Research Question

This post is on how to choose, formulate and present a research topic--the planning stage of your research. A good plan is not only important for carrying out the research and writing up the dissertation, but also essential for communication with your supervisors, examiners and future employers.

Many of you have difficulty in selecting a research direction. Well, you should, because this is the most important stage of planning. A wrong direction leads you to nowhere and you might have to come back to this stage after a long detour. It is like looking for your soulmate or the other half. "If you haven't found it yet, keep looking. Don't settle." as Steve Jobs ever said. My suggestion is to choose a research question from the area you are going to start your career (e.g. commercial banking, investment banking or accounting). At least you will be able to impress your employer during the interview by your knowledge of the industry. However, if you are determined to do something you are really interested in rather than a pragmatic choice, it is totally fine--some of my past students choose to study football economics or economic evaluation of rugby worldcup! Whichever direction you are going to choose, you need to narrow it down to a research question in the form of some testable hypotheses. Something like "the effect of hosting the worldcup" is too wide to be counted as a research question.

A good research question must have some sort of "innovative" contributions to the literature. This "innovation" could be:

1) new concept/idea never identified before in the literature; or

2) new data never collected before in the literature; or

3) new approach never used before in the literature; or

4) a combination of any of the above.

So not all topics you are interested in can be a research question. You need to SELL your research to the readers by clearly spelling out what is the novelty and innovation of your research. Many of you failed to do this in your presentation of your research proposal.

A very key step of this job is literature review. Still many of you are not 100% sure of what can count as an academic "literature". Let me tell you what is NOT:

- a journalistic report such as Financial Times and Economist is NOT;

- a textbook is NOT;

- an unclassified journal paper is NOT.

They are not literature because they are either too descriptive or low quality.

Then what could be eligible? Usually, we use specialised academic journal articles (including working papers targeting for these journals). In the UK, the "ABS journal ranking" (which can be downloaded here) is the "bible" of the academic staff. We use this ranking to classify the influence and quality of a journal article. You should focus on the papers published on these journals (ideally 4 star and 3 star ones) to write your literature review.

A technical problem is how you can obtain the relevant literature from these journals? We usually use "JSTOR" and "ScienceDirect" to search. I suggest you to use the most recent publications, then read their literature review as a start. It will then tell you more references. This technique should be accumulated by yourselves because it is not a supervisor's job to do this step for you.

After reading a comprehensive "literature", you will then have to develop your own research topic. Don't forget to emphasise your "innovative" contributions rooted in the literature, or the so-called "gap in literature". Your research objectives and aims should then developed based on this (as shown in the presentation).

Next, you need to formulate this research topic/idea as a set of "testable hypothesis". This is a step where you trade off between your research ambition and your academic ability. Obviously you try to be great and want to answer all questions in your research, but most of the time you won't be able to win a Nobel prize by your dissertation. So a down-to-ground hypothesis summarised from your topic must be feasible in difficulty and doable in technicality. For example, data for the hypothesis testing should be reasonably easy to get. And the data analysis techniques (such as econometrics) should not be too difficult for you (if you really hate maths).

A usual form of a testable hypothesis takes this form: "variable x positively determines variable y." For example, you may want to test whether "the marginal trading in China significantly destabilises the financial market", or "the monetary policy in the UK is effectively reducing the volatility during the business cycles".

The first point you need to be aware is that we are trying to find out a CAUSALITY relationship between x and y, NOT just correlation. For example, my father's wage and your father's wage are positively correlated, but they are not CAUSING each other. Neither explains either. What you really want to find is what factors determine one's wage, such as education, experience and skills, not what is co-moving or correlating with it.

Another point is that usually the variable y is determined by many x's, not just one. So if you try to explain y by x, you need to control for all the other factors to get the PURE effect. You cannot just discuss the relationship between y and each x separately, because in that way you are missing out and ignoring the cross causality between x's, resulting in biasness in estimates. A simple solution is to run a multiple regression with all x's in your model.

The third point is about possibility of endogeneity. Sometimes, not only x causes y, but also y causes back x. In this case, we call the mutual causality "endogeneity problem". You cannot use ordinary regression because it leads to biased estimates too. If your research question has this theoretical possibility, you will have to use more advanced econometrics such as 2-Stage-Least-Squares or Structural Equation Modelling, instead of just simple OLS. These techniques are ready to use and widely accepted in the current literature. So if you want to have a robust and defendable dissertation with a publication prospect, these techniques are basic requirements.

In summary, I have talked about three things to plan/structure your dissertation:

1) Clarify the innovative contributions of your research topic, based on the gap of the literature;

2) Formulate a testable hypothesis which is technically doable and empirically measurable;

3) Accumulate a set of data analysis techniques (by reading the current literature or studying this module) to test the hypothesis.