Profile picture
, 14 tweets, 4 min read
My Authors
Read all threads
1/n The Great Awokening: Twitter edition

Credits to the folks at storywrangling.org for compiling and sharing these data.
2/n Next step: see how these trends relate to those observed in left-leaning news outlets (I'll have to check, but, on first glance, it looks like the media's awokening trends might actually precede those depicted here). More to come.
3/n Speaking of 'folks'
4/n Interesting that 'white supremacy' was on the rise even before Trump announced his candidacy (June 2015)
5/n Been running some granger tests on the yearly data. So far, most of the relationships I've observed appear to run from news media (e.g. NYT) -> twitter (i.e. the previous year's NYT term frequency predicts its twitter frequency in the following year, but not the inverse).
6/n Obviously, with just 11 data points, caution is warranted. I will try running the tests on the monthly NYT I grabbed from LexisNexis a while back and report back.
7/n So I pulled out the older (only runs through June 2019) monthly-level NYT data that I gathered from LexisNexis (). What I find is that the relationship between Twitter and NYT can be bidirectional for some terms, and unidirectional for others.
8/n Here is an example of a (strong) relationship that appears to run from NYT->Twitter
9/n By 'appears' I mean that NYT 'Racism' significantly predicts Twitter 'Racism', but not the inverse
10/n And here's an example of relationship that appears to run from Twitter->NYT
11/n I'm hoping to use ProQuest to build a newer monthly-level NYT database. But as far as I can observe in the present data, Twitter and the NYT (and/or other news outlets) essentially feed off each other.
12/n Another relationship that appears to run from Twitter->NYT
13/n Fucking knew it (I'm sure we all did)
14/n A lot of problematic hierarchies and systems of power starting coming into relief in 2014.
Missing some Tweet in this thread? You can try to force a refresh.

Keep Current with Zach Goldberg

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

More from @ZachG932 see all

Profile picture
Zach Goldberg
@ZachG932
1/n Interestingly, an almost identical proportion of respondents (55%) gave this response in a 1988 AP poll.

The difference today is in the distribution of responses across different subgroups.
2/n Specifically (I could graph but I'm lazy), ~59% of democrats and 55% of republicans selected 'yes' in 1988 as compared to 82% and 30%, respectively, in the 2020 WSJ poll.
3/n In other words, the parties have become more polarized/better sorted on race (which is an important reason why the awokening has been largely limited to Dems).
Read 6 tweets
Profile picture
Zach Goldberg
@ZachG932
1/n Using the Washington Post Shooting Database (2015-2020), I entered the names of all 'unarmed' black and white police shooting victims into ProQuest and created a tally of search results.
2/n What I find is that news media content covering black victims is about 9x greater than that of white victims (whether one compares the medians or the means).
3/n To account for the skewed distribution of the data, I used a quintile/median regression to regress the # of results onto the incident/victim-relevant variables included in the Washington Post dataset. None of these variables explain the difference.
Read 11 tweets
Profile picture
Zach Goldberg
@ZachG932
@Inductivist Okay. I can control for 5. I don't have variables for 3 (just population size). 1 is already controlled for in the sense that virtually all reports of force occurred during officer-initiated contact.
@Inductivist I can't control for 2 (there's one variable asking if reason for being stopped was DUI suspicion, but none that ask if respondent was actually under the influence)
@Inductivist You're right. Force is less likely to be used when it's a single officer relative to two or more. The inclusion of this variable further narrows the BW gap from (0.008->0.005)
Read 4 tweets

Embed code for your website

×
Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!