Thoughts on Choosing a Dissertation Topic

François Nielsen

The following are general scenarios / thoughts / recommendations on choosing a topic for a dissertation.  Note that each recommendation has exceptions -- without exception!

People find ideas for dissertations from a wide variety of sources; so relax and be open-minded.

Examples:

• I began my dissertation on the Flemish movement in Belgium reluctantly, because my adviser Mike Hannan had become interested in the topic, as an instance of ethnic resurgence in an industrial society, and I happened to be from Belgium. At that time I would have preferred a topic related to social stratification. But eventually I got into it and completed the dissertaiton successfully.
• Art Alderson (now at Indiana) chose his topic on effects of globalization in relation to, but separately from, a joint research program with me on income inequality
• Reading the literature on a substantive topic to understand the theoretical and research design issues (as in soci707) is often a useful way to get research ideas
• Kathi Harker Tillman used the AddHealth data she had been working on as an RA to develop a new classification of types of  non traditional family structures and estimating effects of family structure on adolescent schooling
• Susan Bailey worked several years at RTI analyzing data on substance abuse. Eventually she decided to do her dissertation on determinants of adolescent substance abuse, using one of these data sets. Her boss at RTI was added as a member of her dissertation committee.

Choose a topic that is "big" enough.

A very narrow topic may be easier and quicker to execute, but you can get bored, experience La Nausée (i.e., become depressed), or experience a feeling of emptiness ("so what?").  Example: redo a small group experiment with the door to the lab open rather than closed.

Don't worry, you will have plenty to talk about.

There is no need to be anxious about the length of your dissertation.  Once you start writing and you add the tables and the figures and the references and the footnotes and the appendices, before you know it you will have a huge document.

Do not stake everything on a specific empirical result.

This may simply mean keeping an open mind and avoiding premature conclusions.  If one sets out to prove that variable X interacts with variable Z to cause variable Y, what does one do if the interaction is non significant?  Change topic, drop out of graduate school, beat up your dissertation adviser, start drinking heavily...?  Better to define topic as a study of the interrelationships of X, Y, and Z, without committing to a specific conclusion in advance. E.g., in a paper on home cleaning services I once read where the author was determined to show that the power relations described by Hochschild were at work for the new cleaning services, even though her data showed exactly the opposite (relations between cleaning services and customers were essentially impersonal and bureaucratic).

Don't start out with a SEM with 47 variables, 5 reciprocal effects, and 7 correlated errors.

People sometimes get too heavily invested in a complicated, difficult model, and eventually discouraged when their reciprocal effects cannot be estimated or their modifications indices tell them to correlate more errors.  Keep it simple! (See next point.)

Subordinate the methodology to theory & data, not the other way around.

For example, define a topic as "determinants of interracial and interethnic dating relations among adolescents", not "a LISREL model of interracial and interethnic dating relations among adolescents".  AFTER you are done with your research you may want to use the "LISREL model..." title, but it is better to keep your methodological options open beforehand.

In rare cases one might run into unusual synergy between theory, data, and methods; but don't count on it.

In my research with Art Alderson we hit on the idea that statistical techniques for unbalanced pooled time series of cross sections (i.e., unequal numbers of observations per unit over time) were ideally suited to analyzing cross national data on income inequality.  We were able to produce several successful papers out of that realization. Art used these techniques in his dissertation.

Q: How long does it take to do a dissertation?  A: One year.

If, that is, you have a straightforward topic, you use existing data, and you don't do too many things besides you dissertation. E.g., Biz Pressler-Marshall's dissertation on intergenerational transmission of attitudes on gender roles was done in one year, a period during which she also had her first baby!  In general, try to avoid collecting your own data!

Weigh the pros and cons of choosing a topic in which you are emotionally, philosophically, or politically engaged.

Being engaged may be a powerful motivation for research.  But on the other hand it is crucial to be able to look at a topic dispassionately.  If one sets out to prove a predetermined conclusion preferred on external grounds one runs the risk of being empirically wrong and not being able to cope with this, or of distorting the importance of a topic out of reasonable proportion.  Examples: a student active in a battered women shelter studying the decisions of judges to make the mother or father legal guardian of children, with the theory that judges systematically decide in favor of the father; deciding to study the gay-lesbian community in the wake of discovering one's own homosexuality; deciding to study AA as a recently sober alcoholic.

Last modified 16 Oct 2007