Friday, October 30, 2009

The Future of Research Economics

Tyler Cowen notes some relevant trends here . I talk to people and they say that applied micro has suffered over the last 15 years as top Americans have gone to Wall Street rather than the professor route. When I was a graduate student 20 years ago, my entering class was 50% American. Now I believe that at Chicago it is 15%. So what? I'm go back and forth on the causes and consequences of the globalization of research economics. I look at the MIT faculty. It is a highly international group and they are certainly churning out great research except when the World Cup is played and then the place turns into the United Nations.

A deeper question might ask; "Why are Americans not entering academic economics?" Is it simply the "pull" of Wall Street big money? Unfortunately, I am slightly worried that economics has hit diminishing returns. There is a huge intellectual payoff from starting to know basic economics and statistics but are there increasing returns here?

I think of Watson and Crick and the Double Helix with many research teams all simultaneously trying to crack the same research question. They were working in a "stationary" environment. I don't believe that DNA changes over time.

Economics is harder. The agents we are studying form expectations of the future, are highly heterogeneous, their choices are often strategic and some claim that they even make mistakes. These agents have private information concerning their past history and this past set of choices and outcomes affects their decisions today (non-separable) On top of this, the economies we study are not stationary as they are bombarded with shocks (climate change, financial news, new products) that change the equilibrium we observe. If we could study a stationary process where people play the same games and situations over and over, I'm highly optimistic that we could use revealed preference methods to tease out interesting facts about people, firms and governments and then use our "structural findings" to inform policy debates.

I do have the sense that modern applied micro is too spread out. Our collective research covers a vast number of topics. I would find it satisfying to be in a field where there is one core question and everyone is trying to make progress on that question. In contrast, we have sprawled. Perhaps I know too little about other fields and they are equally sprawled.

Well, I just received a revise and resubmit email so please let me end this blog
to get back to the research frontier. Lucas please call me!

UPDATE: I see that people are reading this blog entry so I want to offer a quick
follow up.

I am proud to be a Ph.D. applied economist. If I was 22 again, I think I would make the same career choice. This is certainly a very good life. But in 2009, are the next 30 years brighter for academic economics or academic biology meets computer science or neuroscience? If in the year 2039, economics is in the midst of a "golden age" --- this would be a very happy surprise to me when I am a 73 year old man.

I want to be wrong here because it would be quite exciting to be part of a field whose cumulative insights are accelerating. What can economics do to maximize the likelihood that this remains an exciting time for our field?

A couple of "micro" ideas

1. Google could create a centralized data clearinghouse so that all data sets would be archived there after publication. No more ISSR. Journals such as the AER that require such archiving could forward on the data sets to Google.

2. The National Academy of Sciences could strongly lobby the government of the U.S and others around the world to reconsider what types of data are collected and to reconfigure panel data sets.

3. The U.S tax authorities should offer special tax deals for data companies to share proprietary data with academic nerds. I pay a lot of money for micro data sets from companies that sell real estate data or vehicle registration data. There is no reason for why I need to face such monopoly pricing when the true marginal cost of data delivery equals zero. This is slowing down science.

#1- #3 would improve data quality and guarantee replicability. It would lower the barrier to entry -- we know that the number of papers written would rise but would average quality rise? Given that we only care about the max of the set --- so what? but someone would have to review all of these papers.

I just stumbled across this recent speech by Alan Krueger. He appears to agree with me.

I am not smart enough to answer the riddle of the future of macro. We started with the 1 sector growth model with representative agents and common production technology. We now agree with the 2000 Nobel Laureate concerning the fundamental role that heterogeneity plays along many many dimensions in our modern economy. Once we allow the "heterogeneity genie" out of the bottle, how do we generalize from the field experiments we run? How do we do empirical work when we are accumulating high numbers of unobserved state variables that the agents are aware of and responding to but that the econometrician does not observe?