Dissertation Econometrics

Students studying with Andrew Seltzer

Andrew’s contact details

Phone: 01784 443-475
e-mail: [email protected]

Office hours: tba. Outside these times, I am available by prior appointment only. I am very busy and can not meet outside of my office hours on a regular basis. If you are unable to attend these hours, I 英國dissertation網would advise you to request another supervisor.

Data

Your dissertation requires you to use econometrics, unless you have been given a waiver. As all of the topics I have assigned are empirical and appropriate for econometric testing, don’t even think of the waiver route. I will not directly supply data, but below I outline how to down load one commonly used data set. (There are a number of other publicly available data sets at this web site. The directions below only need to be modified slightly for these.)

To do the empirical section you will need a data set. You may want to use the Labour Force Survey data set. This contains a random survey of English workers which asks several hundred questions of about 60,000 workers, and has information on things such as wages, occupation, education, age, sex, etc. I would advise students to use data from a single year exclusively unless you are specifically looking at changes over time and thus need multi-year data. The appendix to this handout outlines how to download the LFS data set to your computer from the UK Data Archive. There are a number of other interesting data sets on line at the UK Data Archive, which you may also want to use for your dissertation. Several of the topics that I have assigned this year require other data sets such as the British Household Panel Survey and the British Social Attitude Survey.

I would strongly advise you to use a readily available data set. If your topic does not lend itself to using one of these data sets, my recommendation is to change topics. If you do decide to proceed with a topic that requires a different data set, you should locate a suitable data set as early as is possible. One of the main reasons that students do poorly on the dissertation is that they fail to obtain a suitable http://www.ttc-switch.com/dissertation_writing/Ecommerce/data set early in the academic year. Under no circumstances will I find a data set for you; it is your dissertation, not mine.

In order to use the LFS data set properly, you will need to understand how the variables are coded. The UK Data Archive also allows you to download the LFS codebook accompanying each year’s data set. The actual variables in the data set change from year to year, and you need to be careful that your main variables are defined consistently if you are using data from more than one year.

You can find more information about the LFS at http://www.data-archive.ac.uk/findingData/lfsTitles.asp#p#分頁標題#e#

The LFS Stata files are huge. You will probably need to allocate more memory to Stat before you can open the files. The command for this is:

set mem 66m

(this allocates 66mb of memory to Stata)

I would suggest that you delete variables that you don’t need in order to economise on memory. You should decide on which variables you do and don’t need only after doing some background reading on your topic and thinking about your regression model. Remember, it is your dissertation and it is up to you, not me, to decide on your regression model.

When you have decided on the variables you want in your data set you can keep the variables you need or drop the variables that you need or don’t need as follows:

keep age sex wage

This will create a file with the variables age, sex, and wage. Everything should be in lower case letters. It is important that you get all the variable names exactly right or the program will not run.

drop age sex wage

This will keep everything but the variables age, sex, wage.

You should also spend a lot of time checking the data to make sure that they are in the right format for your work. Don’t assume that the data are “correct” and therefore you are ready to go. Check to see whether dummy variables are 0/1 not 1/2 or something else. Check to see that the code for missing observation is ‘.’ not -99, or some such. Do some cross tabulations to check whether the underlying relationship that you are addressing in your dissertation is really likely to exist. These checks take time, so start early.

Consultation

I am happy to meet with students about their dissertation. Many questions can be answered as easily by e-mail and I encourage students to use this option. My e-mail address is [email protected]

My office hours will be posted on my door and you must make an appointment if you want to visit outside these hours.

If you have routine questions about Stata or about econometrics (as opposed to big questions about the organisation of your dissertation, your list of readings, or the interpretation of your results) you should first contact the coordinator of the dissertation econometrics sessions.

What to expect from your supervisor

Early in the dissertation – I am happy to provide advice on how to go about doing a literature survey. If I suggested the original topic, I can usually suggest specific readings. If I did not suggest the topic, chances are that I won’t know the literature and won’t be able to suggest readings.

Comments – I will provide feedback on your preliminary report. Obviously the more work that you do, the more feedback that you will receive. I am willing to read one draft of your work, but please allow at least two weeks for me to provide comments.

