The Raj Chetty course “Using Big Data to Solve Economic and Social Problems”

Syllabus, slides, and videos of the lectures, here you go.


The front page graphic is pretty striking:

Man, those people born during the 1940s almost all made more than their parents. Pretty sweet deal.

I'm not sure the American dream is to make more than your parents. In fact that seems like a pretty poor approximation, and one doomed to disappoint given the historical inevitability of lower rates of GDP growth of advanced economies.

The "fading dream" graph does support a "great stagnation" view.

But I'm with Mathesar, “Never give up, never surrender!“

If only Chetty could understand that fancy econometrics is not a substitute for having high-quality data.

Your bitterness feeds my soul.

You're welcome!

He could well use the world famous rebuke of econometrics "evidence":

"Correlation is not causation"

This is a weird comment, since Chetty has made a career of getting access to higher-quality (and "bigger") data than anyone else. He literally does nothing fancy in econometric terms.

His data are bigger, not higher quality. The tax records dataset in his mobility studies is a case in point. He doesn't observe several essential variables for his analysis, e.g. race or income beyond early career, but tries to overcome this statistically (and largely fails). There are many smaller but much higher quality datasets, e.g. the various NLS surveys, that can be used for analyzing the same questions.

You cannot do what Chetty does with the tax records with surveys like the NLS. The density at the neighborhood level is far too low to get anywhere near the level of precision you need to extract anything other than noise from the data, and the measurement of income is much less reliable. In the tax data, you can measure income beyond early career for anyone except the children of parents also observed in the sample period, and this will naturally improve with time. And I am aware of work underway to do a better job adding race to the analysis (and I completely disagree with the notion that you can't learn anything useful without observing race). Argue with his methods if you want, but you can't argue with the fact that the tax data are great data, and Chetty's been instrumental in getting insights from that data to the public.

“Using Big Data to Solve Economic and Social Problems”

... WHAT economic/social problems has Chetty solved with his BIG DATA ??

no results = no credibility

Big data is extremely dangerous if misused

Yes. So is small data. So is anything with the potential to be useful.

Medium-sized data are the best.

21 cm is the best size, says my wife.

90% of data considers itself medium sized

Looks good. Thanks.

Mo' data, mo' problems.

Rather than criticize the research (which I think is quite interesting), I find this to be a strange course. It doesn't look like he has the students actually analyze the data themselves - it looks like a promotion of his own work. It's find for him to self-promote, but for students (even without a statistics background) it should be a course that exposes them to the data directly. Maybe he does that - but from the materials it is not obvious. It looks like a bunch of provocative graphics that are derived directly from his analysis, not from having students work with the data themselves.

So, does this mean we study Chetty's course and then there is noting left to do in Econ?
All problems are solved? Nothing left to be done?
Should we start closing out all Econ Departments around the country, starting with MIT's?

Thanks Raj! Finally I can do something else than running regressions!

New Theory Cracks Open the Black Box of Deep Learning

In the talk, Naftali Tishby, a computer scientist and neuroscientist from the Hebrew University of Jerusalem, presented evidence in support of a new theory explaining how deep learning works. Tishby argues that deep neural networks learn according to a procedure called the “information bottleneck,” which he and two collaborators first described in purely theoretical terms in 1999.
Discovered theory of everything, but do not realize it yet.

Version 1.0 of the singularity is happening as financial centers learn the new theory.

Ya, when Chetty found he could get monopoly access to big datasets, he lost interest in his previous role as identification sheriff. Opportunistic guy to say the least.

This Raj Chetty?

"In 2008, The Economist and The New York Times listed Chetty as one of the top eight young economists in the world.[10] Chetty is among the most cited young economists in the world.[11] In 2010, he received the Young Labor Economist Award from the Institute for the Study of Labor for his paper "Moral Hazard Versus Liquidity and Optimal Unemployment Insurance" in the Journal of Political Economy.[12] In 2012, he was one of 23 fellows to receive $500,000 over the following five years from the John D. and Catherine T. MacArthur Foundation as a recipient of one of the Foundation's 'Genius Grants'.[5]

Chetty was the recipient of the 2013 John Bates Clark Medal, awarded by the American Economic Association to "that American economist under the age of forty who is adjudged to have made a significant contribution to economic thought and knowledge."[13]

Chetty was awarded the Padma Shri, an award for distinguished service in any field, by the Government of India in 2015.[14]

George Mason University economist Tyler Cowen has described Chetty as "the single most influential economist in the world today."[15]" From Wiki

Great. What we need is a merger between bad data and central planning. What could go wrong?

Comments for this post are closed