Assessing Parenting Behaviors Across

1
Assessing Parenting Behaviors Across Racial
Groups Lawrence M. Berger Christina
Paxson Center for Health and Wellbeing Princeton
University prepared for Fragile Families
workshop Columbia University July 2004
2
In-home addition to Fragile Families

• Added an in-home assessment to the Fragile
  Families Study.
• In-home assessments added at 36 and 60 months.
• Information includes
• Child health and behaviors
• Parenting (through interviews and direct
  observation)
• Childrens cognitive development (PPVT,
  Woodcock-Johnson)
• Parental physical and mental health and cognitive
  ability (stress anthropometrics PPVT)
• At 60 months ask families about involvement with
  the child welfare system.

3
Questions to be addressed

• How do factors such as economic status, family
  structure, and parental health affect the
  environments in which children live?
  (Environments include aspects of the physical
  environment as well as parenting.)
• What are the determinants of very poor parenting
  behavior?
• How does the care that children receive (across a
  variety of domains) affect their cognitive and
  developmental outcomes?
• Do policy changes that alter the economic status
  of parents affect childrens environments and
  outcomes?

4
This paper (in progress!) Racial Differences in
Parenting

• Why is this topic of interest?
• A large literature documents racial and ethnic
  differences in parenting of young children, along
  a variety of domains.
• It is possible that these parenting differences
  may account for differences in early
  achievement/school readiness.
• Question how much (and which) socioeconomic and
  parental characteristics can account for racial
  and ethnic differences in parenting?

5
Another motivation Child Maltreatment and Race

• The issue of racial bias in the assessment of
  parenting is an important topic in the child
  maltreatment literature.
• Based on official statistics, blacks represent 15
  percent of the U.S. children and 25 percent of
  child maltreatment victims. Whites represent 65
  percent of children and 51 percent of
  maltreatment victims.

6
Statistics from Child Abuse and Neglect 2002
(NCCAN) (cases/1000)
By age 0-3 16.0
4-7 13.7
8-11 11.9
By type Neglect 7.2
Physical abuse 2.3
Other 5.6
By race Black 20.2
White 10.7
Hispanic 9.5
7
Problems of Interpretation

• Validity of official statistics as accurate
  measures of maltreatment rates can be questioned.
• Poor childrenespecially those on welfaremay be
  more likely to come to the attention of mandated
  reporters, especially at young ages.
• Some evidence of racial bias in reports of
  maltreatment (Lane et al, JAMA).

8

• What might explain racial differences in
• maltreatment rates?
• There is an active debate over the source of the
  disparity in official maltreatment rates.
  Potential explanations include
• Differences in parenting behaviors that are
  correlated
• with race.
• Racial bias in the child protective services
  (CPS)
• system.
• Racial bias refers to a racial double standard,
  such that otherwise identical families of
  different races are assessed differently by
  reporters of maltreatment or by the social
  workers, judges, etc. who verify the validity of
  reports of maltreatment.

9

• What might drive racial bias?
• There are several reasons in theory why racial
  bias in reporting and substantiation might be
  observed
• Stereotyping Individuals expect to see worse
• parenting among some groups, and observe what
  they
• expect to see. (Could be driven by genuine
  differences
• in average behaviors in the population.)
• Miscommunication Individuals do not know how to
  
• interpret information from members of some
  groups.
• Assessments may not be systematically biased,
  but
• may be more error-ridden (so more chance that
  an
• extreme value is observed.

10

• Implications, for maltreatment rates and for
  surveys
• Maltreatment Greater error in the system,
  possibly for both blacks and whites.
• Surveys The same kinds of factors that result
  in biased assessments by reporters and social
  workers could result in biased assessments by
  interviewers.

11

• Research Objectives
• To investigate whether there are differences in
  parenting across racial groups (blacks versus
  whites).
• To assess whether these differences can be
  accounted for by household and maternal
  characteristics other than race.
• To study whether parents reports of their own
  behaviors and interviewers assessments of
  parental behaviors are related to the race of the
  interviewer.

