Economists who are interested in
#DataScience
: I'm teaching a master's-level course this semester entitled "Data Science for Economists." Feel free to follow along---all materials are publicly available on GitHub here:
If there are any PhD or pre-PhD students out there who are interested in remotely taking my structural modeling course this semester, feel free to DM me and I will be happy to add you to the course (up to a capacity constraint). I have already added 2.
I'm pleased to announce that I've been awarded tenure and promotion to associate professor!
Thanks to everyone who mentored and supported me in any way. Now it's my turn to pay it forward!
I detest teaching on Zoom because the audience barely participates at all.
I love holding department meetings on Zoom because the audience barely participates at all.
The good news: I recently had a paper accepted by JHR!
The bad news: I had to convert from LaTeX to Word
Here is a checklist for my process, in case anyone out there needs to do the same.
For anyone who wants avg income by US ZIP code, here is some code:
library(tidyverse)
zipdata <- read_csv("") %>%
mutate(avg.salary = A00200/N00200) %>%
select(STATEFIPS,STATE,ZIPCODE,avg.salary)
save(zipdata,file="IncomeByZIP.rda")
If anyone out there is wondering about structural models, I've just posted this slide deck that might be helpful.
If you feel I have mischaracterized your tribe, feel free to submit a pull request or issue! 😆
I'm using
#Rstats
to teach introductory econometrics this semester. Most class periods, my students will be doing a lab in small groups. I'm posting the labs here in case they are of use to anyone else:
Yesterday was a big day: The paper I wrote for my PhD 2nd year field paper nearly 10 years ago (and now coauthored with
@docashworth
,
@VJosephHotz
and Arnaud Maurel) got accepted at Journal of Labor Economics!
I'm pretty much feeling like
#GronkSpike
I've never had major problems with depression, but these past two weeks (since I found out the kids aren't going back to school) have mentally been absolutely brutal. Toughest time since the job market 5½ years ago.
Stay positive, everybody!
Big congrats to
@jenniferdoleac
and
@benconomics
for winning the H. Gregg Lewis Prize for best paper published in Journal of Labor Economics over the past two years!
Hey
#EconTwitter
:
OpenAI's embeddings can be used for fuzzy string matching! And best of all—API calls are super cheap (1¢ for every 25,000 characters).
Here's a GitHub gist that will produce the following image in <75 lines and basically for free:
Finally, if you're interested in using the NLSY surveys, I highly recommend checking out our replication kit (Careful: the ZIP is nearly 2 GB). We put a *ton* of work into making the two NLSYs compatible, so please don't reinvent the wheel!
New paper out, with Gizem Kosar and Wilbert van der Klaauw: We estimate migration preferences using a choice experiment.
People have strong preferences for living near family + those with shared norms/values. Psychological moving costs are large!
I'm thrilled that this paper has just been accepted at Journal of Econometrics!
The editor and referees made it a better paper.
I'm glad to have worked with
@gizemkosar
and Wilbert van der Klaauw---two stellar economists!
Replication kit available here:
New paper out, with Gizem Kosar and Wilbert van der Klaauw: We estimate migration preferences using a choice experiment.
People have strong preferences for living near family + those with shared norms/values. Psychological moving costs are large!
Hey
#EconTwitter
:
OpenAI's embeddings can be used for fuzzy string matching! And best of all—API calls are super cheap (1¢ for every 25,000 characters).
Here's a GitHub gist that will produce the following image in <75 lines and basically for free:
Our colleague
@tyleransom
was recently granted the
@OUArtsSciences
Irene Rothbaum Award, which is awarded each year to the outstanding assistant professor out of more than 100 in the college. Congratulations, Tyler!
Mind-blowing paper about how the estimated coefficients of a logit model (or odds ratios) are not identified separately from the variance of unobservables.
