Drew Johnston
@drew_m_johnston
Followers
731
Following
677
Media
6
Statuses
75
Economics PhD candidate @harvard . I like computers, cities, and social influence.
Joined November 2018
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#เมเจอร์xหลิงจังจอมแก่น
• 566186 Tweets
LINGLING MAJOR FANDOM
• 542860 Tweets
SAY SOMETHING
• 153480 Tweets
#انقاذ_النصر_مطلب_عشاقه
• 116151 Tweets
線状降水帯
• 42884 Tweets
アニサマ
• 35003 Tweets
#TPOPStagexนุนิวฤดูของเธอ
• 32416 Tweets
#HAPPYGYUVINDAY
• 31097 Tweets
台風どこ
• 27178 Tweets
土砂降り
• 25790 Tweets
Your Season On Stage
• 25047 Tweets
KND x PHUWIN
• 24571 Tweets
#온우주가말해_규빈아_생일축하해
• 23270 Tweets
#ابعاد_لودي_مطلبنا
• 21620 Tweets
#THE夜会
• 20292 Tweets
OUR AUGUST SUNSHINE KIM GYUVIN
• 15873 Tweets
Jabar
• 15532 Tweets
コロンブス
• 13541 Tweets
ハチャウマ
• 12418 Tweets
#HAPPYRIHITODAY
• 11240 Tweets
Pinned Tweet
I'm excited to announce the release of the zip code Social Connectedness Index!
It captures the rate of Facebook friendships between zip codes.
This data set allows for much more granular uses than before. A 🧵 of examples:
8
124
583
Dissertation defended! I'm excited to announce I'll be coming on as a postdoc at Meta to keep working on a few projects I'm pretty excited about--stay tuned!
9
0
169
But there are certainly more stories left to be told using this data, so download it for yourself and play around in your favorite statistical software.
A number of other granularities (US and international) are available if zip codes aren't what you need.
1
0
33
There's a ton more in here. It's possible to see that friendships tend to coincide with public transit lines (), and possible to see that there's a ton of variation in how clustered areas' social networks are.
1
2
25
There are a lot of micro stories in the zip code social connectedness data--for instance, the areas in the Bay Area most connected to 08540 (Princeton NJ) are home to research universities.
(Check my pinned thread for a link to the data!)
1
0
25
It's possible to dive more deeply than before into the connections of specific neighborhoods. For instance, you can see that Flushing (home of the best Chinese food in NYC) is closely connected to NYC's other Chinatowns, in Brooklyn and Manhattan.
1
0
21
It's possible to see how sharply connections fall along class lines as well--compare the social networks of the neighboring zips 11218 (the ritzy Upper East Side) and 10029 (East Harlem)
...this one is so stark that the NYT wrote an article about it ()
1
2
15
@MelissaLDell
This is amazing and I wish this had been around five years ago! I spent a lot of time doing this by hand for my first RA job :)
1
0
14
You can also dive deep into the connections between neighborhoods ACROSS cities--for instance, you can map connections between Flushing and the Boston metro, and see that a number of areas with large Chinese populations pop out.
1
0
9
I'm very happy to see the culmination of our team's research out today! Check out the data visualization in the NYT as well:
1
0
8
Huge thanks to
@michaelcbailey
,
@stroebel_econ
,
@DomRussel
, Theresa Kuchler, Pat Farrell, and Alex Pompe for making this possible.
1
0
7
Let us know if you find anything interesting using this data!
2
1
7
We're also making aggregated data on social capital at the HS/college/zip code/county level here:
1
0
3
@avi_collis
@stroebel_econ
@michaelcbailey
@DomRussel
Looking at how the results generalize to other migrant groups/countries would definitely be interesting! With earlier migrants we don't observe migration time though, which rules out some of the analyses. Other countries would be cool tool, though some inst. knowledge is needed!
0
0
2
@stephkestelman
@RFisman
@Harvard
@Princeton
@Columbia
You would not BELIEVE how useful sword fighting is in the world of econ research
0
0
2
@PopulismUpdates
You might be right--this website says 18 for local government positions, but doesn't seem to be mentioned in the charter.
0
0
2
@paula_slspi
@CarolinPflueger
@stroebel_econ
@Ogoun
I think just because this project is old! If I remember correctly, when we started working on this project ivreghdfe was new and did not let you cache residuals between regressions. This was 5 years ago though so my memory might be hazy. Big picture, feel free to use ivreghdfe--
1
0
1
@MarketUrbanism
Might make a difference for low-income renters if the fee is paid upfront vs factored into the rent over time
0
0
2
@stephkestelman
@RFisman
@Harvard
@Princeton
@Columbia
The NCAA fencing -> econ PhD pipeline is shockingly strong... the causality is a bit murky though :p
1
0
2
@paula_slspi
@CarolinPflueger
@stroebel_econ
@Ogoun
not sure why they give slightly different answers, but I think ivreghdfe is better maintained these days (we had to set a flag called "old" to get our legacy code to keep working with reghdfe)
1
0
2
@CarolinPflueger
@paula_slspi
@stroebel_econ
@Ogoun
We used this code in the Peer Effects in Product Adoption paper--you can check it out in context in our replication package (). I never really intended for it to be used more widely, but hopefully it solves your problem!
1
0
1
@ben_golub
Pricing in markets with more options seems wackier than in markets with fewer--I'd imagine it's a function of PredictIt charging you a 10% charge on gross (not net) profits in each market, so you can't short everything easily.
0
0
1
@aleszubajak
This is an updated version of the data, using connections from this month!
Not sure I'm allowed to share a user number.
0
0
1
Seems plausible that decking could be a better solution for BQE in Williamsburg, Prospect Expressway... Would add housing and still leave room for biking/walking infrastructure
1
0
1
0
0
1
@lara_putnam
@urbanclio
@PrdctblyIrrtnl
(We then re-scale the values to mask user counts, but this is the basic intuition)
1
0
1
This is awesome, hopefully a sign of a bigger shift in thinking. I wonder if funding will be available for decking over highways too... Boulevards are nice in a lot of cases, but in others decking could add land for housing/parks/whatever instead of a still-wide street...
1
0
1
@lara_putnam
@urbanclio
@PrdctblyIrrtnl
Between two places with 10 pop, there are 100 friendships if everyone is connected (the numerator). So the SCI if all friendships are formed is 100 / (10*10) = 1.
Between two places with 100 pop, there are 10k possible friendships, so the SCI if fully connected is 10k/(100*100)=1
1
0
1
@TBroekel
Sure thing! We don't currently have zip-level data for Europe but on the site where the data is distributed we also have something at the NUTS3 level.
0
0
1
@lara_putnam
@urbanclio
@PrdctblyIrrtnl
The denominator self-sci is n^2-n (since friendships ARE counted both ways for self sci)
0
0
1