Econometrics research suggestions

andrew

I'm doing an econometrics* module this year, and a significant portion of the mark for the year goes toward a research project. As far as ideas go, I've got one idea so far (to do with drug prices), but I'm not sure yet how feasible it is etc.

So I was wondering if anyone had any ideas, something you'd like to preform some econometric analysis on, but just don't have the time or inclination to do so. I'll post the results here/mail them to the person who suggested the idea, so that you get something in return too.

*For those of you who don't know, econometrics applies statistics to economic data. So what I'm looking to do is to show, using numerical data, the plausible existence (or nonexistence) of a causal relationship between several economic variables and another, Dependant variable. An example would be seeing how one's wages (the Dependant variable) are affected by one's IQ, Time spent in school, Race etc.

Time Magazine

Check if distance from Cork is positively related with people's self-reported happiness.

andrew

I'll title it 'Happiness in the mid Atlantic.'

Time Magazine

BTW if you really want to pwn this project, be very careful about your use of the word "causal". Running an OLS regression of wages on IQ, education, race (etc) will by no means trace out a causal relationship.

(Why? Because you're leaving out something important (innate ability) that is likely correlated with both education (an X variable) and wage (the Y variable). Thus you'll suffer from an omitted variable bias and your errors will not be random. To trace a causal relationship you'll likely need an instrument (like quarter of birth) or a natural experiment (something like a large, unexpected influx of immigrants) that you can argue overcomes the omitted variable. This kind of complexity is absolutely fine to ignore while an undergrad, but it's better if you avoid the use of the word of "cause" anywhere in the project. Bonus points if you include a paragraph explaining why the results are not causal. A+ if you implement an instrumental variable approach or some such.)

Addendum: the pretty girl explains it reasonably well here.

Anonymous1987

Data is always a pain so I suggest first finding a large dataset and then think what might be an interesting topic to explore using this data. The European Social Survey has a wealth of information covering topics such as politics, well being, immigration, religion, and socio-economic factors.

If you are set on the drugs prices topic and are struggling to find Irish data, check out the UK and US.

Économiste Monétaire

Head size and intelligence...

Time Magazine

Anonymous1987 wrote: »

Data is always a pain so I suggest first finding a large dataset and then think what might be an interesting topic to explore using this data. The European Social Survey has a wealth of information covering topics such as politics, well being, immigration, religion, and socio-economic factors.

If you are set on the drugs prices topic and are struggling to find Irish data, check out the UK and US.

Also check out the General Social Survey. It's US data, but in addition to the usual stuff like demographics, religious views, political persuasion etc, it asks hundreds of questions like "Should Muslim clergymen be allowed to preach hatred of the United States?" and "Did you use a condom the last time you had sex?" Classic stuff. I attach the full set of questions.

andrew

Time Magazine wrote: »

Also check out the General Social Survey. It's US data, but in addition to the usual stuff like demographics, religious views, political persuasion etc, it asks hundreds of questions like "Should Muslim clergymen be allowed to preach hatred of the United States?" and "Did you use a condom the last time you had sex?" Classic stuff. I attach the full set of questions.

I've just skimmed through it so far but it looks excellent, thanks a lot.

Ostrom

You could register for access to the uk data archive (via. Exeter Uni) - you might find something useful. Cant remember seeing drug data in any of the big irish sets, maybe try statcentral.ie? I think the QNHS may have had a supplementary module on health a few years ago (cant check from my phone, but they are listed at cso.ie). If there was one, you should be able to get an anonymised sample.

andrew

There's a criminology lecturer in DIT who I was talking to, who mentioned he had a big dataset on Irish drug statistics, so I was going to ask him. Though I'm not sure how protective researchers tend to be over their data?

Économiste Monétaire

If the researcher collected the data himself, or paid a few minions to do it, then he might not give you the dataset.

Do you have a particular model in mind?

andrew

Économiste Monétaire wrote: »

If the researcher collected the data himself, or paid a few minions to do it, then he might not give you the dataset.

Do you have a particular model in mind?