Since nobody asked, here's my 15-step process for doing referee reports:
1. when I first accept a request, I add the line to my CV. I also create a folder for the journal and a sub-folder for the month-year that I perform the review
2. I also add it as a card on my Trello board
@amandayagan
@arpitrage
@maxkasy
Also if you want a quick cheat sheet for how to code basic econometrics estimators in R, I made a cheat sheet that I distribute to students (I will be updating it in the coming weeks based on twitter feedback):
Life with
#autism
: last night I came home from a conference and my flight got in 30 minutes early. When I showed up at my house, my son was extremely upset and told me to go back to the airport, that I wasn’t welcome because I wasn’t following the schedule. 🤣
My brother Todd would have been 38 today. My family remembers him by grabbing slurpees from 7-11 since that’s what we did when he was alive. It’s a nice tradition. RIP bro
My paper with
@docashworth
@VJosephHotz
and Arnaud Maurel is now forthcoming at JOLE. We generalize the Mincer model in four ways to estimate the wage returns to various types of work experience using the NLSY79 and the NLSY97.
This surprised me, but twitter economists actually have more research experience than non-twitter economists (and more citations, higher h-indices, etc.).
Hey
#EconTwitter
, my master's student Rachel Thatcher scraped data from
@repec_org
on economists' research records and which ones have a twitter profile (registered with RePEc). Her Q: "Do twitter followers predict citations?" Answer: Yes (but beware reverse causation) 1/N
Recent realization:
Academic writing boils down to repeatedly doing referee reports on your own manuscript
Late-stage PhD advising boils down to repeatedly doing referee reports on your student's manuscript
Hey
@Andrew___Baker
and other JMCs, back when I was on the market 4 years ago I put together some code to automate the process of collecting listings (and figuring out which listings were newly posted). You might find this helpful:
This was truly one of the “crucible” moments of my life. On top of striking out on the market, my oldest son had just gotten diagnosed as autistic, no one in our house was sleeping well, and my wife was pregnant with our other son.
But I got through it, and so can you!
Our paper "Divergent: The Time Path of Legacy and Athlete Admissions at Harvard" is now forthcoming at
@J_HumanResource
. We show that legacies and athletes (LA’s) saw their Harvard admission bonus *more than double* from 1996-2013 (Classes of 2000-2017).
Hey Job Market Candidates, my department
@OUeconomics
is hiring! There is no JOE ad up yet, but it should be posted fairly soon. Hiring is for "macro" and/or "big data" (both very broadly defined).
What's worse than code that won't run? Code that runs, but doesn't give the correct results.
This means you're past the conformability bugs and into the more insidious bugs that are harder to find.
While I was on vacation last week I came across a venue called the "R Bar." Sadly, no one wanted to debate dplyr vs. data.table or discuss code piping with me.
Today I'm grateful for a TT job at a good university. I have first-hand experience with how hard landing an academic job can be, and how much noise there is in the process. I am grateful to be able to work with good colleagues and bright and capable students.
#GiveThanks
As I was grading my final exam for my upper-level undergraduate Economics of Education course, I decided to check how ChatGPT would do on the exam.
Answer: ChatGPT scored a 77%, which was just slightly below the class average.
I'm reading *Working Backwards*, a book about Amazon culture. I wonder why in academia we don't structure our research seminars in the same way AMZN does its meetings:
1. Everyone reads summary of paper in first 20 mins
2. Discussion ensues
3. Presenter uses slides for backup
The
@causalinf
bump quantified:
49 visits to my website on Sat after this tweet
8.22 visits avg across pvs 9 Saturdays
+496% treatment effect
Thanks, Scott!
OK folks, we're all going to have to start following Valerie, because I started following her last week, and the same exact thing happened to me today!
2 PAPERS ACCEPTED ON THE SAME DAY!!! What is even happening to me?!? Is this real life? 🤯🍾🍾
"Nevertheless She Persisted? Gender Peer Effects in Doctoral STEM Programs" (with Bruce Weinberg) is forthcoming at JOLE
New working paper from me, Peter Arcidiacono, Esteban Aucejo, and Arnaud Maurel!
The paper looks at how information frictions affect college dropout and post-college labor market sorting.
This IZA DP is a revision of an NBER WP from many years ago
If anyone wants a broad overview of Julia for folks with mild stats backgrounds, I've just posted some slides that you might find helpful. (These will feature as reference material for my PhD-level structural modeling course this fall.)
Migration researchers (e.g. myself,
@Richard_Florida
, and others) are often interested in how "rooted" the population is.
Today, I propose a new measure of "rootedness": a person's roots are where his/her cell phone number area code is.
Last week, I taught time series in my econometrics class. Yesterday I had the following exchange:
Student: Why does covariance stationarity mean that we can make causal inferences?
