The longer you write, the more weaknesses you reveal. Many IO job-market papers are 50+page long (and mine, too, was as long), but they don't have to be. Take time to rewrite & shorten before journal submission.
Before submitting your "first neural-nets paper in economics that revolutionizes everything" to Econometrica, read & cite Xiaohong Chen's many papers:
Chen & Shen (1998 EMA)
Chen & White (1999 IEEE info theory)
Econ PhD students: Let me know if you happen to see any of my papers on graduate syllabi or reading lists, and I'd be happy to provide data and code in return (except for confidential/proprietary ones).
Students often "accept" AI/ML (with the zeal of the convert, forgetting everything from econometrics), while some professors just "reject" it as irrelevant. Here's the third way:
Everybody talks about "competition & innovation" as if we knew anything about it. No, we didn't... until this paper!
[Published] Mergers, Innovation, and Entry-Exit Dynamics
Presented new paper at DICE, Düsseldorf Institute for Competition Economics. Apparently, this is the main economics department of the university, with 50!? IO economists, PhD students with many good ideas, and no coursework in macro.
Machine learning is good for pattern recognition, so let's use it for recognizing patterns that matter.
DETECTING EDGEWORTH CYCLES
—with Simon Scheidegger
@comp_simon
(HEC Lausanne) & Timothy Holt (USI)
My first deep-learning paper. Comments welcome!
Thanks, David, and here's the replication package (data & code in MATLAB, plus a 6-page README with detailed instructions) for your PhD students!
*Interested in Dr. B. Douglas Bernheim's expert witness report (2002a)? I can share it personally:)
Someone at REStud kindly shared download stats for replication packages. Our vitamin paper was the most downloaded in 2021 ("Measuring the Incentive to Collude" w/
@takuo_sugaya
)
If you don't mind contributing to the 2022 stats:)
Revised & resubmitted "Mapping Firms' Locations in Technological Space"—Thanks to advice from editors & reviewers, now it comes with an IO model, detailed case studies, & tons of sensitivity analysis.
My smuggling of Empirical IO into undergrad teaching continues... this time moving on to the supply-side: translating cartel markups into equivalent numbers of symmetric Cournot competitors
Revised & resubmitted "Privatization and Productivity in China" to RAND, with Mo Xiao (Arizona), Masayuki Sawada (Hitotsubashi), and Yuyu Chen (Peking).
We augment Gandhi, Navarro, & Rivers (2020 JPE) and study selective/endogenous privatization of SOEs.
Updated my topological-mapping-of-technological-space paper to ver. 3. Now it's contextualized in the innovation literature since 1962. Because it's the first TDA paper in economics, I wrote a 1-page "TDA 101" section (that non-topologists can understand)
presenting new paper at NBER-SI Innovation, Wednesday 7/17 at 11:30am.
Livestreaming, paper, slides here:
welfare analysis of product & process innovations —and how the incentive to innovate changes with market structure & mergers
will present "Welfare Gains from Product and Process Innovations" (with
@ShokiKusaka
@jeffyjqiu
@tweetanhtl
) at SI 2024 NBER Innovation meeting on July 17 at 11:30am
...and Kevin Bryan
@Afinetheorem
will be discussing—thanks!
So I'm presenting my 2020 paper (on mergers & innovation) with Kosuke
@Chanman_ECON
one last time at the International Finance Corporation's conference in DC next Monday at 2:15pm.
You can register here and watch online, too!
IIOC: This Friday at 5:30pm, I'll be discussing job-market paper by Quan Le
@qlquanle
(PhD Princeton-->HBS), one of the Rising Stars
@IIOC_IO
!
And then on Saturday at 2:00pm...
Thanks to Adam Jaffe's swift handling at ResPol, our "first TDA paper in economics" is now forthcoming, with data & code available here!
*TDA = Topological Data Analysis. With Python 3.8 or later, you can construct a Mapper graph in 10 minutes or less.
World Congress: I was waiting for my session chair's e-mail to coordinate on prerecording. It never came. So I checked who the lazy chair was ...and rushed to write an e-mail.
Teaching evaluation for my first (and last) MBA class at UBC was much better than any of my undergraduate or PhD courses at Yale—I was expecting the opposite
Nice application of "Approximate Dynamic Programming" tricks!
—to solve a massive resource allocation problem, with application to NIH research funding
Super-nice paper by M. Igami "AI as Structural Estimation.." appeared in Econometrics Journal recently. Econometrics Journal now enjoys a great stream of papers, thank to revamping editorial policy (shorter papers, "minor revision" only policy, very fast turn around times).
Thanks to Adam Jaffe's swift handling at ResPol, our "first TDA paper in economics" is now forthcoming, with data & code available here!
*TDA = Topological Data Analysis. With Python 3.8 or later, you can construct a Mapper graph in 10 minutes or less.
My first talk in Mexico
@ITAM_mx
was virtual but very interactive and fun. I presented my topology-technology paper (pictured). Thanks for the invitation, Francisco
@fgarridoGU
!
AMES2022 in Tokyo/Zoom: Invited IO talks by
@PhilHaile
&
@vaguirregabiria
will be 8:00PM-9:30PM EDT *Sunday* (= 9:00AM-10:30AM Monday in Japan). Good time to update your beliefs about demand estimation & dynamics!
@florianederer
I do like the Hard Drive paper by
@MitsuruIgami
and
@Chanman_ECON
- another great case study! I thought this was the "important" graph. Captures a similar mechanism/tradeoff.
Data and code for "Privatization and Productivity in China" (RAND 2021, with Chen,
@masa1123
, & Xiao) is now available. You can estimate Gandhi-Navarro-Rivers nonparametric production function—with endogenous change in firm type—in <15 minutes!
