Tips for Empirical Newbies

When I started teaching four years ago, the advice for junior scholars on empirical research was pretty clear: Don’t do it unless you have graduate training in statistics or a related field. Since I shied away from math until becoming a professor, this advice should have led me far away from empirical projects. But a few years ago, I decided to take on a fairly simple empirical project, and I have spent the last few years trying to get up the courage to take on a more complicated project. There has been plenty of trial and error, and there are certainly things about my original project that I would change now. I have nonetheless learned a few things that might be helpful to other empirical newbies:

Start Small. I respect the people who do fancy modeling and cutting-edge statistical tests, but not all empirical scholarship needs to look like that. Some of my favorite empirical work (like those here and here) takes a more descriptive approach, uncovering an interesting corner of the legal world and describing it. There certainly can be pitfalls to this type of work, but if you do it right, you can make a real contribution to legal scholarship while staying within your comfort zone.

Go to the Workshops on Conducting Empirical Legal Scholarship put on by Lee Epstein and Andrew Martin. I have been to both the introductory and advanced conferences, and I found them both to be excellent. I think the advanced workshop is essential if you want to do your own research. Both conferences could also be helpful if you have no intention of doing your own empirical scholarship, but simply want to be a more informed reader of other people’s scholarship. I know other schools are starting to offer similar programs, so it may be even easier to attend these programs in the future.

Take statistics courses at your own institution. People often say that they couldn’t possibly find the time to take a statistics course, especially if they are a junior professor and are still prepping their own courses and getting their publishing legs under them. But I am not sure it is actually that hard. One of the great things about already having a teaching job is that we don’t have to impress anyone in these courses. If you audit the course, you can do only the work that complements your own goals, which may mean skipping the tests, group projects, etc. I started taking statistic courses at our business school two years ago. I took two introductory courses and then started taking more advanced Ph.D level courses last year at a nearby university. I probably spend 4-5 hours per week on these courses, which is a significant chunk of time but definitely doable. I do think it is key to take courses that start in the early morning or late afternoon so that they don’t cut into your day too much.

Regularly read empirical journals. Our library routes new issues of the Journal of Empirical Legal Studies to me. It is helpful to see what tests other people are using and try to figure out why. It is also valuable to see how people describe their methodology and their findings.

Learn a statistical program. Preparing to do your first empirical project is partly about learning the underlying statistics and partly about learning the relevant statistical program. I haven’t taken my own advice here yet, but I do think it would be helpful to attend a 2-3 day training program on STATA or SPSS. I have stumbled along using STATA (after a very ill-fated attempt to learn SAS), but I am sure that I could make things easier for myself if I took the time to really learn the ins and outs of the program.

Anyone else have any tips?

You may also like...

5 Responses

  1. Dave Hoffman says:

    This is great advice. I’d also recommend:

    1) that folks considering co-authorship generally prefer political scientists and economists to statistics professors;

    2) that you do coding using Access or through a web collection form rather than inputting data directly into excel;

    3) that you focus on what you have a comparative advantage at — i.e., bringing legal situation sense to bear.

  2. Joe Doherty says:

    A great topic. My 4 cents:

    1) Replicate. When you read an interesting article, contact the author(s) and ask for the data. Try to re-create the tables. You will learn a lot about how the analyses were conducted and what choices were made in reporting the results. Most importantly, you will get ideas about how to extend the research for a paper of your own.

    2) Find existing data. Familiarize yourself with the various data archives. ICPSR is fantastic, but some of the longitudinal datasets are no longer updated; the updates are available on project websites, instead. If your campus has a data archivist, pay a visit. If commonly used data seems unmanageable (the Current Population Survey), you are probably not alone. Look around, someone else might have solved the problem for you and posted scripts online.

    3) Get certified with your campus IRB. It takes 2 hours and you may never need it, but you’ll learn much about research ethics.

    4) Hire a good graduate student. Social science graduate students will, for a fee, run or help you run your data. Since you are in the law school, be very clear (if asked by administrators) that this is not a mentor/apprentice relationship but an exchange of money for expertise. At the very least, such an arrangement can serve as a check on your own inexperience.

  3. Thank you. I’ll keep it as a guideline. I just started with this.

  4. William Gallagher says:

    My advice to add to the above sensible advice: Have an interesting research question and be able to explain why the data you’re analyzing are the most appropriate means of exploring that question. It seems to me that part of the research process is under-developed in much empirical work by beginners. (At a seminar on empirical methods sponsored by the American Bar Foundations a couple of years ago, the first question from an aspiring newbie to empirical research was “Where can I find a data set to work on.”) Letting the data or method drive the research is less than ideal–leads to bad, uninteresting research.

  5. Tracy Lightcap says:

    Just two things:

    1. Learn R. It’s the wave of the future, it’s updated continuously by people who know what they are doing, and it’s FREE! It’s quirky, but there’s a good windowing front end to it, John Fox’s R Commander. I use R in my classes since the grad schools are all going to it. Along with businesses, governments, think tanks, consultants …

    2. Learn to frame research questions so that the datasets you develop aren’t the size of the GM parts inventory. See Lane Kenworthy for examples:

    You can even do classroom demos with data you put together yourself!