Thread by @SethGershenson

so happy this is out. always a pleasure to work with @MichaelSHayes & @DrJingLiu. quick thread 1/n
sciencedirect.com/science/articl…

the main contribution here is extending/replicating a clever @J_HumanResource paper by @ajmceachin @JonValant Barret & Mills 2/n

jhr.uwpress.org/content/early/…

along with another contemporaneous paper by @mariabzhu & Ying Shi, all 3 papers use different datasets and slightly different methods to document substantial "intentional discrimination" 3/

edworkingpapers.com/ai21-383

intentional discrimination refers to racial differences in the punishments meted out for similar (or in this case the exact same) offenses. the term originates in an Obama era Dear Colleagues letter 4/

www2.ed.gov/about/offices/…

a common argument/justification is that racial disparities in suspension outcomes are "fair" because they're due to underlying racial differences in behavior; these papers show that that's not true... 5/

the intuitive identification strategy compares the disciplinary outcomes (suspension decisions on extensive and intensive margins) of students involved in the SAME EXACT multi-student infraction (a fight, cutting class, etc). 6/

In Louisiana, North Carolina, and a large urban district in California, all three papers find similar patterns: Black students are more likely to be suspended, and are suspended for more days, than white students involved in the EXACT SAME INCIDENT (on average) 7/

our main innovation is observing the office referrals that necessarily precede a suspension. this is important because (i) it allows us to study the extensive margin of whether an incident resulted in any suspensions and 8/

(ii) it allows us to control for students' prior disciplinary histories. here's the main results table (models control for incident fixed effects) 9/

the extensive margin is really important: Black students are 2 to 3 percentage points (2 times) more likely to be suspended than the white student they fought/cut class/misbehaved with. you miss this without using referrals to identify incidents. 10/

we also just explore patterns in referrals more generally. referrals are understudied since many datasets don't record them well (or at all). here, we show they follow similar patterns as suspensions (which makes sense) 11/

for instance, another "justification" of racial disparities in suspensions is that it's due to "sorting into schools"... again, we show that's not so. figure 1 decomposes racial referral / suspension gaps into within and between school variation. lots of within-school var! 12/

the bigger descriptive point is that there's a racial disparity in referrals, but it's not a 1:1 mapping into the racial gap in suspensions. there's ALSO a racial gap in the conversion rate of referrals to suspensions. 13/

why? to tease our next paper, there's bias in the referral process (how and when teachers make referrals). but also at the adjudication stage (typically a principal's decision). the latter of course could be influenced by how teachers report 14/

our paper cites some qualitative studies on these issues. so what to do? as is often the case in matters of education policy, an obvious course of action is to read @Dr_ConstanceL 's CV. 15/ specifically, this, w/ @JOkonofua : science.org/doi/10.1126/sc…

some closing thoughts:

(i) I love love love JUE's new insights (15 double space page limit & 1 round of R+R) option for shorter papers. several journals have now adopted this format and I think 90% of papers would benefit from this sort of editing / tightening 16/

(ii) like I said, @DrJingLiu, @MichaelSHayes , & I have two more projects in the works that really dig into the referral data and specifically the role that teachers play there (they're the modal source of referral). I know others are working in this space as well. stay tuned 17/

(iii) the evidence that suspensions (exclusionary discipline) is harmful and that students of color (esp Black students) are disproportionately subjected to this type of punishment is pretty clear. research and policy should turn to remedying this. 18/

adjudication is a part of it... but it all starts referrals, classroom management, and student-teacher relationships. again, this is why @JOkonofua & @Dr_ConstanceL 's work is so important. 19/

finally, even referrals that don't convert to a suspension harm student-teacher trust/relationships. 20/20

PS: part of why I love working with Mike and Jing is that it's a masterclass in comparative advantage & gains from specialization: Jing knows the data inside and out, Mike writes the code, and I forget when our meetings are. we each play our part

PPS: shout out to @NeumarkDN for being a helpful and demanding editor who pushed us to be clear, precise, and organized while staying in the 15 page limit. again, this JUE Insight option is a fantastic outlet.