abstract
DESCRIPTION
Tracking Mobility at the Household Level Kate Bachtell, Ned English and Catherine Haggerty. Residential movement. DISCUSSION. ABSTRACT. DATA AND METHODS. - PowerPoint PPT PresentationTRANSCRIPT
This work examines the methodological challenges associated with tracking mobility at the household level. We describe a retroactive approach for linking individual household members across waves that was employed for the Making Connections Survey. We compared individuals at three different points in time using a combination of probabilistic matching software, data queries, and human review. The process produced personal identifiers that enable us to examine mobility across a gradient of stability in household composition. Our work advances past studies in two ways. First, our definition of adding or losing individuals is calculated based on the presence or absence of a specific person, rather than numerical change in household size. Second, we demonstrate a more nuanced understanding of household mobility by examining specific types of change in household composition in combination with physical relocation. A series of maps compare the patterns of residential movement and household composition change among sample members.
ABSTRACT DATA AND METHODS
Limitations• These seven sites are not representative of poor urban communities nationwide.• There may be unmeasured differences due to sample attrition.
Conclusions• Personal identifiers should be assigned at the time of data collection (though retroactive
assignment is possible).• Mobility should be measured in at least two dimensions: (1) residential movement and (2)
household composition change.
DISCUSSION
REFERENCES
FINDINGS
BACKGROUND
CONTACT INFORMATION
Kate BachtellSenior Survey Director, Economics, Labor, and Population Studies 55 East Monroe Street, 20th Floor, Chicago IL [email protected] (312) 759-5095
(Figure 3) Residential Moves among San Antonio Households Over Six Years
MobilityMobility has been shown to impact economic and social well-being and is studied intensely in research on neighborhood effects. Mobility presents a methodological challenge for longitudinal data collection in that it is both labor and cost-intensive to follow families to unknown destinations (Ansolabehere and Schaffner 2010, Marshall and Bush 2010). It can introduce coverage bias if those respondents who are not successfully located are somehow different from those who are located. Household mobility is especially challenging for longitudinal studies in that entry and exits among individuals may occur for a wide range of reasons, including births, deaths, marriage, divorce, financial necessity, etc. (Marshall and Bush 2010, 32; Duncan and Hill 1985, 361).
Mobility among Low-Income PopulationsAfter three waves of data collection for the Making Connections Survey, we continue to observe movement that exceeds national levels. This is consistent with abundant literature demonstrating that disadvantaged neighborhoods experience higher rates of mobility (Coulton et al 2009).
The ProblemConceptually, we aim to approach mobility as a dynamic family process rather than the behavior of an individual. This presents a methodological challenge:
How do we track a moving and morphing target?
Data SourceThis research features data from the Making Connections Survey (http://mcstudy.norc.org), a longitudinal and cross-sectional study of ten low income neighborhoods. The survey is part of a larger community change initiative funded by the Annie E. Casey Foundation.
Methods1. Retroactively assign personal identifiers to link individuals over time2. Create measures of residential movement and household composition change (see Figures 1-2)3. Use GIS and statistical methods to examine both components of mobility (see Figures 3-6 and Table 1)
TWO DIMENSIONS TO MOBILITY
Residential movement Household composition change
(Changing place) (Changing people)
Wave 1 Wave 2 Wave 1 Wave 2
(Figure 1) Our Conception of Mobility Involving Two Distinct Processes: Residential Movement and Household Composition Change
(Table 1) Comparison of One and Two-Dimensional Approaches to Mobility: Percent of Households Defined as Experiencing Mobility (Seven Sites)
Waves 1-2 Waves 2-3
n % n %One-dimensional: just residential movement 2,152 51.1 2,800 54.1
Two-dimensional: residential movement and composition change 2,838 71.0*** 3,509 67.9***
Missed mobility when employing one-dimensional approach 686 17.2 709 13.7
After filtering out “natural” changes in household composition (birth of new children to the respondent and teenagers and adult children age 18-30 moving out), we find that 71.0% and 67.9% of households between waves 1-2 and 2-3, respectively, are found to have experienced mobility using the two-dimensional approach. This represents a statistically significant increase over the percentage observed using a one-dimensional approach focusing just on residential movement (see Table 1).
(Figure 4) Residential Moves and Household Composition Change among San Antonio Households between W1 and W3 (Six Year Interval)
(Figure 5) Percent of San Antonio Households with Residential Moves Over Six Years: Aggregate for Census Tract Based on W1 Address
(Figure 6) Percent of San Antonio Households with HH Composition Change Over Six Years: Aggregate for Census Tract Based on W1 Address
Ansolabehere, Stephen and Brian F. Schaffner. Summer 2010. “Residential Mobility, Family Structure, and the Cell-Only Population.” Public Opinion Quarterly 74(2): pp. 224-259.
Coulton, Claudia; Theodos, Brett; and Margery Turner. November 2009. "Family Mobility and Neighborhood Change: New Evidence and Implications for Community Initiatives." Report prepared for the Annie E. Casey Foundation by the Urban Institute. Washington, D.C.
Duncan, Greg J. and Martha S. Hill. 1985. “Conceptions of Longitudinal Households: Fertile orFutile?” Journal of Economic and Social Measurement 13: pp 361-375.
Marshall, David and Robert Bush. March 2010. “The Ipswich Study: A Review of Longitudinal Methodology.” Final Report prepared by the Healthy Communities Research Centre. University of Queensland, Australia.
Using San Antonio as an example, in Figures 3 through 6 we demonstrate the distinction between residential movement and household composition change. We find that composition changes were far more common than residential moves among San Antonio households over six years (87.2% and 50.0%, respectively). Figures 5 and 6 portray the differences in the aggregate sum of residential movement and household composition change over six years at the Census tract level. The overall “darkening” of the neighborhood in Figure 6 supports the depiction of households as dynamic collections of individuals and highlights the degree of turnover occurring among neighborhood residents.
Person Matching• Retroactive process• Compiled data for every individual ever found in an interviewed
household • Used Link Plus to establish matches • Performed 100% human review• Output = unique PERSONIDs
Analytic Dataset• Households that participated in wave 3 from seven sites: Des
Moines, Indianapolis, Denver, San Antonio, White Center (Seattle), Providence, and Louisville
• 3 points in time: W1 (‘02-’04), W2 (’05-’07), and W3 (‘08-’11)• Household composition data (from 32,394 individuals)• Household-level address data (18,043 addresses)
(Figure 2) Example of Two-Dimensional Approach Tracking Residential Movement and Household Composition Change
None One Two Three Four or more0
102030405060708090
Gained Adults over Six Years
Households without Residential Movement Households with Residential Movement
None One Two Three Four or more0
102030405060708090
Lost Adults over Six Years
Households without Residential Movement Households with Residential Movement
***p<0.001
None One Two Three Four or more0
20
40
60
80
Gained Children over Six Years
Households without Residential Movement Households with Residential Movement
None One Two Three Four or more0
102030405060708090
Lost Children over Six Years
Households without Residential Movement Households with Residential Movement
(Figures 7-10) Detail of Household Compositional Change Over Time by Residential Movement Status (Percentage, Seven Sites)
Tracking Mobility at the Household Level Kate Bachtell, Ned English and Catherine Haggerty