<Civic/> Data Engineering

Building a system to track Milwaukee evictions.

Branden DuPont

Medical College of Wisconsin

 Datashare @ MCW IHE = Local IDS

Eviction as Cause, Not Just Consequence of Poverty

• Displacement
• Affects education outcomes for children

• Housing conditions (worsened by cycle of eviction)

• Among many others

Milwaukee Policymakers, Researchers, & Community Members Need Eviction Data

• Difficult to understand drivers of eviction and design prevention strategies
• Community unable to view a properties eviction record
• Conducting additional research prohibitive

Consider a Basic Question on Eviction

• "He'll evict you in a minute" - Milwaukee Journal Sentinel on Youssef "Joe" Berrada's Use of Eviction
• High eviction filing rate or does Berrada just own a large number of properties?

Data Needed To Construct an Eviction Filing Rate

• Court records on eviction - small claims case
• Address the eviction took place
• Taxkey information for that address
• The number of units at that property
• The primary land use of the property (multi-family or single-two family)

ETL: A Common Pattern

Dave Guarino, ETL for America

• Extract: getting data out of [many] systems where it is stored

• Transform: reformatting and reshaping the data in ways that make it usable. Creating meaningful variables (eviction filing rate)

• Load: putting the transformed data into another system, generally something where analyzing it or combining it with other data is easy for end-users

ETL: a hard ^#&@ing problem

Dave Guarino, ETL for America

• "Many of the problems government confronts with technology are fundamentally about data integration"
• Or, there is a reason no one was tracking evictions before

Foundational Work Comes First

Source:Monica Rogati’s  “The AI Hierarchy of Needs”

Getting Eviction Data

Got Eviction Data:

Deeply Nested JSON

• Data needs to be parsed into 8 tables
• Re-pull entire case for any slight change
• Retention rules: data needs to be stored and deduplicated

Get Eviction Location

• defendant in cases earliest address is used to represent an eviction location
• check to see if an address is updated during a case. Extract address from court events
• Remove PO boxes or out of country addresses if other defendant address is available

Match an Eviction Record to an Address

• Defendant address formatted exactly as in city address
• Extensive series of steps to clean addresses

Match Eviction Address To Taxkey

• 94.8% match rate compared to 93% match for Eviction Lab

Grab City Property Data

• Bring in data on primary land use, unit size, and ownership
• Easy and painless thanks to the Open Data Portal

Calculate Eviction Filing Rate

• How many evictions filed compared to the numbers of units
• To get a long term trend, calculate an average over three year period
• Visualize Berrada's properties.

Get Eviction Filing Rate

Imagine:

Repeating Every Single Step Again Next Week

• Desmond's Eviction Lab ~ 3 years old
• Difficult step of pulling, cleaning, and performing record-linkage

Track Milwaukee Evictions

• Civic Data Engineering = automates this complexity
• Provide useful information to the public, policymakers, and other researchers