As a general point, I can not help you very much with work that you haven’t done. In general the only thing I can say to “I don’t know what to do next”, etc. is something to the effect of “go away and do some work and then we’ll talk about it”. I will help you with specific questions about the literature, the data, the econometrics, the write-up, etc. I won’t outline the agenda of your dissertation for you. This prevents you from thinking through the larger problems and thus reduces the amount that you learn from the dissertation. I won’t find you a data set. I won’t proof-read your draft for spelling mistakes, grammar, etc. Both of these things are fundamentally your duty, and I simply don’t have the time to do them for you.#p#分頁標題#e#

Saving your work:

Regardless of what data you are using, it is imperative to save your work, much as you would with a WORD or any other file. Accordingly, you should keep logs, and program using .do files, rather than directly typing commands into Stata and losing this work when you exit. To use a .do file, follow these steps.

1) open a WORD document
2) write your Stata commands, eg.
set mem 50m
use “c:\mydata.dta”
gen var3=var1+var2
replace var3=0 if var3<0
3) FILE>SAVE AS> text file [make sure the extension is .do, not the default which is .txt]: suppose you call it program.do
4) open stata
5) do program.do

This will do all of the commands in your program.

6) save as – chose a different name, this way you can go back and edit your program and run it again on the original data set

To created a log of your Stata session:

1) type log using c:\filename.log
2) do your work in stata
3) type log close

This creates a text file called c:\filename.log that contains all of your work for the session.

You can also copy all of the commands from the review window and paste them into a word file.

There are a number of things that can go wrong with your data: it can become corrupted during an internet transfer, you may inadvertently erase or transform it, you may find that variables are not constructed the way you need them for your model, etc. If you use .do and .log files even the worst problems can be undone with less than half hour of work. Extensions will not be granted to students who lose time after failing to save their work properly.

Supporting coursework

If you are doing a dissertation in a topic in labour economics, I very strongly recommend taking EC3311 Labour Economics and EC3333 Econometrics. If you are doing a topic in economic history, I very strongly recommend taking EC3342 Topics in Economic History and EC3333 Econometrics.

Some frequently asked questions

By and large these questions show a lack of thought about the economics of what you are doing. Focus on that alone. The quality of your work is far more important the number of anything (words, regressions, variables, articles, etc). You should set up an interesting economics question and answer it. The rest is secondary.

1) I don’t know what to do next.

You need to work on the literature, the data, the econometrics, the write-up. I would suggest doing the literature first and the write-up last, but other than that the order doesn’t matter much. If you can’t think of interesting questions to be asking the data, then you should read more about what others have done.

It is your dissertation, not mine. You are responsible for setting the agenda.

2) How many variables should I have in my regression?

There is no magic formula. Data mining to achieve a target is bad economics.#p#分頁標題#e#

3) How many regressions should I have in my dissertation?

See the answer to 2

4) How many articles should I refer to in my literature review?

See the answer to 2

5) My dissertation is 1000 words under/over the word limit. Will I be penalised?

As long as you are reasonable, we don’t count exact numbers of words. We care far more about the content than the formalities. That said if you are 10,000 words over the limit, you should think about condensing things such as the literature review. Look at articles in professional journals and see how they handle the literature review.

6) I can’t find any literature on my topic. What do I do?

You need to think about how you define your topic. Say your dissertation is on adoption of mobile phones in Botswana. Perhaps no one has written an economics article on precisely this topic. However, there will be articles on diffusion of new technology, product markets in developing countries, regulation of utilities in developing countries, and a number of other closely related topics. There may also be relevant papers with useful econometric methodologies. If you can’t find any related literature in mainstream economics journals, then you should consider choosing a different topic.

If you don’t know what the mainstream journals are, see: http://www.le.ac.uk/economics/research/rankings/econ-rankings.html
 

Your dissertation requires you to use econometrics, unless you have been given a waiver. As all of the topics I have assigned are empirical and appropriate for econometric testing, don’t even think of the waiver route.

To do the empirical section you will need a data set. You may want to use the Labour Force Survey data set. This contains a random survey of English workers which asks several hundred questions of about 60,000 workers, and has information on things such as wages, occupation, education, age, sex, etc.

You can down load the LFS data set for any available years via Athens.

Go to http://www.rhul.ac.uk/information-services/library/

Click on resources via metalab

Click on find resource

Click on subject economics, sub-category databases

Click on Economic and Social Data Service

Click on login/register

Complete the registration

After registering click on login/register

Click on for academic research

Click on continue via Athens account

Click on UK Data Archive

英國dissertation網Type “labour force survey” into Search <enter>

Click on labour force survey

Add the quarterly data sets that you need, the download will provide a stata file and a pdf file documenting exactly what is in each of the variables.