mathematics, and try to get in some graduate math classes. Standard graduate math classes for economists are Math 202A: Topology and Analysis, Stat 205/Math 218: Probability.
Statistics
Probability (Stat 134)- I don't see how you could try to take graduate econometrics without this class.
Statistics (Stat 135)- Much less mandatory, but still very useful.
Stochastic Processes (Stat 150)- I'm showing my biases here, but I really love this material. If you are an aspiring econometrician, economic theorist, financier, or model-builder (that's pretty much everyone), you will appreciate this class. Taking Math 104 (and 105!) beforehand will give you a deeper appreciation for the material.
Time Series (Stat 153)- pretty techie, at Berkeley. A lot more about curve fitting, engineering-style statistics, than the models and regressions you see in undergraduate economics classes. You can get pretty far from "social science" if you start down this road, but if you are interested in finance, macro, international, etc you should take it.
Other Statistics classes: Stat 133, Statistical Computing and 155, Game Theory. Unlike math, almost any statistics class will have some application in economics. These are just the ones I know of. I understand the Game Theory class is good; Linear Modeling would probably be overkill, considering its extensive coverage in econometrics.
Economics (yes, there are a FEW economics classes on this list.)
Microeconomics (Econ 100A or 101A)- It is far more likely you will be a microeconomist than macroeconomist. There are just a lot more firms and individuals than nations. Plus, with the density and spread of computers throughout the globe, we are seeing an explosion of datasets on the microeconomic level. It should be obvious that this is an important class to take. Of course at Berkeley you want to take 101A if you can, but it is not essential. If you can take 101A with David Card you should do everything in your power to do so.
Macroeconomics (Econ 100B or 101B)- Since the financial crisis, this class has become increasingly important. It is obvious that many of the theoretical insights that led to the monetary policy successes of the 1980s and 1990s were insufficient in the financial crisis of 2008. The aftermath of the crisis has radically altered the priorities of the whole field of economics, and it is no longer acceptable for a serious economist to ignore the macroeconomy.
Math for Economics (Econ 103, Math 103)- Take Analysis first. People have told me the class is actually pretty easy, but you don't understand anything if you don't know real analysis.
Advanced Microeconomic Theory (Econ 104)- Again, take Analysis first.
Econometrics (Econ 140 or 141)- Might as well get started. Economics today is largely about econometrics. This is the most useful thing they teach in the undergraduate curriculum, if you take to it. It is also the most important class to take for getting a job as a research assistant, or for doing your own research. It would be good to take Stat 134 before 141.
Applied Econometrics (Econ 142)- This class will tell you if you will enjoy graduate school in economics. (Okay, I'm exaggerating a little.) You should really take this class before you try running any regressions, since there are some important warnings they don't give you in the theory classes.
Other economics classes- well, there are a bunch of great undergraduate economics classes. It might surprise you then, that I don't go on and on about them here. Basically, I recommend you take classes in the fields you are interested in, with a slight bias for technical classes, like international trade, finance, and industrial organization, over "softer" fields like health, development or history. It is not the end of the world to take the soft classes, but I think you will find that you could learn the material on your own far easier in these classes than in some of more technical fields.
Graduate Classes in Economics (and other departments)
You should strongly consider taking graduate economics classes if you are applying. The warnings are that the first-year PhD curriculum can be very challenging, and is not at all like what economists do on a day-by-day basis, from what I have seen. That being said, getting an "A" (and letter of recommendation) in one of the first-year courses improves your application more than any other class. However, there is no expectation that you will take graduate courses, provided you do well in undergraduate courses, and you are taking a risk by taking one of these classes. Additionally, it is often impossible to take the first-year courses if you are not a graduate student. The best advice I can give you is to do well in your undergraduate courses, and try to find an advisor who is willing to recommend you to one of the first-year profs.
And that is it?
Well, not really. There are tons of classes that look really good when you are applying for graduate school in economics. Physics classes are at the top of this list- I am always surprised at how many great economists, in all fields of economics, credit physics as giving them their "edge." Also, the MATLAB class, Engineering 7 at Berkeley, would be helpful. Engineering students tend to do well in economics programs, as do computer scientists.
If you are reading this and thinking "what about history/poli sci/development studies/psychology/public policy/sociology..." These classes can look good, but they are neither necessary nor sufficient for admission to a top PhD program. A major in these fields will lead the application reader to ask "does this student have the technical skills required to succeed in an economics PhD program?" And you should ask yourself the same thing. A few classes in these areas can help, and knowing HOW to take one of these classes will teach you a lot about how to do interesting research; however, they will do nothing to make the first year of your program any easier, or make you stand out when applying. (Sorry, but at least you're not an English major!) That being said, if you can publish in one of these fields, go for it.
In undergraduate world, classes take students of many different abilities, a few prerequisites, and gives everyone a common basis for conversation. Even if you struggle, the classes are slow enough for you to work hard and make up any deficiencies. As far as I can tell, graduate school is not like this. In graduate school, the priority is exposing the student to as much as possible. Instead of taking students with diverse backgrounds and bringing them together, graduate school takes students with diverse backgrounds, and gives each the chance to grow into their unique abilities. Although there will be many chances to make up ground quickly once you get to graduate school, your opportunities will largely be limited only by those you have the time, energy, and training to take advantage of.
Getting technical classes out of the way makes a lot of sense, both in that it shows that you can handle the material in graduate economics programs, and it prepares you to jump right in. Who knows, you might even find out you love math, statistics, and/or mathematical economics.
Ask your advisor about classes you should take, and skills you will need. Working as a research assistant is another way to help you figure out where you want to invest your time.