Disentangling Ethnic and Contextual Influences Among Parents Raising Youth in High-Risk Communities

Abstract

This article reports on analyses examining contextual influences on parenting with an ethnically and geographically diverse sample of parents (predominantly mothers) raising 387 children (49% ethnic minority; 51% male) in high-risk communities. Parents and children were followed longitudinally from first through tenth grades. Contextual influences included geographical location, neighborhood risk, SES, and family stress. The cultural variable was racial socialization. Parenting constructs created through the consensus decision-making of the Parenting Subgroup of the Study Group on Race, Culture, and Ethnicity (see Le et al., 2008) included Monitoring, Communication, Warmth, Behavioral Control and Parenting Efficacy. Hierarchical regressions on each parenting construct were conducted for each grade for which data were available. Analyses tested for initial ethnic differences and then for remaining ethnic differences once contextual influences were controlled. For each construct, some ethnic differences did remain (Monitoring, ninth grade; Warmth, third grade; Communication, kindergarten; Behavioral Control, eighth grade; and Parenting Efficacy, kindergarten through fifth grade). Ethnic differences were explained by contextual differences in the remaining years. Analyses examining the impact of cultural influences revealed a negative relation between racial socialization messages and Communication or Monitoring.

Fast Track is a randomized longitudinal clinical trial targeting adolescent problem behaviors and conduct disorder. The longitudinal study is grounded in a developmental model (CPPRG, 1992) positing multiple influences—neighborhood risk, problematic parenting, chaotic classroom settings, and antisocial peers—that converge during the course of development to promote an escalation of problem behaviors that culminate in deviant and risky adolescent behaviors. The randomized clinical trial featured the ten-year delivery of a comprehensive multicomponent intervention to three successive cohorts of children recruited from high risk communities and selected for high levels of behavioral risk when entering first grade (CPPRG, 2004).

A second goal of Fast Track was to examine the normative development among a group of children with varying problems who grow up in high-risk communities. This latter sample of African American and European American families was the focus of analyses reported here. These analyses were conducted as part of the collaborative project on ethnic differences, cultural processes, and parenting outcomes (Le et al., this issue) in which common analyses were conducted on consensually derived constructs with different datasets. With this sample, the following questions were addressed: 1) Are there ethnic differences in parenting that remain when contextual variables are controlled? and 2) When ethnic differences remain, do cultural processes predict parenting outcomes?

METHOD

Participants

The participants in Fast Track were recruited in four sites in the United States: Durham, North Carolina (N = 100), Nashville, Tennessee (N = 100), rural central Pennsylvania (N = 100), and Seattle, Washington (N = 87). Within each of the four sites, schools serving communities with historically high rates of crime and poverty were identified. These schools were then grouped into multiple paired sets that were matched for demographics (ethnic composition, size, percentage of children receiving free or reduced lunch), and one from each pair was randomly assigned to intervention and control conditions. Using a multiple-gating screening procedure (Lochman & the Conduct Problems Prevention Research Group, 1995; the Conduct Problems Prevention Research Group, 2006), all kindergarten children in three successive cohorts were screened for inclusion in the study. After teacher reports of disruptive classroom behavior (using the Teacher Observation of Child Adjustment–Revised [TOCA-R] Authority Acceptance Score; Werthamer-Larsson, Kellam, & Wheeler, 1991) were obtained, parents of children scoring in the top 40% within cohort and site reported on home-based problem behaviors (using the Child Behavior Checklist [Achenbach, 1991] and similar scales). A final sample of high-risk children was selected, using a standardized combination of teacher and parent scores and specific decision rules for inclusion (see Conduct Problems Prevention Research Group, 2007 for more details). The schools in the control condition served as the source for the normative sample.

Children in the normative sample were selected to represent the site population in regard to gender, race, and level of severity of behavior problems. At each site, children were randomly selected from each decile of the distribution of scores on a teacher- and parent-report screen for behavior problems, resulting in a sample with a range of risk for chronic conduct problems. Of the 387 children comprising the initial sample, 51% were male, 49% were ethnic/racial minorities (predominantly African American), 60% were from one of the two lowest SES groups as classified by the Hollingshead system (1979), and 40% resided in single-parent families. The mean age at the start of the study (entrance into first grade) across the sample was 6 years, 4 months (SD = 5 months).

