Patterns of Residential Mobility

1
Patterns of Residential Mobility

• Using Cluster Analysis to Identify Different
  Types of Movers,
• Stayers, and Newcomers
• in the Making Connections Sites

2
High Rates of Family Mobility
3
About Half the MC Households Moved
4
Some Movers Stayed Nearby
5
Spatial Patterns of Mobility Vary
Des Moines
San Antonio
6
Implications for Early Childhood Initiatives?
7
Mobility and Neighborhood Change
8
Mobility and Neighborhood Change
9
Less Engagement Among Newcomers
10
Why Are Families Moving In and Out of the MC
NeighborhoodsCluster Analysis Hypotheses and
Methods
11
Why are families moving?

• Few direct survey questions re reasons for moving
  
• Milwaukee and Louisville Wave 2 survey
• Lots of information about possible push and pull
  factors
• Literature inventory of relevant factors
• Three illustrative survey questions
• Volunteer in neighborhood (attachment)
• Trouble w/housing expenses (instability)
• Housing tenure (home purchase)

12
Intro to Cluster Analysis

• Analytic technique to classify observations into
  groups based on variables of interest
• Measure distance between individual observations
  and the centroids of groups of observations
• Can use dichotomous and continuous variables
• No independent confirmation of cluster groupings

13
Methods

• Step 1 Create cluster predictions
• Guided by theory, previous research, population
  in question, variation in data
• Making Connections cluster predictions
• (following slides)

14
4 Separate Cluster Analysis Models

• 1. Out-movers with children Wave 1 and 2
• 2. Childless out-movers Wave 1
• 3. Stayers Wave 1 and 2
• 4. Newcomers Wave 2

15
Lots of variation among out-movers with children
(5 site pooled data)

• Household change
• 11 got married 13 separated
• 33 added a child 13 have fewer kids
• Employment change
• 12 became employed 13 lost their jobs
• Tenure change
• 18 became homeowners 11 shifted to rental
• Perception of neighborhood
• 63 think new neighborhood is safer
• 24 think its a better place to raise kids

16
Hypothesized clusters of Out-movers with children

• Moves reflect a step up to better housing and
  neighborhood circumstances
• Moves reflect a change in household composition
  ( housing needs)
• Moves reflect instability insecurity

17
Some variation among stayers

• Neighborhood engagement
• 38 attend neighborhood events 29 volunteer in
  the neighborhood 31 work with neighbors for
  change
• Perception of neighborhood
• 46 score safety high
• 55 think its getting better 12 think its
  getting worse
• Satisfaction with services
• 86 highly satisfied with kids school 6
  dissatisfied
• 74 highly satisfied with banking services 3
  dissatisfied
• 90 highly satisfied with parks 7 dissatisfied

18
Hypothesized clusters of Stayers

1. Staying reflects attachment and satisfaction
2. Staying reflects dissatisfaction lack of
   alternatives

19
Lots of variation among newcomers

• Employment
• 26 have no employed adults 37 have a stable
  job
• Income
• 6 have incomes gt 300 poverty 66 have incomes
  below poverty
• Housing
• 22 are homebuyers 26 live in subsidized
  housing 40 report difficulty paying housing
  costs
• Perception of neighborhood
• 65 think its a good place to raise kids 47
  think its likely to get better
• Engagement
• 29 attend neighborhood events 18 volunteer in
  the neighborhood 15 work with neighbors to
  solve problems

20
Hypothesized clusters of Newcomers

1. Affluent newcomers investing in expectation of
   neighborhood change (gentrifiers)
2. Newcomers similar to current residents
   optimistic about neighborhood quality
3. Newcomers whose moves reflect instability
   insecurity

21
Methods (contd)

• Step 1 Create cluster predictions
• Step 2 Identify variables of interest for each
  model
• Different variables selected for the four models
  based on theory and data availability
• Individual factors
• Demographic/family composition,
  employment/income, hardship, homeownership,
  neighborhood services and perceptions,
  neighborhood attachment
• Neighborhood factors
• Housing market, poverty, racial composition

22
Methods (contd)

• Step 3 Test for correlations among variables
  that reflect push pull factors
• Correlation Matrices
• Step 4 Principle components analysis to identify
  possible composite factors
• Collapse data where appropriate
• Step 5 Look at the data
• Scatter diagrams, tree graph

23
Methods (contd)

• Step 6 Cluster Procedures
• Standardize coefficients
• Jaccard coefficient is a reliable and simple
  method
• Hierarchical or Non-hierarchical (k-means)
  cluster analyses
• SPSS, SAS, and STATA have established commands
• Specify number of clusters
• Run cluster procedure multiple times with
  different numbers of clusters specified

24
Methods (contd)

• Step 6 Cluster Procedures (contd)
• Review generated clusters
• Investigate clusters, interpret, profile groups
• A heuristic Local maximum of pseudo F statistic,
  with local minimum of R-squared
• Step 7 Robustness tests
• Run multiple cluster tests
• Compare with different variable specifications
• Split sample, cluster again
• Step 8 Use the findings!
• Compare groups along key measures

25
Why Are Families Moving In and Out of the MC
Neighborhoods Cluster Analysis Illustrative
Findings
26
Illustrative Results

• 4 Types of out-movers with kids
• Optimistic Homebuyers
• Changed Family Circumstances
• Reluctant Movers
• Unstable Families

27
Illustrative Results (contd)
Out-Mover Demographics
28
Illustrative Results (contd)
Out-Movers Differ By Sites
29
Illustrative Results

• 3 Types of Stayers
• Subsidized
• Attached
• Trapped

30
Illustrative Results (contd)
Stayer Demographics
31
Illustrative Results (contd)
Stayers Differ by Sites
32
Illustrative Results

• 3 Types of Newcomers
• Subsidized
• Attached
• Trapped

33
Illustrative Results (contd)
Newcomer Demographics
34
Illustrative Results (contd)
Newcomers Differ by Sites
35
Analysis Next Steps
36
Cluster Analysis Next Steps

• Apply cluster analysis to 9 site pooled data
• Conduct robustness tests
• Analyze clusters to find
• Distribution of households across clusters by
  site
• Service utilization, demographic characteristics,
  and key outcomes of cluster groups
• Use clusters to characterize MC neighborhoods
• Incubators, launch pads, traps, gentrifying
• Map locations for different types of out-movers
  with kids, childless movers, stayers, and
  newcomers

37
Cluster Analysis References

• Afifi, Abdelmonem, Virginia Clark, and Susanne
  May. 2003. Computer-Aided Multivariate
  Analysis. Chapman and Hall.
• Finch, Holmes. 2005. Comparison of Distance
  Measures in Cluster Analysis with Dichotomous
  Data. Journal of Data Science, 3.