Japanese economists are obsessed with "sunset" industries. Not a surprise, given our shrinking population. The irony is that these papers' introductions have to emphasize "the importance of ...declining industries" and "the big share in GDP of ...shrinking industries"
The main takeaway is that ANN (either single-layered = "shallow" or multi-layered = "deep") is useful for approximation when the object is a complicated unknown function of high-dimensional covariates.
First time being included in a Finance session at
#NASMES2022
. Take a closer look, and 2 of the 4 papers are in IO. No, we are not studying bank mergers. The session title is "Economic Applications of Machine Learning" = anything goes.
I don't know how many people are coming to Miami, but in case you are, please come & enjoy our "Economic Applications of Machine Learning" session—11:00AM Friday at AGB 531.
First time being included in a Finance session at
#NASMES2022
. Take a closer look, and 2 of the 4 papers are in IO. No, we are not studying bank mergers. The session title is "Economic Applications of Machine Learning" = anything goes.
@takuo_sugaya
@RevEconStudies
...and here's the replication package (data & code in MATLAB, plus a 6-page README with detailed instructions) for your PhD students!
This Friday at 5pm, come & see rising stars with brand new JMPs on:
(1) pharmaceutical bundling,
(2) mergers, offshoring, & jobs,
(3) common ownership + innovation.
"Rising Stars: Antitrust II" (Caspian Room, 3F)
@IIOC_IO
Don't have institutional access to the journal?
Please use the open-access link from my personal webpage, which is item
#2
on the PUBLICATIONS section.
This year's IIOC went virtual (sort of) and here's my session on innovation, with Jihong Lee (Seoul National), Andrew Toole (USPTO), Priyanka Sharma (Illinois Inst. of Tech.), and Lucy Xiaolu Wang (Cornell).
Some curation might help:
Chen & Shen (1998) and Chen & White (1999) do artificial neural network (ANN) sieve estimation of conditional mean function in time series models, and compared it against splines and wavelets.
A Network/Topology conference program: Somehow our paper on firms and patents ended up in "Ecosystems and biology" session. (Well, yes, we are all biological organisms.)
We are proud to announce that the preliminary programme of
#TopoNets20
at
@netsci2020
is out!
2 days
6 invited talks
18 contributed talks
Check out the full list here:
Ai & Chen (2003, 2007) use ANN to approximate unknown functions of endogenous variables.
Chen, Hong, & Shum (2007) do sieve MLE for model comparison.
Chen & Ludvigson (2009) use ANN to estimate unknown habit function.
Chen (2007) reviews all of them.
Had a nice dinner over Zoom with my U Tokyo master program classmates
@MitsuruIgami_JP
and
@Chanman_ECON
. I was lucky that I was a part of a hard working and friendly cohort at the early stage of my career.
"IO: Innovation & Patents" (ESWC session 342) - please join our live Q&A this afternoon at 2PM (New York Time) with
@ganglmair
@DeepakHegde_EFL
& Daniel Gross (Duke)!
Updated Zoom link
The
@econometricsoc
World Congress site is online. Check for 700 hours of recorded content. Live Q&A sessions start next week. I'll be doing a Q&A with
@DeepakHegde_EFL
,
@MitsuruIgami
, and Dan Gross next Wednesday, 2 p.m. ET. Tune in!
If you don't have access to the journal, please use the free-access link on my personal website, at the top of the publication section (sorry, such links can't be tweeted).
@comp_simon
Python code, data, & working paper version are here:
*With colorful figures, WP version is easier to read (UChicagoPress journals handle black-&-white only)
...and getting accepted! Its 8-minute slot looks challenging but doable, because I won't have to explain the "method" part to this audience. (BTW less-than-10-minute slots feel about right for most econ papers outside IO.)
Updated our TDA-of-patent-portfolio paper to ver. 2:
Changes from ver. 1 (Aug 2019): (1) all graphs & tables have been overhauled, and (2) a systematic comparison with Jaffe's (1989) method is added.
@aoziparis
Wonderful... I'm glad! The main purpose of this paper is to contextualize everything in our plain econ language, so that our students (and ourselves) don't get lost in the elusive quest for "something new out there".
@florianederer
@eizenb_alon
@Chanman_ECON
Goettler & Gordon ('11) feature forward-looking consumers playing against forward-looking duopolists. It's a cool model and I like it, but I don't think it has a unique equilibrium (i.e., not really identified/estimable). So we don't really know what their true conclusion is.
@eizenb_alon
This looks like a good reading list for me! Because it coincides with my interests so much that, in fact, I've already read at least one or two item in each section.
@eizenb_alon
Thanks, Alon! The average turnaround at the IJIO seems to have improved a lot over the past few years, so please do seriously consider sending papers when your discount factor pushes the RAND etc. out of range!
@levysamuelisaac
Don't have institutional access to the journal?
Please use the open-access link from my personal webpage, which is item
#2
on the PUBLICATIONS section.
The purpose of writing short should not be "to hide weaknesses"—I wrote "the longer you write..." as a counter-weight against some students' belief that writing a JMP is an "arms race" to write longer.
I realized my "Live Q&A" at ESWC2020 is 3:00AM in Japan. If you are located in Asia/Pacific, please go ahead and watch this (rehearsal) video, and I'll respond to your questions posted in the comments section.
@profeski
Right! Political economy seems the right/best angle to say such industries matter (instead of the big shrinking fraction of GDP etc.). And it makes sense that trade economists write a good paper on that topic! Thanks for sharing this.