12
Data Our data are drawn from an in-home module
of the Fragile Families and Child Wellbeing Study
(FFCW). This contains self-reported and
interviewer-observed parenting measures. Data
are available for 20 cities from the 36-month
interview. Our sample consists of 1,149 blacks
and 367 whites. All completed the full in-home
module, and had interviewers for whom race was
known.
13

• Parenting Measures
• Based on maternal reports
• (1) absence of nonviolent discipline
• (2) use of psychological aggression
• (3) use of physical assault
• Based on interview observations
• (1) harshness
• (2) lack of warmth

14

• Measures of Maternal and Child Characteristics
• Based on interviewer observations
• Maternal verbal/social skills
• Maternal understanding/attention
• Maternal hostility/suspicion
• Problems with childs appearance
• Problems with childs behavior

15

• Sociodemographic Characteristics
• and Maternal Behaviors (core survey)
• natural logarithm of needs-adjusted family
  income
• the number of children and adults in the
  household
• indicators for maternal education
• indicators for presence and employment of
  father/other man in household
• an indicator for whether the mother worked in
  the week before the core survey
• maternal depression score
• Indicator variables for whether the mother
  smoked cigarettes, drank alcohol, or used drugs
  during her pregnancy

16

• Question 1
• Are there racial/ethnic differences in parenting?
• Cross-tabulations of parenting measures indicate
  there are significant differences in the
  distributions across whites and blacks.

17
Examples Frequency mother uses time-outs,
explains why something is wrong, gave something
else to do, etc.
18
Examples Called child lazy or dumb shouted,
yelled or screamed at child threatened to hit
threatened to kick out of house
19
Examples Spanked, slapped, hit on bottom with
object pinched shook.
20
Items Parent did not shout at child Parent did
not express annoyance with or hostility toward
child. Parent neither slapped nor spanked child
during the visit. Parent did not scold or
criticize child during visit. Parent did not
interfere or restrict child more than 3 times.
21
Items Parent spontaneously vocalized to child
twice. Parent responded verbally to childs
vocalizations. Parent told child the name of an
object or person during visit Parent
spontaneously praised child at least
twice. Parent's voice conveys positive feelings
toward child. Parent caressed or kissed child at
least once.
22
Items Parent's speech was distinct and
audible. Parent initiated verbal exchanges with
visitor. Parent conversed freely and easily.
23
Items Respondents attention to interview (coded
0-3) Respondents understanding of questions
(coded 0-3) Respondents ability to articulate
answers (coded 0-3) Respondents level of
cooperation (coded 0-3)
24
Items Did respondent appear to
be Suspicious? Uncommunicative? Anxious/nervous?
Hostile? To be on drugs? (Each on 3 point scale)
25
Examples Childs clothes dirty, ill-fitting, or
inappropriate for season Child appears to be
recently bathed Childs hair is combed and clean
26
distributions not significantly different across
races
Items (each scored 0-3) Displays of negative
emotions during interview Displays of positive
emotions during interview Cooperation during
PPVT Cooperation during height and weight
measurement Persistence during PPVT
27
Sociodemographic characteristics
white black t-test
Ln(income/a.e.) 9.45 (0.95) 8.49 (1.05) 15.42
children 1.94 (1.14) 2.44 (1.42) 6.20
adults 2.00 (0.61) 1.84 (0.85) 3.41
Mother works 0.594 0.564 1.01
Mother has high school degree 0.259 0.356 3.45
Mother has more than high school degree 0.569 0.285 10.23
28
Sociodemographic characteristics
white black t-test
Childs father at home and not working 0.071 0.091 1.21
Other man at home and working 0.071 0.085 0.87
Other man at home and not working 0.008 0.029 2.27
Neither father nor other man at home 0.218 0.505 9.95
29
Maternal depression and risky behaviors
white black t-test
Maternal depression score 1.40 (2.53) 1.64 (2.67) 1.54
Smoked while pregnant 0.294 0.211 3.28
Drug use while pregnant 0.046 0.071 1.69
Alcohol use while pregnant 0.158 0.102 2.89
30
Question 2 Do Sociodemographic and Maternal
Characteristics Account for Racial Differences?

