QM2 Econometric Project
As part of the course requirements you have to undertake an econometric
evaluation of an economic issue using data that you have gathered either
from the host of data sets now available on the web, or that you have
assembled yourself from data published in journals or in official sources.
The project is worth 20% of the total marks for this course
The data set must contain a minimum of 3 variables, (one dependent
variable and at least 2 right hand side variables to be added to a constant
term). You should concentrate on estimating a multiple (rather than simple)
regression model. You cannot use the data sets used in the problem sets or
lectures. You must find your own data (with 2 exceptions see below)
You should have a minimum of 50 observations in your regressions.
Anything less will be penalised heavily.
The model should be a causal one, (ie the right hand side variables should
explain the dependent variable, not the other way round). This also means, for
example, that you should not estimate identities like a national accounts
model of the form Y=C+I+G+(X-M). This is an accounting identity and so the
coefficients shouldn’t be anything other than one, barring measurement error).
The econometric results should include a thorough statistical evaluation using
the full range of (relevant) diagnostic tests highlighted during the course. Do
not use tests just for the sake of it. There are no marks to be gained by doing
countless irrelevant tests. Only use tests that are relevant to the type of data
and economic relationship you are trying to estimate. (For the purposes of the
project assume that a sample of 50 or more observations is enough to make
any asymptotic tests valid).
This project should be completed and handed in (via Turnitin) during earlier
weeks of the second term (date and time tbc)
In order for the project to be marked at all, you must also provide a disk or
USB containing i) the data you have used ii) the Stata output log containing
your regression output (any output – including graphs – generated using any
other package will be penalised) and iii) a copy of the project
Do NOT use a package other than Stata to do the project. Do NOT use Stata
commands unless they were covered in the course. This can often cause
confusion that this is your own work. Nor is it necessary. The project is a
demonstration that you understand and can implement the (relevant)
techniques/tests covered in the course. This includes mastering and
understanding relevant Stata commands.
Do NOT write your name on the project – just give your student number
You should aim for a maximum of 2,500 words or around 8 pages of text, 2
to 3 tables of results and 2 or 3 figures (not including the log file)Please
provide a word count on your cover page.
The idea is to choose an economic issue which you find interesting, outline a
theory and a set of testable hypotheses that follow on from that. Then test the
theory empirically using the tools you have learned during this term’s course.
The dissertation should read something like a typical article that you would
find in a (non-technical) academic journal like the Journal of Economic
Perspectives or the May Papers & Proceedings volumes of the American
Economic Review, (the collection of back issues are in the library).
Choosing a Topic
The most important thing is to choose a project that is feasible, that can be
finished within one month from start to finish and still allow you time to work
on your other subjects. This means confining your topic to a simple issue.
Also it is a good idea to choose a topic that you are interested in, rather than
one you fell you ought to do. The more you are interested the easier the
project will be. Do not write the theoretical part of your project until you know
you have data that can be used to test your hypotheses.
One good way to find a topic to study is to read the economic pages of the
broadsheet newspapers and/or academic articles summarised in overview
journals like the
Journal of Economic Perspectives
Journal of Economic Literature
both of which are in the library. In addition there are specialist journals, (and
therefore more technical), such as the
American Economic Review, Economic Journal, Quarterly Journal of
Economics, Journal of Labor Economics, Journal of Industrial
Economics, Journal of Development Economics, Journal of Finance
which should all be good sources of current issues concerning academic
There are a large variety of data sources on the internet that should meet your
Many UK macroeconomic statistics, (inflation, unemployment, gdp etc), can
be downloaded from the Office for National Statistics website
UK Regional data can be found here
The bized site also contains access to official UK data alongside company
account data and some international data.
You can find stock market data at the Stock Exchange’s web site
or from yahoo http://uk.finance.yahoo.com/q/hp?s=%5EFTSE
or from the Bank of England
A very good source of international data both cross section and time series is
given at the Resources for Economists website
and also the Statlib website http://lib.stat.cmu.edu
The World Bank also has data http://worldbank.org
and there are lots of data and ideas at
http://www.economagic.com/ and http://pwt.econ.upenn.edu
The library also has a useful link to some sites
For those of you interested in working with cross section data. I have put 2
different UK cross section data sets on the course web site
Health, Wages: GHS_project.dta – which has information on wages, health,
smoking, drinking, education and other socio-demographic characteristics of
individuals taken from the General Household Survey
(you can find a codebook giving details of the variables at
Consumption: Food_project.dta – which has information on household
spending on various consumer items taken from the Expenditure & Food
(you can find a codebook giving details of the variables at
You will have to choose which variables to model to make sure the data are
free of missing values and give economic reasons for your choice.
These are just guides to help you. You may, of course, find your own data.
Ideally your project should look and be structured like an article you can find
in any of the economic journals listed above. You are strongly advised to read
some articles to get a feel for how they are presented.
(there is an example article on the course moodle page)
So your project should include the following sections:
Set out the economic theory underlying your project and use it to specify a
model and the resulting hypotheses to be tested. Set out your prior
expectations of the likely signs and magnitudes of the coefficients. Discuss
any econometric problems you expect to encounter.
Discuss the sources for your data. Give the exact definition of variables (in a
Table in an appendix) and sample period, Describe the main features of the
data using a table of sample means and their standard errors. Graph the
trends in the dependent and, perhaps, the independent variables. Comment
on the main trends/features.
Outline the econometric techniques used to estimate your model, (eg.
ordinary least squares with corrections for heteroskedasticty/autocorrelation).
You need to convince the reader that you have made the right choice of
estimation technique. Evaluate the model using the set of (relevant)
diagnostic tests covered in the lectures. (Eg, Box-Cox, Ramsey Reset,
Forecasting). Do NOT report the results of the tests one after another like a
shopping list. Report the tests for each model at the bottom of a column of
estimates. (Again read a journal article for hints on presentation).
Outline your results in tabular form, (check with a journal if you are unsure as
to how to present your results). The Stata command “outreg” will help
considerably with you inputting the results in tabular form. State whether your
hypotheses are accepted or rejected. Comment on the results and on any
diagnostic tests you have used.
Give an overview of your hypotheses and main results
Always list the data sources and articles that you may have read at the end of
the discussion. Tables and Figures should come after the references.