Gender Differences in the Developmental Cascade from Harsh Parenting to Educational Attainment: An Evolutionary Perspective

Abstract

This study utilized life history theory to test a developmental cascade model linking harsh parenting to low educational attainment. Multi-group models were examined to test for potential gender differences. The sample consisted of 1,482 adolescents followed for nine years starting in 7^th grade (M_age = 12.74). Results supported indirect links between harsh parenting and low educational attainment through the development of extreme peer orientations, early sexual behavior and delinquency. Among males, harsh parenting was related to the development of an extreme peer orientation, which further led to increased delinquency, and subsequently lower educational attainment. Among females, harsh parenting predicted extreme peer orientations, which increased both delinquency and early sexual behavior. Early sexual behavior further predicted lower educational attainment in females.

Two out of every three people in the United States disengage from formal education before obtaining a college degree (National Center for Education Statistics, 2013), thereby divesting many Americans of the benefits correlated with higher educational attainment, such as increased income over the lifespan, greater life satisfaction, and lower rates of criminal behavior (Baum, Ma, & Payea, 2013; Cutler, Deaton, & Lleras-Muney, 2006; Sabates, 2007). Children exposed to early adversity, including harsh parenting and maltreatment, are at a greater risk of having problems in school and dropping out of the formal education system (Alexander, Entwisle, & Horsey, 1997; Jimerson, Egeland, Sroufe, & Carlson, 2000). Despite the acknowledgement that school dropout rates pose a significant public health concern (Organization for Economic Co-operation and Development [OECD], 2010), little is known about the developmental processes and mediating mechanisms that account for why early exposure to harsh parenting relates to poor school outcomes. While researchers in the education and psychology fields have yet to propose a clear theoretical framework that explains the multitude of risk factors associated with early dropouts, Battin-Pearson and colleagues (2000) identified five potential predictor models of early high school dropout: (a) general deviance, (b) deviant peer affiliation, (c) low school bonding and academic achievement, (d) poor family socialization, and (e) structural strains (e.g., gender, race, and socioeconomic status). However, no one single model could account for school dropout rates, as all theories partially explained the data. Additionally, these models were largely differentiated based on structural and organizational features of the environment (e.g., family versus peer versus academic contexts) rather than the functional goal of each system.

Recent advances in evolutionary theories of development have led to calls within the developmental field to examine both the distal and functional causes of more proximal development (Bjorkland & Ellis, 2014). Specifically, evolutionary life history theory offers a promising approach by which to address current gaps in the field and develop an integrated theory of how family, peer, and individual level factors unfold in predicting educational attainment. In this study, we applied the evolutionary developmental theory of life history strategies to examine a novel set of hypotheses predicting a developmental cascade that links harsh parenting to low educational attainment.

Life History Theory

According to life history theory, evolution favors the adoption of a diverse set of strategies aimed at capturing and allocating resources and energy in order to maximize the fitness between an individual and their ecological domain (Del Giudice, Gangestad, & Kaplan, 2015). Fitness goals can often be differentiated along the lines of reproductive goals (i.e., replication of genes) and individual somatic goals (i.e., acquiring resources in order to protect and enhance the individual’s current development and fitness) (Del Giudice, 2014; Geary, 2002). Somatic goals also include the development of embodied capital, which is the acquisition of strength, coordination, skills, and knowledge that can increase ultimate fitness and longevity (Del Giudice, 2014). Because resources and energy are finite, allocation of these assets towards different and often competing goals results in a trade-off of evolutionary benefits and disadvantages. For example, energy expended toward finding and securing mates advances evolutionary fitness through increased likelihood of reproduction; however, the increased energy spent on mating may come at the expense of time spent on other endeavors that could also benefit fitness in the long run, such as the amount of time spent on increasing knowledge and education. While a focus on mating and reproduction may offer short-term advantages for fitness, the allocation of personal resources towards education may have the potential for long-term benefits, perhaps most notably a potential increase in income. Income itself allows for greater access to other resources, including higher quality food, mates, and health care. Yet, by focusing on these long-term benefits, there is generally a trade-off of delayed reproduction. With such a delicate balance at hand, how does one decide whether to pursue short-term gains (e.g., increased offspring) or long-term rewards (e.g., increased health and longevity)?

