老师评价:推荐信

发布时间:2011-03-30 10:14:39 论文编辑:第一代写网

I have taken an initial look over your submitted dissertation and the examiners' reports. I think that you will have to do a rewrite with some re-analysis and re-formatting to pass the examiners. This is feasible, but will require a fair bit of work on your part. I have some initial comments that I want you to think about as you start to plan your reworking of the dissertation (taking account of the examiners' comments).

I divide my comments into substantive and formal comments.

First, the substance of your dissertation is light. Your proposed research question is:

how does US trade affect the US dollar exchange rate?


You need to follow such a statement with an explanation of:


1) why the question is important


2) what the pre-existing theory and evidence on the proposed relationship is


3) how the previous research on your question is lacking and how you plan to address these shortcomings in your own research on the question


Unfortunately, I see little of any of these issues addressed in your dissertation. In particular, why are the two related? What mechanism connects trade flows and the exchange rate? What does theory tell us? What are the difficulties that an empirical researcher faces when trying to evaluate the relationship? Why not look at the effect of the exchange rate upon trade flows (this seems to be an equally valid effect)?


The main thing that distinguishes any good empirical study is its use of a research design -- a method by which we can disentangle the causation between two variables. For example, a two-stage least squares strategy uses an instrument in a linear regression framework to address possible endogeneity. The hardest part of any empirical study is coming up with a research design that is convincing or at least plausible.


There are some empirical problems with your specification:


a) You should look at the relationship between the rate of nominal exchange rate appreciation and the size of trade flows over the corresponding period (e.g., 5-10 years or something). There is no reason for the level of the exchange rate to be related to trade flows. The level of the exchange rate in equilibrium is just a function of the relative money supply sizes.


b) You mix fixed and floating exchange rate regimes. Why? They are surely different. For your purposes, I would think only floating exchange rate regimes make sense. Otherwise, you can get huge trade flows associated with no exchange rate movement under fixed exchange rates. Look at the regime classification schemes of Reinhart and Rogoff (2004) and Shambaugh (2004). You would need to make sure that countries are floating over the entire period that you look at.


c) Ideally, you should have something beyond a cross-country sample (e.g., a panel dataset). You make a huge assumption that the effect is the same across countries when you pool.


d) Even so, endogeneity will be a big problem in the simple relationship. Are there instruments that you could use for trade flows (assuming you stick with the interest in the effect of trade on the exchange rate)? For example, what about using the Frankel-Romer instrument and doing two-stage least squares? This is not perfect, but better than nothing.


e) In general, you need to include a lot more on the economics -- why is this an important question for an economist? How do the economics inform your choice of specification? Right now, you spend a lot of time talking about various econometric outputs, but you don't give much of a sense for why an economist should care nor what the estimated effects actually mean (do they indicate something about theory?).


f) Recall that the dissertation is supposed to be an original piece of research. You need to think about what your contribution to the literature is -- how is your approach different or innovative? How does it help us learn about the relationship? Is it convincing? This is all about external and internal validity, which you discuss some (that's a good thing to be aware of and good to include something on these issues).


Second, there are some formatting issues. You have embedded quite a few tables directly from the Stata output, which is not advisable. These don't look very good and include a lot of extraneous information (e.g., model sum of squares and other stuff). Ideally, your tables for linear regression outputs only include stuff that you discuss and that is relevant for your question. Typically, the most that this includes is number of observations, r-squared, f-statistic for the regression, coefficients, standard errors, and p-values (or t-statistics/significance levels). You should make the tables and graphs look nice and friendly for the reader (we don't want them to have information overload).


You don't need to include the data as a giant table in the appendix. It would be far better to create a table that documents your variables and their associated data sources, so that another researcher could look at the table and reconstruct your data set and re-do your regressions.


Your list of references is generally either quite dated (from the 1960s) or from textbooks. You want to try to use more recent references, in addition to classic references in the field (on exchange rates and trade). Ideally, a large component should be published articles (published in well-known economics journals).


These are just some initial comments. My recommendation at this point is to take a look over these comments and look over your dissertation again. Then, have a think about how you can change your dissertation and the analysis to make for a stronger piece of work. After some reflection, then start to write out an outline for the revised dissertation/work, taking account of the comments here and the comments from the examiners. I am happy to try to answer any questions that you might have as you work through this.