• Estimate regressions that control for city of
  residence then add sociodemographics then add
  maternal depression and risky behaviors.
• Dependent variables are indicators being above
  the 75th and 90th percentiles for each of the
  scales.
• I show OLS results, because I move to fixed
  effects models later (although conditional logits
  show similar results.)

31
Parenting Measures at 75th Percentile Cut Points
Lack of non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
Mother Black 0.144 (0.027) 0.046 (0.038) 0.073 (0.028) 0.083 (0.024) 0.160 (0.028)
Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls
Mother Black 0.098 (0.030) 0.030 (0.031) 0.073 (0.031) 0.026 (0.026) 0.077 (0.031)
Test Socio-demographics 0.008 0.100 0.250 0.000 0.000
Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors
Mother Black 0.091 (0.030) 0.031 (0.031) 0.069 (0.031) 0.028 (0.027) 0.076 (0.031)
Test Socio-demographics 0.002 0.211 0.364 0.000 0.000
Test Mater-nal controls 0.037 0.000 0.035 0.702 0.216
32
Parenting Measures at 90th Percentile Cut Points
Lack of non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
Mother Black 0.055 (0.018) 0.020 (0.020) 0.046 (0.017) 0.070 (0.019) 0.069 (0.018)
Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls Add sociodemographic controls
Mother Black 0.021 (0.020) -0.001 (0.022) 0.049 (0.019) 0.027 (0.021) 0.026 (0.020)
Test Socio-demographics 0.000 0.001 0.401 0.000 0.000
Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors Add maternal depression and risky behaviors
Mother Black 0.016 (0.020) -0.006 (0.022) 0.051 (0.019) 0.025 (0.023) 0.024 (0.020)
Test Socio-demographics 0.000 0.015 0.527 0.000 0.002
Test Mater-nal controls 0.039 0.000 0.329 0.928 0.677
33
Parenting Measures at 75th Percentile Cut Points
Non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
Ln(income) -0.048 0.015 0.008 -0.015 -0.031
children 0.002 -0.007 -0.018 -0.008 0.005
adults 0.008 -0.020 0.003 -0.028 -0.011
Mother works 0.026 0.009 -0.017 -0.035 -0.028
High school -0.023 -0.033 0.044 -0.015 -0.032
gt High school -0.027 -0.084 -0.014 -0.081 -0.109
Dad, no work 0.061 0.005 0.015 -0.024 0.013
Step, work -0.029 -0.024 -0.007 0.015 0.055
Step, no work 0.076 -0.101 -0.068 -0.047 -0.033
No dad/step 0.006 0.015 0.013 0.045 0.034
Depression -0.009 0.016 0.010 0.001 0.007
Smoking -0.053 -0.002 -0.002 -0.012 -0.008
Drug use -0.010 0.137 0.102 -0.007 0.081
Alcohol use 0.017 -0.006 -0.053 0.043 -0.007
34
Question 3 Does the Race of the Interviewer
Matter?

• Estimate regressions that control for the race of
  the interviewer (including all other controls
  used so far).
• Estimate models with interactions between the
  race of the mother and the interviewer.
• Estimate models with race interactions and
  interviewer fixed effects.

35
Descriptive information on interviewers

• Interviewer characteristics

36
of interviewers 100
Average interviews per interviewer 15.2
Range of interviews per interviewer 1-54
of interviewers who are black 39
of black interviewers who interview both black and white mothers 77
of non-black interviewers who interview both black and white mothers 66
37
Descriptive information on interviewers

• Interviewer characteristics
• Assignment of interviewers to respondents

38
Interviewer non-black Interviewer black Total
Mother white 227 (15.0) 140 (9.2) 367 (24.2)
Mother black 580 (38.3) 569 (37.5) 1,149 (75.8)
Total 807 (53.2) 709 (46.8) 1,516 (100.0)
39
Descriptive information on interviewers

• Interviewer characteristics
• Assignment of interviewers to respondents
• Is assignment random?