To answer this question, life history theory proposes that an individual’s early rearing environment may provide clues as to the most beneficial life strategy, thereby priming individuals towards either slow or fast life strategies (Del Giudice et al., 2015). Over time, the slow and fast life strategies have coalesced into correlated suites of morphological, biological, and behavioral differences that determine how individuals allocate resources (Del Giudice, 2015). In stable environments with adequate access to interpersonal and material resources, a slow life strategy that delays reproduction and expends energy toward pursuits with long-term rewards may be the most beneficial for evolutionary fitness (Ellis et al., 2012). Because the slow strategy is inherently future-oriented, conditions where immediate survival is not endangered allow individuals to invest more resources toward individual development and higher quality future offspring. In addition, individuals with a slow life strategy are further proposed to be risk-averse in order to increase their health and chances of survival for future reproduction (Del Giudice, 2014). Thus, environments that do not require or reward risky behaviors (e.g., fighting for limited resources) are particularly suitable for the slow life history strategy.

Conversely, a fast life strategy that focuses on early mating and larger quantity of offspring may be more favorable in rearing environments characterized by harshness, unpredictability, and signs of early mortality. As survival is uncertain in these chaotic environments, individuals must be attuned to rewards readily available within the environment; hence, energy expenditures switch toward accessing immediate rewards versus delayed gratification (Sturge-Apple et al., in press). As a result, individuals with fast life strategies are particularly susceptible to risk-taking and impulsivity (Bjorkland & Ellis, 2014). While the fast life strategy may proffer advantages in reproductive fitness, the proclivity towards immediate gratification, impulsivity, and risky behavior indicative of youth with a fast life strategy can create a mismatch between these individuals and the modern school setting (Richardson, Castellano, Stone, & Sanning, 2015).

Developmental Cascade Model

Below, we build on developmental theories of life history strategies during middle childhood and adolescence to present a developmental cascade model of mediating processes linking harsh parenting in middle childhood to low educational attainment by early adulthood (see Figure 1). While specific definitions and labels change across studies, harsh parenting comprises punitive and dysregulated behaviors and negative emotional reactivity aimed at the child, such as yelling, hitting, and coercive behaviors (Chang, Schwartz, Dodge, & McBride-Chang, 2003).

Figure 1.

Conceptual model of developmental cascade of effects from harsh parenting to low educational attainment.

Link 1: Harsh parenting and extreme peer orientation

Middle childhood marks a particularly salient period for the development of life history strategies, as information about the individual’s early social and physical environment informs the onset and coordination of biological, pubertal, and peer relationship processes involved in the expression of the fast and slow phenotypes (Del Giudice, 2015). During middle childhood and early adolescence, peer relationships become increasingly salient and friendships begin to fulfill key attachment functions that were once entrusted to parents (Nickerson & Nagle, 2005). While this is a normative transition, children reared by harsh parents receive an implicit message that they are unlovable and that others are undependable, which can lead to later insecure attachments with peers and a mistrustful orientation toward others (Belsky, Steinberg, & Draper, 1991). Del Giudice (2015) further argued that youth on the fast life history spectrum are more likely to develop insecure patterns of relating to peers, including a preoccupied style that seeks constant validation and closeness with their peers. Indeed, Nickerson and Nagle (2005) found that adolescents who did not have secure attachments to parents were more likely to turn to peers to fulfill attachment functions (e.g., proximity seeking, safe haven, and secure base).

In addition, issues of popularity and competition among peers heighten during middle childhood and early adolescence. For some children, increased pressure to belong to a peer group can result in developing an extreme peer orientation, a type of insecure orientation that over-emphasizes conformity with peers and detachment from parental figures (Fuligni & Eccles, 1993). Middle school children who perceive their parents as being more strict and less autonomy-granting are particularly likely to endorse high levels of extreme peer orientation (Fuligni & Eccles, 1993; Goldstein, Davis-Kean, & Eccles, 2005). Thus, as the first link of our developmental cascade model, we predict that youth who experience harsh parenting during the middle school years will develop an extreme peer orientation that signals an over-reliance on peer relations (see Figure 1).

