Undergraduate Thesis

Geoff_rey

Hello everybody,

I'm currently pursuing an undergraduate degree in Economics, and, as stipulated by the final year of the course, I have to submit an economic paper on a topic of my choosing.
I'm currently trying to gather some data in relation to cross-border shopping (hey, it seemed current at the time). So far, I've followed a number of leads, but I've enjoyed little success. The problem, you see, is either, given my status as an undergraduate, the quantitative methods I have viewed have not yet been covered, and so are - for the present - a little too advanced, or, alternatively, the data required for the types of quantitative methods that we have covered remain rather elusive.
Given the above, has anybody any advice on where I should be focusing my research?:o

Thanks a million!

Time Magazine

Welcome to Boards. Fairly obvious first point of contact would be your supervisor - do you have one?

If you have a read through the journal of economic perspectives, you can get a solid understanding of a subject (along with some good references) presented in a non technical matter.

As above, your supervisor should be able to point you towards some good reading and there is always REPEC (http://ideas.repec.org) which gives you free access to a lot of material.

If your topic is macro or financial based, things like The Economist or reports from banks or other financial institutions can also give good groundings.

Flamed Diving

Would literature of this nature exist for neighbouring US states with different VAT?

Geoff_rey

Hey guys. Thanks for the replies.

Yeah, my supervisor said that the thesis can be predominantly qualitative, but implied that some econometric analysis should be accommodated where possible - namely, multivariate linear regression models.
In light of this, I've considered what I should use as variables. I arrived at the following: indirect taxes (VAT and excise duty), price differentials, search costs, transportation costs, wages, etc. as my independent (x) variables; however, I'm not quite sure if I'm barking up the right tree.
Also, I've considered various other quantitative methodologies: optimization approaches, elasticities, wage-setting and price-setting relations, etc., but data for these models has proven quite difficult to acquire.
I'm pretty sure that some of these approaches are appropriate, but if I can't quantify them, are they of any value to the thesis?

Économiste Monétaire

A simple log-log model would estimate elasticity (or log-lin if you're using sales ~ VAT). Avoid the optimisation thing, you'll fall into dynamic programming which is beyond most undergrads.

Flamed Diving

Are these variables backed by the literature? Is there a precedent where someone took these variables and used a linear regression? Instead of trying to guess what variables to include, you should always back the inclusion of each variable by theory or literature.

How many countries do you intend to include? How many years? Why? Try to tackle these questions first. Try to justify why you are doing this, to yourself.

Geoff_rey

Deleted User wrote: »

As above, your supervisor should be able to point you towards some good reading and there is always REPEC (http://ideas.repec.org) which gives you free access to a lot of material.

Zaraba: Yeah, I've been looking at a number of papers on that site, and whilst they have been of much assistance, the quantitative methodologies on most seem rather arcane.

Flamed Diving

This paper might lead you somewhere and isn't too exotic, but it isn't exactly what you need:

http://web.utk.edu/~dbruce/sej06.pdf

Geoff_rey

Flamed Diving wrote: »

Would literature of this nature exist for neighbouring US states with different VAT?

Flamed Diving: I've been looking at a similar type of paper dicussing indirect taxation differentials in Scandinavia, but I'd presume the approaches are fairly similar.

Économiste Monétaire wrote: »

A simple log-log model would estimate elasticity (or log-lin if you're using sales ~ VAT). Avoid the optimisation thing, you'll fall into dynamic programming which is beyond most undergrads.

Thanks for that. Now, this is probably going to sound ridiculous, but this is kind of where I'm coming unstuck: where will I get data on the relevant elasticities to aplly the model?

Flamed Diving wrote: »

Are these variables backed by the literature? Is there a precedent where someone took these variables and used a linear regression? Instead of trying to guess what variables to include, you should always back the inclusion of each variable by theory or literature.

How many countries do you intend to include? How many years? Why? Try to tackle these questions first. Try to justify why you are doing this, to yourself.

Thanks a million again! This has already provided me with a newer focus.

Flamed Diving wrote: »

This paper might lead you somewhere and isn't too exotic, but it isn't exactly what you need:

http://web.utk.edu/~dbruce/sej06.pdf

Cheers! It's really appreciated!

Économiste Monétaire

Geoff_rey wrote: »

Thanks for that. Now, this is probably going to sound ridiculous, but this is kind of where I'm coming unstuck: where will I get data on the relevant elasticities to aplly the model?

You estimate the elasticity in the model. Say you have a model for wages and years of education

[latex]\displaystyle W_{i} = \beta_{0} + \beta_{1}E_{i} + \mu_{i}[/latex]

If you want to estimate a % increase in wages from a one year increase in education

[latex]\displaystyle log(W_{i}) = \beta_{0} + \beta_{1}E_{i} + \mu_{i}[/latex]

You'll get a coefficient number for [latex]\displaystyle \beta_{1}[/latex] when you run the regression, which, if you multiply by 100, you'll get your answer.

Or, imagine if you work on commission,

[latex]\displaystyle W_{i} = \beta_{0} + \beta_{1}S_{i} + \mu_{i}[/latex]

where S is 'sales'. You can make this a 'log-log' model, which satisties the definiton of an elasticity (a percentage increase in sales, increases W by X%).

[latex]\displaystyle log(W_{i}) = \beta_{0} + \beta_{1}log(S_{i}) + \mu_{i}[/latex]

The [latex]\displaystyle \beta_{1}[/latex] in this model will by the elasticity of wage w.r.t. sales.

These a pretty basic examples, but think how they would apply to your multivariate model. What variables do you think affect the volume of cross-border sales? What should the 'functional form' of your model be? (log-lin, like the first example, or log-log in the second, etc.).

Flamed Diving

Another tentative idea is if you are considering doing a cross-sectional study of, say, 15 countries (EU-15?), you could not only measure changes across countries, but also across time using a panel data model. They are really simply to use, say you get your model for the year 2009:

[latex]\displaystyle Y_{it} = \beta_{0} + \beta_{1}X_{it} + \epsilon_{it}[/latex]

Well, in a simplistic explanation, it is like getting this regression for the year 2000, 2005 and 2009 and then saying "Well, in each year the countries had such and such a relationship, but between 2000-2005 and 2005-2009 the value of Beta increased, suggesting that sensitivity increased, etc, etc".

The model is easy to run and easy to interpret (plus it helps you increase n!) and it gives you an extra dimension to your thesis. Anyway, talk to your supervisor if you are considering that idea, and if you need help, ask us on here.