Machines Learn the Chicago School

Modeling Neighborhood Dynamics with Spatial Markov Chains

Elijah Knaap, Ph.D | @knaaptime
Predicting Neighborhood Change Using Big Data and Machine Learning

Jacobellis v. Ohio (1964)

I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description “hard-core pornography”; and perhaps I could never succeed in intelligibly doing so. But I know it when I see it, and the motion picture involved in this case is not that.

—Justice Potter Stewart, concurring opinion in Jacobellis v. Ohio 378 U.S. 184 (1964), regarding possible obscenity in The Lovers.

Gentrification is fuzzy

Quantifying the “I know it when I see it” doctrine

  • socioeconomics
  • race
  • housing markets
  • commercial amenities
  • sights and sounds

Chicago School

Neighborhoods as “Natural Areas”

The Chicago School holds that a natural area involves:

  1. a geographic area physically distinguishable from other adjacent areas;
  2. a population with unique social, demographic, or ethnic composition;
  3. a social system with rules, norms, and regularly recurring patterns of social interaction that function as mechanisms of social control; and
  4. aggregate emergent behaviors or ways of life that distinguish the area from others around it."

  • (spatially-constrained clustering)

Social Area Analysis

A social area is not considered a neighborhood in the same sense as the natural area is. A social area consists of all those urban subareas with similar combinations of residents’ social characteristics on status, familism, and ethnicity. The subareas need not be contiguous. Their similarity arises from the social similarity, not the physical proximity of their residents.

Social area investigations are rigorous in the use of complex statistical techniques. Typically, however, only demographic or housing variables are studied. Direct observations of social interactions are missing from these investigations.

(Schwirian 1983)

Geodemographics

Temporal Geogemographics

Invasion and Succession

The Chicago sociologists recognized change in local population composition as a major mechanism by which natural areas change. The terms “invasion” and “succession,” taken from plant and animal ecology, were used to describe the processes of neighborhood population alteration.

  • Neighborhood change is a spatial process
  • New social groups invade nearby areas until the they ‘tip’, and the new Type succeeds as the predominant social area

Modeling Neighborhood Dynamics as Temporal Geodemographics

  • How often does Type 1 become Type 2 ?
  • How often does Type 1 become Type 4
    • If most of its neighbors are already Type 4 ?
    • If most of its neighbors are still Type 1 ?

Data

  • Census LODES/LEHD
  • block-level, annual resolution
  • limited demographic data since 2010
  • race, ethnicity, education, wages

Model

Cluster Model

  • gaussian mixture model
  • k varies from 2-7
  • race/ethnicity, education, earnings
  • z-standardized within each time period

Spatial Markov Chain

  • KNN weights matrix
  • Modal spatial lag

DC Results

(poorly) Describing Types

  • Cluster 0: white with some diversity, highest education, highest earning,
  • Cluster 1: black/white, lower education, lower income
  • Cluster 2: white/black, high education, high income
  • Cluster 3: white segregated, high education, high income
  • Cluster 4: racially diverse, med education, med income
  • Cluster 5: white/asian, high education high income
  • Cluster 6: black/racially diverse, lower education, lower income

Neighborhood Transitions

  • Type 6 has a 68% chance of remaining Type 6
  • It has less than a 1% chance of becoming Type 0 or Type 3 – the two neighborhood types with the smallest shares of minority residents.
  • There is an 8.7% chance of transitioning into Type 4 (racially diverse, lower education, lower income)
  • Once there, Type 4 has a high probability of transitioning into many different neighborhood types

Spatially-conditioned Neighborhood Transitions

  • Type 6 has 5% chance of transitioning into Type 5 (highest earnings and education) without considering spatial effects (gentrification)
  • Type 6 has a 21% change of transitioning into Type 5 if its modal neighbor is Type 5

https://github.com/spatialucr/geosnap

Conclusions

  • Every transition in every metro area shows significant spatial dependence
    • gentrification is a spatial process!
  • Modern spatial data science validates the Chicago School model
  • Stability is the most common “transition”
  • Neighborhood transitions have important path dependencies that conform to classic sociological theories about social mixing and racial avoidance

Thanks

  • Sergio Rey
  • Levi John Wolf
  • #1733705 Neighborhoods in Space-Time Contexts

  • #1831615 RIDIR: Scalable Geospatial Analytics for Social Science Research

home.knaaptime.com/berkeley