Procedure

Data collection

Two interviewers met with parents and children separately once yearly during home visits, which were conducted each summer, starting after kindergarten, and through the post-tenth grade year. During these two-hour home visits, youth adjustment, parenting, family supports, and stressors were assessed via youth and parent reports, as well as interviewer report. Interviewers received extensive cross-site training and met high rating reliability standards before meeting with parents on their own. Most interviewers were research staff whose racial composition matched that of the communities in which they worked. An additional source of data was census-based ratings of neighborhood quality.

Measures

Data from kindergarten through grade ten were used in this project. Once the items were clustered for each construct based on the consensus model, scores were standardized and internal consistencies for each subscale were assessed for each year of administration. Only those new scales whose reliabilities were over .60 were included in analyses. In some years, certain measures were not administered. Table 1 presents the constructs, year measured, and reliabilities. For years in which certain context measures (e.g., neighborhood) were not administered, the closest previous year's data were used.

TABLE 1.

Reliabilities and Number of Items for Consensualiy Derived Constructs

	Neighborhood		Stress		Monitoring		Warmth/support		Communication		Behavior Control		Self-efficacy
	Alpha	N	Alpha	N	Alpha	N	Alpha	N	Alpha	N	Alpha	N	Alpha	N
Kindergarten	.79	17	.63	9			.67	5	.72	13	*		.79	13
Grade 1	.79	17	.60	13			.70	5	.77	13	*		.80	13
Grade 2	.81	17	.60	11			.71	5			*		.80	13
Grade 3			.72	12			.70	5			*
Grade 4			.72	12	.66	6	.74	12	.71	22	.74	10	.81	14
Grade 5			.72	12	.67	9	.74	12	.82	22	.74	10	.82	14
Grade 6			.73	12
Grade 7	.72	12	.71	12	.74	9	.66	7	.78	22	.74	8
Grade 8	.72	12	.69	12	.79	9	.66	7	.86	22	.73	8
Grade 9			.62	10	.78	9			.80	22
Grade 10	.71	12	.72	12	.80	9

Blank cells: Items in consensual construct not assessed that year.

^*Construct not used; reliability fell below threshold.

Ethnicity was assessed by parent report. For this study, only those parents who were African American and European American were included (N = 368).

Parenting Variables

Items from several measures were included in the consensus model of parenting constructs and variables.

Monitoring was assessed with nine items from the child-report Supervision Questionnaire (Doyle & McCarty, 2000). This measure was adapted from the Supervision/Involvement Scale of the Pittsburgh youth Study (Loeber et al., 1998).

Warmth was assessed using six items from the Parent Questionnaire, a parent report measure (Strayhorn & Weidman, 1988), three items from the Post-Visit Reaction Inventory, completed by the interviewer (CPPRG, 1990), and in grades four and five, three items from the Parenting Scale (CPPRG, 1994), which was adapted from the Positive Parenting Scale, a parent report measure (Loeber et al., 1991).

Communication was assessed in kindergarten and first grade using eight items from the parent-report Family Expressiveness Questionnaire (CPPRG, 1995), adapted from Halberstadt, 1986), and five items from the Post-Visit Inventory (Dodge, Bates, & Pettit, 1990). From fourth grade on, items from other measures were used: seven items from Parent-Child Communication, parent and child report (Doyle & McCarty, 2001), five items from the Post Visit Inventory (Dodge et al., 1990), five items from the Parenting Scale (grades four through seven), and in fifth and eighth grades, five items from the Problem Solving Discussion Rating (OSLC, 1992).

Behavioral Control was assessed with the Interaction Ratings Scale (Crnic & Greenberg, 1990); Being a Parent (CPPRG, 1990); and Parent-Child Communication, parent and child report (CPPRG, 1994). Once the items were clustered for each construct based on the consensus model, internal consistencies for each sub-scale were assessed for each year of administration. Only those new scales whose reliabilities were over .60 were included in analyses. In general, internal consistencies were between .70 and .88, with a few between .60 and .70.

Parental Self-Efficacy was assessed with 13–14 items from Being a Parent (CPPRG, 1990), which was adapted from the Parenting Sense of Competence Scale (Johnston & Mash, 1989). Parents completed this measure following kindergarten through second, fourth, and fifth grades.