40
Dependent variable indicator that interviewer is
black
Probit, no city dummies OLS, no city dummies OLS, city dummies
Mother is black 0.130 (0.033) 0.129 (0.034) 0.010 (0.025)
Test Sociodemographic and maternal controls insignificant 0.244 0.245 0.376
F-Test Joint insignificance of city dummies 91.9 (0.000)
Note All sociodemographic and maternal controls
included.
41
RESULTS
42
Lack of non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
Mother is black 0.090 (0.030) 0.030 (0.031) 0.069 (0.031) 0.029 (0.026) 0.081 (0.030)
Interview is black 0.032 (0.031) 0.025 (0.032) -0.013 (0.032) -0.105 (0.027) -0.380 (0.031)
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
Mother is black 0.016 (0.020) -0.006 (0.022) 0.051 (0.019) 0.026 (0.022) 0.026 (0.020)
Interview is black 0.067 (0.021) 0.035 (0.023) 0.012 (0.020) -0.056 (0.022) -0.192 (0.020)
Note All models include all sociodemographic and
maternal controls and city dummies
43
Interpretation of these results

• The fact that there are systematic differences in
  assessments by the race of the interviewer,
  although interesting, does NOT provide evidence
  of bias.
• We need to examine whether the assessments of
  black relative to white respondents vary across
  black and white interviewers.
• Next set of regressions include a complete set of
  interactions between the race of the interviewer
  and the respondent (omitted category white
  respondent and non-black interviewer.)

44
Non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
M black I non-black 0.088 (0.037) 0.021 (0.038) 0.060 (0.039) 0.065 (0.033) 0.106 (0.037)
M white I black 0.028 (0.054) 0.006 (0.056) -0.031 (0.056) -0.033 (0.047) -0.328 (0.053)
M black I black 0.121 (0.045) 0.050 (0.046) 0.052 (0.046) -0.055 (0.039) -0.284 (0.044)
Bias test 0.926 0.666 0.691 0.066 0.238
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
M black I non-black 0.004 (0.025) -0.005 (0.027) 0.059 (0.024) 0.066 (0.027) 0.070 (0.024)
M white I black 0.042 (0.036) 0.037 (0.039) 0.029 (0.035) 0.023 (0.039) -0.103 (0.035)
M black I black 0.076 (0.030) 0.030 (0.032) 0.067 (0.029) -0.007 (0.032) -0.140 (0.029)
Bias test 0.398 0.949 0.550 0.013 0.002
45
Harshness (90th percentile cut point)
Interviewers race Interviewers race
white black diff
Mothers race black 0.066 -0.007 0.073
Mothers race white 0 0.023 -0.023
diff 0.066 -0.030 0.096
Test for bias (MB/IW MW/IW) (MB/IB -
MW/IB) Or ( 0.066 0 )
(-0.007 0.023) 0 Or 0.096
0 (p-value for test is 0.013).
46
Physical assault (90th percentile cut point)
Interviewers race Interviewers race
white black diff
Mothers race black 0.059 0.067 -0.008
Mothers race white 0 0.029 -0.029
diff 0.059 0.038 0.021
Test for bias p-value0.550 (hypothesis of no
racial bias cannot be rejected.)
47
Lack of Warmth (90th percentile cut point)
Interviewerss race Interviewerss race
white black diff
Mothers race black 0.070 -0.140 0.210
Mothers race white 0 -0.103 0.103
diff 0.070 -0.037 0.107
Test for bias p-value is 0.002
48
With Interviewer Fixed Effects
Non-violent discipline Psych. aggression Physical assault Harsh-ness Lack of warmth
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
M black I non-black 0.089 (0.038) 0.023 (0.039) 0.064 (0.039) 0.053 (0.032) 0.077 (0.034)
M black I black 0.088 (0.046) 0.042 (0.048) 0.089 (0.048) -0.037 (0.039) 0.061 (0.042)
Bias test 0.990 0.744 0.662 0.061 0.761
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
M black I non-black 0.009 (0.025) -0.001 (0.028) 0.063 (0.025) 0.064 (0.027) 0.060 (0.024)
M black I black 0.033 (0.031) -0.003 (0.034) 0.033 (0.030) -0.040 (0.032) -0.021 (0.029)
Bias test 0.527 0.967 0.415 0.009 0.024
49

