fbpx

Machine Learning, Collusion, and Infrastructure, Oh My! Some Observations from the 2018 Annual Meetings of the American Economic Association

Machine Learning, Collusion, and Infrastructure, Oh My! Some Observations from the 2018 Annual Meetings of the American Economic Association

The American Economic Association annual meeting is where economists go to argue with other economists. It is also where soon-to-be-minted PhDs put on suits for the first time and interview for jobs. Truly, you have not lived until you’ve spent a long weekend with 13,000 social scientists attending the Allied Social Sciences Association (ASSA) meetings. For example, you can’t imagine how many jokes there are about price and quantity. Three. There are three. And they don’t get funnier with retelling. This year the meetings were held in Philadelphia, which, based on temperature and time required in transit to get there, was temporarily relocated to the North Pole.

Still, it’s an important conference for keeping up with the state of the art in economics research, especially in areas that are or will be relevant for policy. This post highlights the papers and discussions that TPI’s intrepid Research Fellow, Sarah Oh, found most interesting from the three days of meetings. Given the literally hundreds of sessions and thousands of papers being presented, we make no claim that these are actually the most interesting sessions, but we did our best.

Machine Learning: Causation vs. Correlation

A Friday afternoon session on Machine Learning for Policy Research addressed the difficulties of applying machine learning tools to policy research. Most importantly, machine learning is often used to find correlations, but policy work is about identifying causation. How can we adapt the tool for the job?

This session, presided over by Susan Athey (Stanford), included the latest theory papers on causal inference. The room was packed with overflow seating for more than 200 economists in attendance. One presenter, Victor Chernozhukov (MIT), remarked on the differences between theory and practice in machine learning research. While the academic field is advancing, he reminded the audience that current tools still are not that smart. For example, he noted, a photo recognition tool concluded that he is African when he is actually Russian.

Horizontal Collusion

Next door, former FTC chief economist Ginger Jin (Maryland) presided over a panel on Horizontal Practices: New Analysis of Collusion and Market Structure. The panel highlighted new models on horizontal practices that industrial organization economists are building based on familiar workhorse models. Joseph Harrington (Wharton/Penn) outlined a new two-firm model, with a focus on coordination in list prices and collusion in negotiated price, namely via surcharges. Mark Satterthwaite (Kellogg/Northwestern), in comments on a different paper, however, noted limits of the two-firm model for studying information markets with many firms.

Mechanisms, or Lack Thereof, to Commit to a Promised Activity

David Laibson (Harvard) presented his behavioral economics research as this year’s Ely Lecture keynote speaker. He noted that as a graduate student he needed his faculty advisor, Olivier Blanchard (Peterson Institute), to keep him on deadline, raising the possibility that Blanchard was asked to introduce his former student in order to make sure Laibson prepared for the lecture. Laibson’s work highlights that commitment devices are generally necessary to ensure follow-through, but are rarely used in practice without other enticements.

Global Infrastructure Challenges

Early on Saturday morning, economists trickled into an 8 am session on The Global Infrastructure Investment Challenge. Peter Henry (NYU) noted that low interest rates, low growth rates, and capital flows between advanced and emerging economics are driving global investment patterns. Paul Romer (World Bank/NYU) discussed a novel approach to using land policy to address urbanization and infrastructure problems related to the expected 4x growth in cities by 2050. He emphasized land planning for faster urbanization, likening city grids to platforms that enable growth.

Lawrence Summers (Harvard) compared infrastructure projects to his golf swing. He said that new strategies for infrastructure can sometimes sound like “how to fix your golf swing.” If only he could improve his back swing, follow through, grip, and posture, then he could golf just like Tiger Woods. But just like his golf swing, infrastructure projects can fail for a multitude of reasons, without a single strategy to fix them all. (Perhaps his analogy also demonstrates how, like Tiger Woods, just because infrastructure projects were once well-executed, there is no guarantee they always will be).

Anusha Chari (UNC-Chapel Hill/NBER) noted that political risk remains across the world, deterring private investment. Private investors are looking for projects in the developing world with high marginal products of capital, but expropriation risk from sovereign governments remains substantial.

Central Banking and Safe Assets

Also on Saturday, monetary scholars discussed The Balance Sheets of Central Banks and the Shortage of Safe Assets. Ben Bernanke (Brookings Institution) presided over the panel on global demand for safe assets and money-like assets. While not the focus of the panel, Willem Buiter (Citigroup) suggested that central banks can supply more safe assets, citing innovations in interest, and briefly mentioned the possibility of central banks offering cryptocurrencies.

There was a lot of other stuff, too

Other sessions from #ASSA2018 generated buzz, with topics ranging from tax reform to artificial intelligence, labor markets, and gender in the profession. Next year’s annual meeting will be held further south in Atlanta, and will be held in even warmer conditions in San Diego the year following.

Share This Article

View More Publications by Sarah Oh Lam and Scott Wallsten

Recommended Reads

Related Articles

Sign Up for Updates

This field is for validation purposes and should be left unchanged.