Me: I'm not sure I know the reason why; it's the black magic of time series.
We just completed our latest carpentry project: a “long-run Covid adjustment” home office desk!
#econwoodworkingtwitter
(I say “we” loosely—it was 85% my wife and 15% me. But she’s still giving me coauthorship.)
It is fascinating to me that, every semester when I teach intro econometrics, I get 1 or 2 students who want to research about happiness. It's the most consistently popular topic.
Ditto. I passed first year by the skin of my teeth. Keep in mind that the correlation between first-year performance and research productivity is low (in the population).
Btw, I struggled a lot my first year in classes. I ended up doing well only once I got into field courses and then my own research. So if you are also struggling in first year and not encouraged by your first year profs but excited about research, hang in there.
I put together a slide deck for my master's class about how to use GPT-4 at each stage of the research process. Hopefully someone else also finds this helpful!
I just realized it's been 5 years this weekend since I graduated with my PhD!
It feels like it's been a lot longer than that. Maybe because the tenure track + raising two special needs children has made me age in dog years? My eye bags certainly support that hypothesis.
I spent a non-trivial amount of time this week fighting with Stata's "file write" command to produce LaTeX files. It turns out that inserting ` and $ into said files can be a massive pain because of how Stata interprets these special characters.
Here's what I found to work:
It’s a really tough time. But the important thing is not to over-analyze it. Things that helped me:
1. Working on other non-JMP projects
2. Getting counseling
3. Continuing to apply for jobs as they get posted.
@paulgp
I'll be teaching my Data Science master's level course in the spring and will be encouraging them use copilot/generative AI, no holds barred. But I spend about a week and a half introducing computing tools, so it fits right into my pre-existing syllabus.
I'm going to try something new this semester. I'd like to mentor one-on-one a new assistant professor in an R1 econ (or similar) department.
If you are a new asst prof who would be interested in this, there is an application form here. Details below:
Anyone have a sense for where AER:Insights rejected papers are getting published? Are other top journals accepting shorter articles, or are authors revising these to make them longer? I suspect there are a ton of these out there.
Had a tornado blow by last night within 1000 feet of our house; fortunately we didn't get any damage and no reported deaths. Feeling very blessed and grateful.
Hygienist: You haven't had your teeth cleaned in [checks computer] 2 years? We recommend every 6 months, you know.
Also Hygienist: Wow, your teeth look pretty great
Me: Shouldn't you be congratulating me for solving the optimal stopping problem?
Today, I
#GiveThanks
for:
- PhD advisors who patiently taught (and still continue to) teach me
- Coauthors who forgive my many coding errors
- Exceptional undergraduate advisors that encouraged me to pursue a PhD in the first place
Cool thing I found out how to do in R: Download College Scorecard data directly from NCES!
library(rscorecard)
key <- 'your-key'
sc_key(key)
data <- sc_init() %>%
sc_select_(c('unitid','instnm','iclevel','ugds')) %>%
sc_year(2017) %>%
sc_get()
Today I spent ~3 hours trying to compare Matlab regression output with Stata regression output. I know that sounds mind-numbing (it was), but Stata is a super useful tool for debugging code written in other languages.
I just had an article accepted! 🎉
... and let me tell you how good it felt to be able to just click "upload ZIP folder" to the journal's data repository with no further headache.
Can confirm that it's best practice to start on day 1 as Florian recommends.
Let me provide a few comments from the other side on this thread. If you start thinking about reproducibility of your work only the day the paper gets accepted, it can get rough. Start on day 1! We provide workshops! Let’s take a closer look: 1/n
Behavioral economics at work: If Stata set the default graphics background to be white instead of light blue (e.g. by making “graphregion(color(white))” the default), the world would be a much better place.
#nudge
I truly appreciate all of the support you've provided me here! I feel much better after a day full of chores and playing with the kids.
I know there are more of you out there who feel the same as me, so let's support each other!
I've never had major problems with depression, but these past two weeks (since I found out the kids aren't going back to school) have mentally been absolutely brutal. Toughest time since the job market 5½ years ago.
Stay positive, everybody!
Hey
#EconTwitter
, what is the possibility of moving PhD programs?
I know a first-year student at a low-ranked Econ PhD program who is finding out they're not as enamored with Econ as previously thought and is thinking about applying to policy PhD programs this fall.