Contextual Variables

Contextual variables derived by the consensus model included items from several measures. Socioeconomic status was assessed using the Hollingshead (1975) scale, which combines occupational prestige and educational levels into a nine-point range. Geographic location was included (four locations: Nashville, Tennessee, Durham, North Carolina, rural central Pennsylvania, and Seattle, Washington).

Neighborhood context was assessed using items from multiple measures. First, census-based ratings of neighborhood quality (percent of poverty, unemployment, residential instability, receiving public assistance, renting residents, female-headed households) provided an objective measure. Second, 11 items from three sub-scales, Safety (e.g., “How often are there problems with muggings, assaults, burglaries… ?”), Social Involvement (e.g., “How many neighbors do you know well enough to visit?”), and Services (e.g., “How satisfied are you with the schools?”) of the parent-reported Neighborhood Questionnaire (Conduct Problems Prevention Research Group, 1991), were aggregated into a composite score. This composite included eight four-point, and one each of three-point, five-point, and six-point items.

Family stress was measured with the nine-item Stress subscale from the Life Changes measure (Dodge et al., 1990), a measure of cumulative total of life changes (maximum: 16) experienced by the family. In addition, the nine-point parent-report Financial Stress Measure was used. Parents also reported depressive symptoms on the Center for Epidemiological Studies Depression Scale (Radloff, 1977).

Cultural Variable

Two measures were used to assess cultural processes. The Teenager Experience of Racial Socialization (Stevenson, 1996) was administered to youth following eighth grade. This measure has four subscales: Coping with antagonism, Cultural appreciation of legacy, Cultural alertness to discrimination, and Cultural endorsement of mainstream. These subscales displayed good internal consistency, with alphas above .70. Religiosity (developed for the original project; CPPRG, 1998) was assessed with three items (two on participation in religious institution and one on frequency of prayer).

RESULTS

Consistent with the consensual analytic model of Le et al. (this issue), the first question (“Are there ethnic differences in parenting that remain when contextual variables are controlled?”) was addressed with hierarchical regressions run in the following manner: Step 1 included ethnicity; Step 2 included contextual variables. The findings for Step 1 are reported in the text; the findings for Step 2 are presented in Tables 2 through 6. The second question (“When ethnic differences remain, do cultural processes predict parenting outcomes?”) was addressed with hierarchical regressions run (with African Americans only) in the following manner: Step 1 included cultural processes; step two included contextual variables. These within group analyses were conducted for grades 8–10 only, due to the availability of cultural processes data for this age group. The findings for Step 1 are reported in the text; the findings for Step 2 are presented in Table 7.

TABLE 2.

Monitoring^a

Parameter	Grade 4		Grade 5		Grade 7		Grade 8		Grade 9		Grade 10
Parameter	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.
AA (aa = 1)	−0.07	(.16)	−0.15	(.17)	−0.03	(.17)	−0.31+	(.14)	−0.31*	(.15)	−0.24	(.20)
Neighborhood	0.02	(.08)	−0.02	(.08)	0.03	(.07)	−0.06	(.06)	−0.02	(.07)	−0.04	(.06)
SES	−0.10*	(.07)	−0.02	(.06)	−0.04	(.07)	−0.01	(.07)	0.02	(.06)	−0.08	(.06)
Stress	−0.07	(.06)	−0.01	(.07)	−0.17+	(.08)	−0.12+	(.07)	−0.04	(.07)	−0.06	(.07)
Depression	−0.21**	(.06)	−0.24**	(.07)	−0.10	(.07)	−0.06	(.06)	−0.09	(.06)	−0.13+	(.08)
Site (DURH)	−0.19	(.17)	−0.10	(.18)	0.14	(.18)	0.19	(.18)	0.14	(.18)	0.00	(.21)
Site (NASH)	−0.07	(.17)	−0.01	(.19)	−0.07	(.17)	−0.02	(.17)	−0.06	(.18)	−0.04	(.18)
Site (PENN)	−0.19	(.17)	−0.41*	(.18)	−0.02	(.17)	0.19	(.17)	0.14	(.18)	−0.03	(.18)
F	3.06**		3.52**		2.70**		2.99**		2.51**		2.53*
Degrees of freedon	359		359		359		359		559		359
R2	0.06		0.07		0,06		0.06		0.05		0.05
Adjusted R2	0.04		0.05		0.04		0.04		0.03		0,03

^aTable contains results of step two in two-step regression analyses.

