Berlin school sorting

Schools sorting

• Schelling logic: moderate micro homophily leads to radical segregation
• Is segregation due to moderate homophily?

Confounding factors:

• geography: school segretation given residentital segregation
• school quality: identity used as a proxy?

Setting:

Berlin school system: students assigned based on small zones

Can request a change: select up to three alterantives, with explanation; approved or not

Data:

• school demographics
• for Tempelhof-Schoenberg data on individual requests from students for ca 10 years as well as approvals
• school quality data

Do not have individual group/identity data

Change over time

Requests

Basic exit requests

Given distance and demographics

To, from

Causal analysis? RDD results

RDD

Causal analysis? RDD results

Note: sorting works in our favor here

By subgroup

By subgroup

Identity or quality? 1

Note that in places with more than

Requests given dyad demographies

Identity or quality? 2

Requests given dyad scores

Identity or quality? 3

• Demography is uncorrelated with school quality scores
• Controlling for quality has almost no affect on inferences

[Add RDD graph by sending and receiving quality score]

Directions:

• Best use of best guesses of student identity:

  • Bounding approach
  • EI?
  • Literal interpretation

• Best use of decision data

• Best use of quality data

• Focus on effects or on a model