Nola du Toit Cathy Haggerty Instability Overlooked: Evidence of the Importance of Household Roster Data Collection and Matching Over Time.

Nola du Toit

Cathy Haggerty

Instability Overlooked:  Evidence of the Importance of Household Roster Data Collection and Matching Over Time

2

Background: Instability

• Relationship Instability • Divorce or dissolution of cohabiting union

• Children wellbeing (Amato and Sobolewski 2001) • School performance (Frisco et al. 2007)• Life satisfaction (Hans-Jurgen and Brockel 2007)

• Household instability• Change in number of children/adults

• Household debt (Disney et al. 2008)• Depression (Heflin and Iceland 2009)• Outdoor play (Handy et al. 2008)

3

Measuring Instability: Relationship Instability

• Relationship instability • Change in marriage or cohabiting relationship over

time

TIME 1 TIME 2 OUTCOME

Are you married/cohabiting?

Are you married/cohabiting?

Instability

YES YES 0

YES NO 1

4

Measuring Instability: Change in Number of Adults/Children

• Household instability• Change in number of adults or children

• Are we missing anything?

TIME 1 TIME 2 OUTCOME

Number of children Number of children Instability

3 3 03 4 1

5

• 1959: Are you married? YES• 1964: Are you married? YES

• Instability = NO

• But….

• 1959: Married to Eddie Fisher

• 1964: Married to Richard Burton

• Instability?

Example1:Elizabeth Taylor

YES

6

• 3 children in the household• Lisa, Maggie, and Bart

• Episode 2005• 3 children in the household• Lisa, Maggie, and “BART”

• Still 3 children• But not the same 3 children

• Instability?

Example 2: The Simpsons

YES

7

Research Questions

• Is there another way to measure instability?

• How much instability is overlooked by current measures?

• Does it matter?

8

Data

• Making Connections Survey• Annie E. Casey Foundation• Low income households at 10 sites• Longitudinal

• Baseline (2002-2004)• Wave 2 (2005-2007)• Wave 3 (2008-2011)

• Information on variety of topics• People in household, age, gender, employment,

relationships to one another, children, economic wellbeing, etc.

9

Data

• Household roster data

• Matched across waves

• Link Plus, eyeballing

• Unique identifiers for everyone in household

• Waves 2 and 3 for 6 sites

• 2242 cases

PERSON 1 PERSON 2 PERSON 3

WAVE 3 61116380 A . 61116380 CWAVE 2 61116380 A 61116380 B .

10

Methodology

• Compare “old way” to “new way”

• Relationship Instability• Old way = Is there a spouse or partner present?• New way = Is the Wave 2 spouse or partner id#

present in Wave 3?

• Household Instability in Number of Children and Adults• Old way = count of adults and children• New way = count of same id# at each wave

• Compare old way and new way• Age, gender, log of income, education, and employment

11

Findings: Relationship Instability

***Significantly different p<.001

12

Findings: Relationship Instability

6% is not the same person

***Significantly different p<.001

13

Findings: Household Instability

***Significantly different p<.001

Change in number of children

Change in number of adults

14

Findings: Household Instability

***Significantly different p<.001

Change in number of children

Change in number of adults

15

Does it matter?

Wave 2 CharacteristicsRelationship Instability

OLD WAY NEW WAYUnweighted n (141) (176)Age mean 38.30 40.81Household income mean ($) $35,499.08 $31,915.26 Household income (log) mean 9.97 9.86Percent Female 87% 83%Education - Less than high school 37% 42%Education - High school/GED 32% 30%Education - More than high school 31% 28%Employed respondent 56% 52%

NOTE: *p<.05, **p<.01; ***p<.001

16

Does it matter?

WAVE 2 CHARACTERISTICSRELATIONSHIP

INSTABILITYOLD WAY NEW WAY

Unweighted n (141) (176)Age mean 38.30 * 40.81Household income mean ($) $35,499.08 $31,915.26 Household income (log) mean 9.97 9.86Percent Female 87% *** 83%Education - Less than high school 37% *** 42%Education - High school/GED 32% ** 30%Education - More than high school 31% *** 28%Employed respondent 56% *** 52%

NOTE: *p<.05, **p<.01; ***p<.001

17

Does it matter?

Wave 2 CharacteristicsChange in

Number of AdultsOLD WAY NEW WAY

Unweighted n (670) (786)Age mean 40.29 40.30Household income mean ($) $30,426.91 $29,851.89 Household income (log) mean 9.85 9.84Percent Female 80% 78%Education - Less than high school 33% 34%Education - High school/GED 35% 35%Education - More than high school 31% 31%Employed respondent 58% 58%

NOTE: *p<.05, **p<.01; ***p<.001

18

Does it matter?

Wave 2 CharacteristicsChange in

Number of AdultsOLD WAY NEW WAY

Unweighted n (670) (786)Age mean 40.29 40.30Household income mean ($) $30,426.91 $29,851.89 Household income (log) mean 9.85 9.84Percent Female 80% *** 78%Education - Less than high school 33% * 34%Education - High school/GED 35% 35%Education - More than high school 31% 31%Employed respondent 58% 58%

NOTE: *p<.05, **p<.01; ***p<.001

19

Does it matter?

Wave 2 CharacteristicsChange in

Number of AdultsOLD WAY NEW WAY

Unweighted n (511) (544)Age mean 35.56 35.62Household income mean ($) $26,994.62 $26,187.39 Household income (log) mean 9.73 9.72Percent Female 85% 85%Education - Less than high school 38% 38%Education - High school/GED 33% 34%Education - More than high school 28% 27%Employed respondent 59% 60%

NOTE: *p<.05, **p<.01; ***p<.001

20

Does it matter?

Wave 2 CharacteristicsChange in

Number of AdultsOLD WAY NEW WAY

Unweighted n (511) (544)Age mean 35.56 35.62Household income mean ($) $26,994.62 $26,187.39 Household income (log) mean 9.73 9.72Percent Female 85% 85%Education - Less than high school 38% 38%Education - High school/GED 33% * 34%Education - More than high school 28% 27%Employed respondent 59% * 60%

NOTE: *p<.05, **p<.01; ***p<.001

21

Conclusions

• Is there another way to measure instability?• Roster matching and unique person identifiers add

depth

• How much instability is overlooked by current measures?• Old way overlooks significant amount of instability

• Does it matter?• Characteristics are significantly different

22

Conclusions

• Study is important• Household rosters are important part of questionnaire• Roster matching and unique identifiers add depth• Substantive topics are enhanced by the added details

• Limitations to roster matching and unique identifiers• Burdensome process• Ideal for longitudinal surveys

• But can be incorporated into cross-sectional surveys

• Very beneficial to capturing the whole picture

Thank You!

Nola du Toit: [email protected]

Cathy Haggerty: [email protected]