^*p < .05,

^**p < .01.

Significant parameters are bolded for ease of viewing.

TABLE 6.

Self-efficacy^a

Parameter	Kindergarten		Grade 1		Grade 2		Grade 4		Grade 5
Parameter	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E
AA (aa = l)	0.30*	(.15)	0.24	(.15)	0.34*	(.14)	0.38**	(.15)	0.45**	(.15)
Neighborhood	0.15*	(.07)	0.18**	(.07)	0.21**	(.07)	0.01	(.07)	0.10	(.07)
SES	−0.01	(.06)	−0.09	(.06)	−0.07	(.06)	−0.07	(.06)	−0.02	(.06)
Stress	0.02	(.05)	0.04	(.05)	0.06	(.05)	−0.17**	(.06)	−0.13*	(.06)
Depression	−0.24**	(.06)	−0.31**	(.06)	−0.25**	(.06)	−0.24**	(.06)	−0.26**	(.06)
Site (DURH)	0.28+	(.16)	0.27+	(.16)	0.19	(.16)	0.11	(.16)	0.16	(.17)
Site (NASH)	0.04	(.16)	−0.01	(.16)	−0.01	(.16)	−0.11	(.16)	−0.16	(.16)
Site (PENN)	−0.03	(.16)	−0.26	(.16)	−0.24	(.16)	0.11	(.17)	0.03	(.16)
F	6.05**		9.50**		8.45**		6.39**		8.46**
Degrees of freedom	359		359		359		359		359
R2	0.12		0.17		0.16		0.12		0.16
Adjusted R2	0.10		0.16		0.14		0.11		0.14

^aTable contains results of step two in two-step regression analyses.

^*p > .05,

^**p < .01

Significant parameters arr bolded for ease of viewing.

TABLE 7.

Within Ethnicity: Monitoring, Warmth, Communication, Behavioral Control^a

	Monitoring Grade 8		Monitoring Grade 9		Monitoring Grade 10		Warmth Grade 8		Communication Grade 8		Communication Grade 9		Behavior Control Grade 8
Parameter	B	S.E.	3	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.
TERS—coping with antagonism	0.10	.17	0.22+	.16	0.12	.16	0.03	.14	0.06	.13	0.17	.14	0.01	.13
TERS—cultural appreciation of legacy	0.14	.17	0.03	.18	0.12	.16	−0.15	.16	−0.08	.15	0.04	.15	0.03	.16
TERS—cultural alertness to discrimination	−0.37*	.16	−0.20+	.11	−0.20+	.12	0.08	.11	−0.23*	.10	−0.19+	.10	0.03	.22
TERS—cultural endorsement of mainstream	0.14	.12	0.00	.12	0.15	.14	0.08	.11	0.18+	.10	0.18	.13	−0.02	.10
Religiosity	0.03	.13	0.22*	.09	0.17	.11	0.12	.10	−0.05	.10	0.17+	.10	0.00	.11
Neighborhood	−0.04	.09	0.00	.09	−0.09	.09	0.01	.10	−0.10	.09	−0.11	.09	0.05	.09
SES	0.01		0.02	.12	−0.05	.08	−0.04	.11	0.00	.09	0.02	.08	−0.02	.10
Stress	−0.13		−0.04	.12	−0.01	.11	0.09	.12	−0.06	.11	0.01	.09	0.06	.10
Depression	−0.15		−0.13	.09	−0.08	.11	−0.12	.10	−0.32**	.11	−0.26**	.09	−0.03	.11
Site—Durham	−0.22		−0.11	.28	−0.45+	.25	−0.02	.23	−0.18	.22	0.05	.22	−0.40+	.24
Site—Nashville	−0.60+		−0.44	.35	−0.70*	.31	−0.65*	.27	−'0.50+	.27	−0.18	.27	−0.71*	.28
F	3.69**		3.04**		2.93**		3.69**		4.59**		3.60**		1.67
Degrees of freedom	156		156		156		156		156		156		156
R2	0.206		0.174		0.171		0.204		0.242		0.202		0.105
Adj. R2	0.150		0.116		0.112		0.148		0.189		0.146		0.042

^aTable contains results of step two in two-step regression analyses.

