Live data from Hacker News

Bridging Economics and Data Science

medium.com

11–20 of 24 posts

Re: Bridging Economics and Data Science

#11
post #9
post #4

Earlier quoted context omitted.

I also come from an Economics background, and am now a software engineer/budding data scientist. As I've delved more into machine learning topics, I'm amazed (though not surprised!) at how both academic and industry economists are still mostly focused on running OLS/logit/probit regressions, and not other classification techniques. My undergraduate thesis did use some computational models that sought convergence for…

Macroeconomics and empirical industrial organization are leading the forefront in terms of theoretical and applied technical advances. You ought to look at discrete choice analysis sometime--great stuff. I can't speak for industry economists, but the reason we academics tend to spend so much time with OLS/Logit/Probit is their flexibility and scalability.

Macro was my favorite subject! I was lucky enough to take the first year PhD sequence during my last year, which was my first taste of coding =D

I think in industry (anti-trust at least), they stick with the older models because their value has legal precedent, and using new methods would require some more legal hand waving by the attorneys.

Re: Bridging Economics and Data Science

#13

This is the same in the social sciences. I'm currently a PhD student in one of the top Communication departments in the US, and it's painful to see how far behind in technical skills and tools the curricula are (eg., Excel and SPSS). I've been self-teaching Python, R, and SQL, and extending my knowledge base from simple regression-based stats to data mining and machine learning, to make up for it. Not only does that…

I'm in a well regarded sociology department and there's a huge gap between the sociologists who take technical things seriously (programming, stats) and those who don't. People have actually rolled their eyes in classes when we read papers about simulations. I was lucky: we have a core group of students who program and gather data using modern tools, but I gather this is rare both within my discipline and without.

Re: Bridging Economics and Data Science

#14

Ditto. I am following a similar path. I am curious if there are any groups dedicated to people like us (econ/consulting/finance people learning data science) to help facilitate the learning process?

By facilitate the learning process are you referring to the programming aspect or the math aspect of data science?

Re: Bridging Economics and Data Science

#15

Ditto. I am following a similar path. I am curious if there are any groups dedicated to people like us (econ/consulting/finance people learning data science) to help facilitate the learning process?

Not that I know of. If you're in the Bay Area (contact info on piece), drop me a line and we can grab coffee.

Re: Bridging Economics and Data Science

#16
post #13

This is the same in the social sciences. I'm currently a PhD student in one of the top Communication departments in the US, and it's painful to see how far behind in technical skills and tools the curricula are (eg., Excel and SPSS). I've been self-teaching Python, R, and SQL, and extending my knowledge base from simple regression-based stats to data mining and machine learning, to make up for it. Not only does that…

I'm in a well regarded sociology department and there's a huge gap between the sociologists who take technical things seriously (programming, stats) and those who don't. People have actually rolled their eyes in classes when we read papers about simulations. I was lucky: we have a core group of students who program and gather data using modern tools, but I gather this is rare both within my discipline and without.

That's a real shame, data science will hopefully cut through the theoretical BS that are the results of so many social science thesis'.

Re: Bridging Economics and Data Science

#17
post #5

I have an econ/finance degree and also worked for several years in finance. Then I quit my job and learned how to code. Now I have a web app with a paying and growing customer base. Knowing how to program will give you special powers when combined with your econ background. Congrats!

I have an econs background too. I learned to code in my final year of uni and now do some freelancing. You are right that knowing how to program will give you special powers!

Re: Bridging Economics and Data Science

#18
I'd encourage any economist that has a knack for programming to really put in some time towards the area. I've owned econpy.org for about 3 years now (although I haven't updated it in a long time). I also own economics.io and run econpy.blogspot.com.

It's much easier (relatively speaking) for an economist to pick up some programming than it is for a programmer to pick up some economics. Economists are already familiar with the types of questions that are important to economists, and more importantly, how to frame them. The trouble with economics is that you can't just pick it up overnight as it is a way of thinking more than it is a tool set. Programming on the other hand is something that you can "get working" overnight (economic programmers don't need to be algorithmic theorists -- they just need to be really good at getting/scraping data and organizing it so they can run analyses on it).

Over a year ago, I dropped out of my PhD program in economics because I was not at a school that was going to allow me to do the econ/cs type work I was working on. Leaving my PhD program was one the best things I ever did because it has allowed me to pursue whatever I want to do with the skills I've acquired.

The problem with academic economics is that the data most economists use is so bad and outdated -- such as data from FRED, BLS, and other publicly available sources where everyone and their uncle can download the same CSV dataset that was aggregated by some government employee. The race then is to see who can put together the most elegant econometric model to handle all the issues with the data. The rules of the game change when you create your own dataset and thus have control over while variables to include, the aggregation, the frequency, etc.

Long story short, if you are an economist wanting to do programming, learn to adapt those skills in academia (best way is to find a great advisor -- if there isn't one in your economics department, check the business school as bschool professors are often much more open to highly empirical analyses and care [marginally] less about getting the theory perfect). Or, if you want an easier lifestyle that is much more rewarding, ditch academia for the private sector. You'll find the economists in the private sector to be much more knowledgeable about cutting edge technologies and willing to listen and learn from what you have to say.

Re: Bridging Economics and Data Science

#19
post #8

The biggest issue I see from the ivory tower I'm surrounded by is that economists typically doubt the results of data mining. Neural networks, machine learning, etc. are all well known toolkits in computational economics (one of my specialties) but the results from their application are rarely believed.

Structural models are more 'insightful' than generative models in machine learning. Econ guys are more interested in Pr(y|x) than reproducing the data.

That being said, I really hope computational work gains more traction... this might be a marketing issue.

Re: Bridging Economics and Data Science

#20
He nails the point, and then misses it himself:

the key to success is in cleverly selecting, finding, or creating a data source that answers a particular question

It's about asking the right question and then finding or generating the right data to answer that question. That's what makes it science.

Post reply on HN