Family structure and child outcomes: an illusive relationship

1

• Family structure and child outcomes an illusive
  relationship
• Don Kerr
• Kings University College
• University of Western Ontario
• 2004 Canadian Population Society Meetings
• Winnipeg, Manitoba

2

• To what extent is "family structure
• important in predicting child outcomes?
• 2 issues
• Why should family structure matter?
• What empirical evidence is currently available on
  this issue with the NLSCY?

3
Behavioral Scales for Children 4-11, NLSCY (first
cycle) by Family Type
4

• Why should family structure matter?
• In both step and lone parent families
• child loses out from the lack of co-residence
  with one biological parent
• -gt less parental supervision
• -gt lower transfer of social human capital
• 
  (Amato Booth, 1997)
• There is no consensus on this issue!!!

5

• In working with the first wave of the NLSCY
• Multivariate analyses
• -gt family type is found to be a useful
  predictor of childhood difficulties
• -gt association persists with controls for low
  income, age, educ, etc.
• What does this cross sectional association
  represent???
• gt the impact of being raised in a lone
  parent/step family
• OR
• gt the impact of antecedent factors
• i.e. what lead to the formation of the
  lone parent/step family in the first place?
  (conflict, abuse?)

6

• In Longitudinal analysis
• -gt most common approach
• autoregressive or residual change analysis
• -gt An alternative is now possible (4 cycles of
  the NLSCY)
• Latent Growth Curve Models (LGM)

7
Autoregressive or Residual change approach
1994
2000
Child outcome 2000
Child outcome 1994
Family structure, income, etc.

• Many critiques of this approach
• Rogosa et al. 1982 Rogosa and Willett (1985)

8

• As an alternative
• Latent Growth Curve Models (LGM)
• 
  Duncan et al (1999)
• -gt does not model variance at a specific point
  in time
• -gt attempts to model individual trajectories on
  dependent variable over time

9

Example with consistent measures of
hyperactivity over 4 cycles (1994, 1996, 1998,
2000)
Score on hyperactivity scale
10
Latent Growth Models
Intercept
Slope
Scale on externalizing problems
1994
1996
1998
2000
E2
E3
E4
E1
11
Latent Growth Models
Mean -.565 Var .466
Intercept
Slope
Mean 4.9 Var 6.7
Scale on externalizing problems
1994
1996
1998
2000
E2
E3
E4
E1
12
Latent Growth Models
Lone parent 1994-2000
-.05ns
.176
Mean -.565 Var .466
Intercept
Slope
Mean 4.9 Var 6.7
Scale on externalizing problems
1994
1996
1998
2000
E2
E3
E4
E1
13
Latent Growth Models
Lone parent 1994-2000
Low income 1994-2000
-.05ns
-.033ns
.176
.099
Mean -.565 Var .466
Intercept
Slope
Mean 4.9 Var 6.7
Scale on externalizing problems
1994
1996
1998
2000
E2
E3
E4
E1
14
GFI.970 CFI.971 ?2 355.6 df 13
Latent Growth Models
Lone parent 1994-2000
Step Family 1994-2000
Low income 1994-2000
-.05ns
.132
-.033ns
.107
.176
.099
Mean -.565 Var .466
Intercept
Slope
Mean 4.9 Var 6.7
Scale on externalizing problems
1994
1996
1998
2000
E2
E3
E4
E1
15

• Conclusion
• Results are preliminary
• Results are very mixed as to the importance of
  family structure
• Future research -gt additional controls/behavioral
  scales
• LGM looks particularly promising in the analysis
  of change