^*p > .05,

^**p < .01

Significant parameters arr bolded for ease of viewing.

Monitoring

Across-group regressions were conducted for six grades (fourth, fifth, and seventh through tenth). Ethnicity significantly predicted child-reported monitoring in eighth through tenth grades (Beta ranges from −.27 to −.33, p<.05). African American parents engaged in less monitoring than did European American parents. When contextual variables were entered in Step 2 (see Table 2), ethnicity remained a significant predictor only in ninth grade. Significant contextual influences included SES (fourth grade), and depression (fourth and fifth grades).

Follow-up within group analyses with African Americans assessed the impact of cultural processes (four racial socialization subscales) on monitoring in eighth through tenth grades. In Step 1, Cultural Alertness to Discrimination was a significant predictor (B = −.30, p<.05) in eighth grade, and religiosity was a significant predictor in ninth grade (B = .25, p<.05). After contextual variables were included, Cultural Alertness to Discrimination remained significant in eighth grade (B = .37, p<.0l), and religiosity remained significant in ninth grade (B = .22, p < .05; see Table 7). Parents providing fewer cultural socialization messages about discrimination and whose teens were more religiously active were engaged in more monitoring behavior.

Warmth

Across-group analyses were conducted for eight grades, kindergarten through fifth grade, seventh and eighth grades. Initial ethnic differences were found in kindergarten, third, and seventh grades (Betas ranged from .23 to .35.). African American parents displayed more warmth than did European American parents in all grades. When contextual variables were included in the second step, only ethnicity in third grade (B = .41, p<.01) remained as a significant predictor (see Table 3). Contextual variables that were significant included: neighborhood quaiity, in kindergarten through third, and seventh grades (Betas ranged from .20 to .31, p<.05); SES in kindergarten through third grades (Betas ranged from −.13 to −.19, p<.05); and depression in all grades except second (Betas ranged from −.13 to −.28, p<.05). These findings reflect that living in better neighborhoods and with less depression was associated with higher levels of warmth. Surprisingly, mothers with children in grade three and under who were living in lower SES families displayed more warmth.

TABLE 3.

Warmth^a

Parameter	Kindergarten		Grade 1		Grade 2		Grade 3		Grade 4		Grade 5		Grade 7		Grade 8
Parameter	B	S.E	B	S.E.	B	S.E.	B	S.E.	B	S.E	B	S.E.	B	S.E.	B	S.E.
AA (aa = 1)	0.18	(.15)	0.09	(.15)	0.16	(.15)	0.41**	(.15)	0.13	(.15)	0.07	(.15)	−0.19	(.14)	0.13	(.15)
Neighborhood	0.22**	(.07)	0.31**	(.07)	0.22**	(.07)	0.20**	(.07)	0.07	(.07)	0.12	(.07)	0.20**	(.07)	0.08	(.07)
SES	−0.13*	(.06)	−0.14*	(.06)	−0.19**	(.06)	−0.19**	(.06)	−0.10+	(.06)	−0.06	(.06)	0.07	(.08)	−0.06	(.07)
Stress	0.08	(.05)	0.01	(.05)	0.07	(.06)	−0.03	(.06)	−0.11+	(.06)	−0.08	(.06)	−0.11	(.07)	−0.04	(.07)
Depression	−1.12*	(.06)	−0.13*	(.06)	−0.12+	(.06)	−0.14*	(.06)	−0.18**	(.06)	−0.28**	(.06)	−0.22**	(.06)	−0.20**	(.06)
Site (DURH)	0.49**	(.06)	0.57**	(.16)	0.49**	(.16)	0.30+	(.16)	0.31*	(.17)	0.47**	(.17)	0.66**	(.16)	0.18	(.17)
Site (NASH)	0.16	(.16)	0.35*	(.16)	0.30+	(.16)	0.20	(.16)	−0.11	(.17)	0.03	(.17)	0.26	(.16)	−0.24	(.17)
Site (PENN)	−0.04	(.16)	−0.19	(.17)	0.10	(.16)	0.07	(.17)	−0.12	(.17)	0.12	(.06)	0.23	(.16)	−0.17	(.18)
F	5.38**		6.83**		5.17**		5.26**		4.47**		8.09**		12.69**		5.24**
Degrees of freed	om 359		359		359		359		359		359		359		359
R2	0.11		0.13		0.10		0.11		0.09		0.15		0.22		0.10
Adjusted R2	0.09		0.11		0.08		0.09		0.07		0.13		0.20		0.08

^aTable contains results of step two in two-step regression analyses.