I'm not sure of the exact nature of the data he has, but my idea was to do a linear regression which tested the response of the price of drugs to seizures by the gardaí. As far as I know, he has the kind of data which would be able to show this, though to be honest I'm not exactly sure how to go about it (I would hope that I'll learn to in the next few months).

Économiste Monétaire

Hopefully you'll cover time series before your project is due. Be careful about implying causality, as TM said, but be especially careful when working with time series data. Email the DIT lecturer as soon as possible, because if you can't get data then you're just wasting time on an idea that isn't feasible.

andrew

Économiste Monétaire wrote: »

Hopefully you'll cover time series before your project is due. Be careful about implying causality, as TM said, but be especially careful when working with time series data. Email the DIT lecturer as soon as possible, because if you can't get data then you're just wasting time on an idea that isn't feasible.

We'll definitely have covered time series, as it isn't due until the end of the academic year, though you're right about emailing him ASAP. As far as the idea goes, and assuming I have the data, would it be a particularly difficult topic to do? At this stage I know nothing about time series analysis.

Time Magazine

andrew wrote: »

We'll definitely have covered time series, as it isn't due until the end of the academic year, though you're right about emailing him ASAP. As far as the idea goes, and assuming I have the data, would it be a particularly difficult topic to do? At this stage I know nothing about time series analysis.

Easy. You'll want to add in a couple of other variables (like GDP) to account for other things, but it's not very difficult.

Économiste Monétaire

andrew wrote: »

We'll definitely have covered time series, as it isn't due until the end of the academic year, though you're right about emailing him ASAP. As far as the idea goes, and assuming I have the data, would it be a particularly difficult topic to do? At this stage I know nothing about time series analysis.

Not difficult to do, just one needs to be careful. You'll probably cover the standard issues, such as heteroskedasticity, autocorrelation, multicollinearity, endogeneity and IVs, et cetera, so you'll be well equipped to spot most potential problems. Stationarity is important for questions implying causality in time series; you'll see this when you cover spurious regressions. Best to get the data first, though .

andrew

As it turns out, the drugs data is, and I quote 'all in my head' but 'from reliable sources.' God Damn.

gerry87

andrew wrote: »

As it turns out, the drugs data is, and I quote 'all in my head' but 'from reliable sources.' God Damn.

You'll just have to hit the streets and do some primary research....

You could try and think about something involving financial markets, if for no other reason than there's heaps of data. It can be something novel, I was considering doing an econometrics project on stock market returns and the weather to see if good weather made happy and optimistic investors, i didn't but apparently how bright it is outside does matter.

http://www.behaviouralfinance.net/ would be a good place to try and find an interesting anomaly then you could try and come up with an idea to explain it.

Économiste Monétaire

andrew wrote: »

As it turns out, the drugs data is, and I quote 'all in my head' but 'from reliable sources.' God Damn.

Well, at least you didn't discover this 2-3 weeks from the deadline :pac:. Keep it simple with U.S. financial/macro data, and maybe mix this with a behavioural question as Gerry said, or you could work with the ESS and GSS panel data that Anonymous and TM linked. You might also want to look through OECD.Stat for some ideas on a macro panel project.

andrew

It's looking like that's what I'll do. Ah, it's disappointing though, not being able to get the data you need. I was half thinking about a survey... but that'd be too dodgey. I'll probably keep it simple, as suggested. It's going to be a long christmas break! Thanks for the suggestions though, i'd be sorta lost without them.

Économiste Monétaire

andrew wrote: »

Ah, it's disappointing though, not being able to get the data you need.

Welcome to our world .

andrew

Well, I've settled on examining fertility levels using GSS data (Thanks TM!). Anyway, I was wondering if there exists a measure of the cost of 'all other goods' which i could use in the regression. I know this is what the CPI does, but since I'm only taking into account costs in a single period, I don't think I can use the inflation rate, and if I measure them relative to some other time, then what time do I use?

andrew

Here it is, for anyone who's interested in why women have babies. I have a new found appreciation for people who collect data. And competently analyse said data too.