Home

Evictions project

Notes from first meeting, 29 August 2014, UAL Lab, room 401A.

Present: Paul, Aksel Olsen, Geoff Boeing, Pedro Peterson, Sam Maurer, Akshaj Vij, Tim Brathwaite, Sam Blanchard

Purpose and background

• find early indicators of displacement in areas with significant transit service
• clear policy import: If we make investments, upzone in core urban areas around transit, are we inadvertently pricing out low income residents?
• Karen Chapple had a bunch of students working on it, talking to folks in the community
• Paul's part of the grant: prime urbansim for sensitivity to displacement.

Three main "branches" of this effort, one for each of three models--two sides of the household location dynamics (decision to leave, choice of housing) and the developer model.

Models

A: move out income

*age of householders determine rates of relocation model doesn't take into account rent burden/cost burden/tenure

• paul: change from rate model to binary logit model, more RHS vars, conditioned on tenure.
• Tim: how do we measure displacement?
• --paul: folks that are priced out from their homes
• aksel: how do we observe displacement?
• what is people's current, observed cost burden?
• Can we calibrate a model to generate a pattern

B: move in

loc choice model for households

• Panel data is great, we don't have it. Do we introduce a more explicit budget constraint? Costs are in there, but there is just lower utility to higher costs?
• Paul: We need bldg-level unit size distributions to better match types of households to specific buildings. Census has block (group) level unit size distributions, but we need buildings.
• Aksel: much of bay area is homogenous within the block (group) level. Not so much risk in inferring back.
• Paul: We also need to impute tenure at bldg level.
• Tim:Is there research on the effect on unit size in a building to overall building price?
• Paul: More fancy hedonics could be imagined. Robust regressions, geographically weighted regressions
• Michael Smith-Heimer said landlords are important players and in effect choose tenants. Interact logit probability of household choosing and landlord

C: developer model

• cleaner version than i PBA.
• add aff hsg constraints in the hopper (??)
• density offsets for aff housing
• Tim: what does the spot developer model do?
• Paul: parcel is selected (??? sorry didn't catch this; tuned out)

Other business

• Sam: Pulled together dataquick/corelogic. 10+yrs of data. Centroids, not geometries. Corelogic has parcel genealogy though.

Tasks

• get any aff hsg restrictions--data set from http://www.chpc.net/about/staffboard.html#JamesPappas
• Sam Blanchard is doing data assembly
  • We have craigslist, nationally.

Tim

• imputation of unit size
• imputation of tenure status (was this also Tim?)

Sam M:

• move-out model

Aksel:

• move-in--main task: add budget constraints
• organization:
  • wiki or some such--get everyone on github.com/ual

Pedro:

• aff hsg modeling