^*p < .05,

^**p < .01

Significant parameters are bolded for ease of viewing.

Follow-up within group analyses with eighth grade only were conducted. In each step, although the regression was significant, no cultural process variables significantly predicted warmth (Table 7).

Communication

Initial regressions were conducted on communication in seven grades: kindergarten, first, fourth, fifth, and seventh through ninth. Ethnic differences favoring African American parents were evident only in kindergarten (B = .22, p<.05). African American ethnicity remained significant in kindergarten (B = .33, p<.05) after contextual variables were included in step two (see Table 4). The contextual variables that were significant included: depression in all seven years (Betas ranged from -.18 to -.39, p< .05) and neighborhood quality in fifth grade (B = .13, p<.05). Mothers who were less depressed and lived in better neighborhoods engaged in higher quality communication with their children.

TABLE 4.

Communication^a

Parameter	Kindergarten		Grade 1		Grade 4		Grade 5		Grade 7		Grade 8		Grade 9
Parameter	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.	B	S.E.
AA (aa = l)	0.33*	(.15)	0.00	(.16)	−0.01	(.16)	0.21	(.16)	−0.07	(.14)	0.15	(.16)	0.09	(.14)
Neighborhood	0.07	(.07)	0.09	(.07)	0.01	(.07)	0.13*	(.08)	0.04	(.07)	0.06	(.06)	0.01	(.06)
SES	0.08	(.06)	0.01	(.06)	0.10	(.06)	0.01	(.06)	0.06	(.06)	0.03	(.07)	0.00	(.06)
Stress	−0.02	(.05)	−0.07	(.06)	−0.01	(.06)	−0.01	(.06)	−0.06	(.07)	−0.09	(.08)	−0.02	(.07)
Depression	−0.18*	(.06)	−0.23**	(.06)	−0.26**	(.06)	−0.24**	(.06)	−0.39**	(.06)	−0.25**	(.06)	−0.27**	(.06)
Site (DURH)	0.11	(.36)	0.41*	(.16)	−0.12	(.17)	−0.02	(.17)	0.33	(.17)	0 13	(.18)	0 10	(.17)
Site (NASH)	−0.13	(.36)	0.14	(.16)	−0.12	(.16)	−0.30+	(.17)	−0.10	(.17)	−0.16	(.16)	−0.11	(.16)
Site (PENN)	−0.03	(.16)	−0.04	(.17)	−0.07	(.17)	−0.40*	(.17)	−0.03	(.15)	0.03	(.17)	0.21	(.17)
F	4.33**		4.23**		5.36**		7.34**		13.45**		6.95**		5.55**
Degrees of Freedom	359		359		359		359		359		359		359
R2	0.09		0.09		0.11		0.14		0.23		0.13		0.11
Adjusted R2	0.07		0.07		0.09		0.12		0.21		0.11		0.09

^aTable contains results of step two in two-step regression analyses.

^*p > .05,

^**p < .01

Significant parameters arr bolded for ease of viewing.

Within-group analyses for African American parents were conducted on eighth and ninth grades. Initial regressions with cultural variables revealed no significant influences on communication. However, when contextual variables were included, cultural alertness to discrimination emerged as a significant predictor (B = −.23, p<.Q5), indicating that fewer messages about anticipating discrimination were associated with higher levels of communication (Table 7). Of the contextual variables, depression (Betas = −.32 and −.26, respectively) significantly predicted communication: mothers with less depression had higher levels of communication.