Link 2: Extreme peer orientation and delinquency and early sexual behaviors

Studies have revealed that extreme peer orientation in middle school predicts greater problem behaviors and worse school performance in both middle and high school (Claes et al., 2005; Fuligni, Eccles, Barber, & Clements, 2001; Goldstein et al., 2005). Although the construct of extreme peer orientation has not been examined in relation to early sexual behavior, Zimmer-Gembeck, Siebenbruner, and Collins (2004) found that both highly impulsive temperaments in toddlerhood and friendships characterized by greater intimacy and satisfaction in middle childhood predicted earlier onset of romantic relationships, which in turn predicted greater number of sexual partners by age 19. In a sample of 11^th and 12^th graders, individuals who reported higher concerns with popularity, peer conformity, and peer pressure and lower concerns with conforming to authority figures (e.g., parents) were more likely to endorse sexual attitudes displaying a willingness to have sex, including in situations when they did not want to have it (Santor, Messervey, & Kusumakar, 2000).

Based on these empirical underpinnings, in the second link of the proposed developmental cascade model, we posit that extreme peer orientations are related to increased delinquent and sexual risk-taking behaviors (see Figure 1). According to Ellis and colleagues (2012), adolescence is a developmental turning point in which individuals become primed toward trajectories related to status, resource control, and mating. During adolescence, risky and aggressive behaviors are often implicitly rewarded as signals of bravery and toughness, therefore increasing social dominance, particularly among males (Ellis et al., 2012). Consequently, our model proposes that individuals with more extreme peer orientations (i.e., heightened concerns about conformity with peers and decreased interest in adult figures) might be particularly susceptible to the rewarding aspects of risky behaviors (e.g., peer regard). In addition, the reward-oriented behavior of individuals on the fast life spectrum might also reduce susceptibility to the potential aversive qualities of risky behaviors (e.g., suspension, trouble with the police). Finally, distancing from parental and other adult figures has been reported to orient individuals toward mating opportunities within the peer arena (Ellis et al., 2012); thus, we hypothesize that individuals who endorse extreme peer orientations may also report higher levels of problem behaviors and early sexual behaviors.

Gender differences

Evolutionary theories have additionally suggested that the peer group functions as a context absent of adult members of society (e.g., parents) in which the individual has the opportunity to be sexually active (Del Giudice, 2015; Ellis et al., 2012), but these peer groups operate differently across genders due to differences in reproductive opportunities (Bjorkland & Ellis, 2014; Trivers, 1972). Within females, the peer group is proposed to pull individuals away from adults and increase mating directly through the development of risky sexual behaviors (e.g., early onset of sexual behavior); however, males naturally have higher rates of maximal reproduction due to their lack of gestation and ability to sire more offspring within a shorter period of time (Bjorkland & Ellis, 2014). Since mating is inherently more risky and resource-depleting for females, the maximal reproduction rate of males is constrained by the availability of sexual partners. Thus, males must compete for reproductive partners. Indeed, females tend to be choosier about their sexual partners (Buss, 1989; Schmitt, Shackelford, & Buss, 2001) and value physical strength and the ability to acquire resources in potential mates (Buss, 1989; Ellis et al., 2012). Ellis and colleagues (2012) argued that involvement in delinquency and risky behaviors can be important signals that enhance reputations for bravery and toughness, thus increasing mating opportunities as well as status in the peer group. In support of the proposed function of aggression as a mating signal behavior, a series of experiments have found that adolescent males exposed to photographs of attractive females subsequently engage in more risky or aggressive behavior (Baker & Maner, 2008; Chang, Lu, Li, & Li, 2011; Wilson & Daly, 2004). Thus, we propose that an overemphasis on peer relations to the exclusion of adult norms and rules will increase males’ involvement in risky and aggressive behaviors as a way of signaling dominance and attractiveness as a mate; yet, this same over-reliance on peers and rejection of adult norms and rules will predict early sexual behavior among females.