• Summary so far
• Self-reports of parenting behavior are not
  affected by the race of the interviewer relative
  to the respondent.
• The race of the interviewer is systematically
  related to the assessments of harshness and lack
  of warmth black interviewers are consistently
  less likely to give all parents, black and
  white, more negative ratings.
• We find evidence of racial bias in assessments
  of harshness (at 75th and 90th ptile cut point)
  and lack of warmth (at 90th ptile cut point).
• Could these findings be due to more negative
  interactions of the interviewer with the mother
  in mixed-race interviews, or worse child
  behavior?

50
Maternal verbal/social Maternal attention Maternal hostility Child appearance Child appearance Child behavior
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
Mother is black 0.042 (0.024) 0.046 (0.030) -0.045 (0.031) -0.037 (0.032) -0.037 (0.032)
Interview is black -0.243 (0.025) -0.051 (0.031) -0.132 (0.032) -0.106 (0.033) -0.106 (0.033)
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
Mother is black 0.018 (0.017) 0.014 (0.018) 0.029 (0.017) -0.019 (0.018) -0.020 (0.021) -0.020 (0.021)
Interview is black -0.121 (0.018) -0.090 (0.019) -0.090 (0.017) -0.088 (0.018) -0.062 (0.022) -0.062 (0.022)
Note All models include all sociodemographic and
maternal controls and city dummies
51
Maternal verbal/social Maternal attention Maternal hostility Child appearance Child behavior
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
M black I non-black 0.092 (0.030) 0.066 (0.037) -0.092 (0.038) -0.012 (0.040)
M white I black -0.143 (0.043) -0.010 (0.055) -0.225 (0.055) -0.055 (0.058)
M black I black -0.172 (0.035) 0.006 (0.045) -0.204 (0.046) -0.130 (0.048)
Bias test 0.005 0.365 0.041 0.273
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
M black I non-black 0.045 (0.021) 0.027 (0.023) 0.047 (0.021) -0.022 (0.022) 0.007 (0.026)
M white I black -0.067 (0.031) -0.063 (0.033) -0.053 (0.030) -0.095 (0.032) -0.007 (0.038)
M black I black -0.088 (0.026) -0.069 (0.027) -0.052 (0.025) -0.108 (0.026) -0.067 (0.031)
Bias test 0.033 0.321 0.127 0.789 0.076
52
With Interviewer Fixed Effects
Maternal verbal/social Maternal attention Maternal hostility Child appearance Child behavior
At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points At 75th percentile cut points
M black I non-black 0.067 (0.029) 0.047 (0.037) -0.091 (0.037) -0.028 (0.040)
M black I black 0.011 (0.035) 0.020 (0.046) -0.042 (0.045) -0.074 (0.049)
Bias test 0.195 0.626 0.017 0.437
At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points At 90th percentile cut points
M black I non-black 0.033 (0.021) 0.009 (0.022) 0.031 (0.020) -0.020 (0.022) -0.004 (0.026)
M black I black -0.017 (0.025) -0.005 (0.027) 0.017 (0.025) -0.016 (0.027) -0.077 (0.318)
Bias test 0.111 0.626 0.667 0.888 0.056
53

• Conclusions
• There are racial differences in self-reported and
  interviewer-assessed parenting behaviors.
• For many (although not all) measures, racial
  differences are accounted for by household and
  maternal characteristics.
• The race of the interviewer matters for
  interviewer assessments of parenting.
• There is some evidence of racial bias in
  interviewer assessments of parenting measures.