Behavioral Control

Regressions were conducted on behavioral control in fourth, fifth, seventh and eighth grades. Ethnic differences emerged in fourth and eighth grades (Betas = .22, 37, respectively). African American parents engaged in more behavioral control. After contextual influences were entered in step two, only ethnic differences in grade eight (B = .36, p<.05) remained significant (see Table 5). Contextual variables that were significant included SES in fourth and fifth grades (Betas −.16, −.13, respectively); depression in fourth, fifth, and eighth grades (Betas ranged from − .19 to −.23, p<.05); and neighborhood quality in seventh grade (B = .16, p<.05). Mothers in low-SES families, who were less depressed and living in better neighborhoods, used more behavioral control. A within-group analysis was conducted only for eighth grade; this analysis was not significant in either step (Table 7).

TABLE 5.

Behavioral Control^a

Parameter	Grade 4		Grade 5		Grade 7		Grade 8
Parameter	B	S.E	B	S.E	B	S.E.	B	S.E.
AA (aa = l)	0.06	(.15)	0.13	(.15)	0.15	(.15)	0.36*	(.17)
Neighborhood	0.01	(.08)	0.08	(.07)	0.16*	(.07)	0.09	(.07)
SES	−0.16**	(.06)	−0.13*	(.06)	−0.04	(.08)	−0.07	(.08)
Stress	−0.10+	(.06)	−0.07	(.06)	−0.15*	(.07)	−0.04	(.09)
Depression	−0.19**	(.07)	−0.23**	(.07)	−0.12+	(.06)	−0.19**	(.07)
Site (DURH)	0.39*	(.17)	0.32+	(.18)	0.56**	(.17)	0.05	(.17)
Site (NASH)	0.28+	(.16)	0.28	(.19)	0.26	(.18)	−0.02	(.18)
Site (PENN)	−0.01	(.17)	0.01	(.18)	0.05	(.16)	−0.14	(.17)
F	3.93**		3.94**		6.14**		4.93**
Degrees of freedom	359		359		359		359
R2	0.08		0.08		0.12		0.10
Adjusted R2	0.06		0.06		0.10		0.08

^aTable contains results of step two in two-step regression analyses.

^*p > .05,

^**p < .01

Significant parameters arr bolded for ease of viewing.

PARENTAL EFFICACY

Regressions were conducted on parental self-efficacy in kindergarten through second, fourth, and fifth grades. Ethnic differences favoring African American parents were found in all grades but fourth, where a trend was found. (Betas ranged from .23 to .29, p<.05.) When contextual variables were entered, significant ethnic differences were found in all grades but first (Betas ranged from .30 to .45, p<.05; see Table 6). Contextual influences that emerged as significant included neighborhood in kindergarten through second grade (Betas ranged from .15 to .21) and depression in all five grades (Betas ranged from −.24 to −.31). Living with less depression and in better neighborhoods were associated with feelings of higher self-efficacy. No within-group analyses were run on this variable.

SUMMARY

In sum, in this data set of youth growing up in high-risk communities, regressions were conducted on five parenting outcomes across multiple years (ranging from 4 years – efficacy – to 8 years – warmth). Initial ethnic differences were found in monitoring (grades 8, 9, 10); warmth (grades k, 3, 7), communication, (k), behavioral control (grades 4, 8), and efficacy (k, 1, 2, 4, 5). The bolded years represent those in which ethnic differences remained significant when context variables were included. African American mothers monitored less and displayed more warmth, better communication, more behavioral control, and higher self-efficacy. Ethnic differences were explained by contextual influences among the remaining years. Thus, these results indicate that ethnicity and contextual influences have a different impact on parenting, depending upon the specific parental behavior under investigation and the child's developmental level. Within group analyses examining .. the impact of cultural processes were run only in grades 8 through 10. Cultural Alertness to Discrimination was a significant predictor of monitoring (grade 8) and communication (grade 8). In both cases, parents who monitored or communicated less had youth who reported receiving more racial socialization messages. It is important to note that although ethnic (and in two cases, cultural processes) differences were significant predictors of several parenting outcomes, in some cases (notably, monitoring) the percent of variance explained by these differences was low. Perhaps ethnic group membership is not sufficiently sensitive to capture differences in these specific parenting behaviors that may vary due to ethnicity; other variables not measured in this study may more be sensitive to variations in parenting. Indeed, within ethnic-group, differences in parenting may be more meaningful than between ethnic-group differences. In light of this, more refined exploration of cultural differences in parenting may yield a more nuanced understanding of between- and within-ethnic group differences in parenting.