Link 3: Delinquency and early sexual behavior and educational attainment

The final links of the hypothesized developmental cascade model assert that early sexual behavior and problem behaviors will predict low educational attainment (i.e., early departure from formal schooling). Although risky behaviors are considered adaptive strategies toward increasing dominance and status for individuals on the fast life history trajectory, many of these behaviors violate school norms, leading to higher discipline rates that can ultimately impede educational attainment (Richardson et al., 2015). While there is consistent support for the hypothesis that early problem behaviors lead to early school dropout rates (e.g., Battin-Pearson et al., 2000; Wang & Fredericks, 2014), research linking sexual behavior to educational attainment is relatively scarce, with the majority of research focusing on examining antecedents of risky or early sexual behavior. However, some correlational analyses have indicated that early adolescent sexual activity relates to lower academic performance and interests (Miller & Sneesby, 1988).

Gender differences

Increased investment in sexual relations and mating behaviors may pull resources away from educational goals, hence decreasing performance in school and leading to early attrition from formal education (Richardson et al., 2015). In addition, early sexual behavior increases risks for unplanned pregnancies, and females are disproportionately affected by pregnancies, as they must invest considerably more resources during the gestation period and, historically, the caregiving of offspring (Trivers, 1972). Therefore, we expect that early sexual behavior will be particularly detrimental to educational attainment for adolescent females.

The Current Study

The goal of the current study is to test a set of novel, precise hypotheses on the developmental cascade between harsh parenting and low educational attainment. To our knowledge, this is the first study to use life history theory as a theoretical framework for examining links between parenting and education. In addition, the prospective, longitudinal nature of the current study allows for a more holistic examination of the direct and mediating influences between harsh parenting and educational attainment over time. The time frame assessed in the current study also enables us to test current evolutionary theories about the importance of particular developmental periods for the emergence of insecure peer orientations, sexual initiation, and delinquent behavior.

Our specific research questions are: (a) How does harsh parenting relate to educational attainment over time?; (b) In what ways does it predict the development of extreme peer orientations, sexual behavior and delinquency, and how do these developments further predict educational attainment?; and (c) Do these direct and indirect links between harsh parenting and educational attainment differ for males and females?

To investigate these questions, we utilized a longitudinal data set that spanned nine years, beginning with youth in the 7^th grade and ending three years after the students’ expected high school graduation year. Drawing from life history theory, we hypothesize that (a) harsh parenting in 7^th grade will predict greater extreme peer orientation in 8^th grade; (b) extreme peer orientation, in turn, will predict early sexual behavior in females and greater delinquency in males in the 11^th grade; and (c) low educational attainment three years post-high school will be predicted by early sexual behavior among females and greater delinquency among males.

In order to rule out a competing theory that intelligence, poor academic performance, and low educational values account for later problem behavior and school attrition, we also included indices of standardized test performance, educational value beliefs, and grade point averages in the developmental model. For example, previous research has linked lower intelligence and grades to early sexual behavior (Halpern et al., 2000), while theories of criminology have suggested that low academic performance causes delinquency (Maguin & Loeber, 1996). However, much of this research has relied either on cross-sectional data or unidirectional effects linking early academics to later problem behaviors (Hoffmann, Erickson, & Spence, 2013). As such, we include multiple assessments of academic functioning and educational utility beliefs at different time points in order to provide a strong test of our hypothesis that early sexual behavior and delinquency directly predict lower educational attainment over and above early academic performance, school grades, and low educational values.

Method

Participants

This study used data from the Maryland Adolescent Development in Context Study (MADICS), which was designed to examine the influences of social contexts on adolescent academic and psychosocial development. MADICS is an ongoing longitudinal study that includes participants from a large county near Washington, DC. Twenty-three of the 25 middle schools in the county participated; one nonparticipating school catered to children with special education needs, and the other nonparticipating school was undergoing an extensive multiyear restoration project. Recruitment letters were sent to the homes of 1,700 7^th graders of select schools, requesting parental permission for their child’s participation in the study. Then if the 7^th grader assented to participating in the study, they were administered a 50-minute structured interview and a 30-minute self-administered questionnaire. At the first wave of data collection, 1,482 (87% of those originally contacted about the study) 7^th graders (M_age = 12.74; 51% male; 56% African American, 34% European American, 10% biracial or other ethnic minorities) were enrolled in the study.

Data from this sample were drawn from Waves 1, 3, 4, and 6, which were administered between 1991 and 2000. These waves were chosen so that we could adequately test our developmental hypotheses relating to the mediating processes emerging from middle childhood through adolescence and into early adulthood. Wave 2 was not used because it was a qualitative study of parental management strategies conducted at the end of 7^th grade. We chose to examine educational achievement in Wave 6 rather than Wave 5 in order to extend the variability in education levels obtained post-high school. Participants were in 8^th grade (M_age = 14.24) at Wave 3 (n = 1060) and 11^th grade (M_age = 17.07) in Wave 4 (n = 1057). Wave 6 (n = 899) took place three years after high school (M_age = 21.43). Participants reflected a broad sample of socio-economic backgrounds, with pre-tax incomes ranging from $5,000 to over $75,000 (M = $45,000 – $49,999). Education levels of the primary caregivers ranged from completing 5^th grade to graduate degrees, with 92.5% completing high school and 28.5% completing a bachelor’s degree. The location where participants resided is ecologically diverse, thus the resulting sample ranged from children living in urban, suburban and rural environments.

The retention rate over the nine-year period from Wave 1 to Wave 6 was 60.66%. In order to determine if there were group-level differences between those who completed the Wave 6 assessment and those who did not, a series of t-tests were conducted for all study variables assessed at Wave 1. Significant differences emerged between the two groups on several demographic variables, though the effect size was relatively small. Those who completed the assessment at Wave 6 had higher family incomes, more educated primary caregivers, and were more likely to have a European American than an African American background. Those who did not complete the Wave 6 assessment also had lower prior academic achievement than those who remained in the study at Wave 6. There were no differences in Wave 1 harsh parenting between those who did and did not complete the Wave 6 assessment.

Measures

Harsh Parenting

At Wave 1, youth reported on both their primary and secondary caregivers’ use of physical and verbal discipline. This construct was measured by six items adopted from the Conflict Tactics Scale (Straus, 1979), including two items assessing the primary caregiver’s use of harsh physical and verbal parenting, two items assessing the secondary caregiver’s use of harsh physical and verbal parenting, and two items that asked how their parents reacted when the child broke a rule. Example items include “During the past month, how often did your parent hit, push, grab, or shove you?”, “During the past month, how often did your secondary caregiver yell at you?”, and “When you break one of your parents’ rules, how often do they physically punish you?” Each item was scored on a scale of 1 (“almost never”) to 5 (“almost every day”). The six items were averaged together to create a composite scale of harsh parenting. The resulting scale showed adequate internal consistency (α = .66).

Extreme Peer Orientation

At Wave 3, adolescents reported on the extent to which their peer group was more important than other responsibilities (e.g., completing homework, following rules). This scale has been used in prior large-scale studies and has shown good internal consistency, as well as adequate construct and predictive validity (Fuligni & Eccles, 1993). Example questions from the five-item scale include “How often is it OK to break some of your parents rules in order to keep your friends?” and “How much does the amount of time you spend with your friends keep you away from doing things you ought to do?” Due to differences in item-level scales, each item was standardized and then averaged together to form a composite scale of extreme peer orientation (α = .66).

Academic Values

Perceived importance of academic skills was assessed at Wave 3 by two self-report items adopted from Eccles’ well-validated academic task value scale (Eccles, Lord, & Midgley, 1991). Adolescents reported on a scale of 1 to 7 how important they felt (a) math and (b) other school subjects were compared to other kids their age, with high scores reflecting greater perceived academic values. The two items (r = .67) were averaged together to form a single measure of academic values.

Early Sexual Behavior

At Wave 4, adolescents reported on (a) perceptions of their chances of becoming pregnant or getting someone else pregnant before finishing high school and (b) their chance of starting to have sex too young, on a scale of 1 (“very low”) to 6 (“already happened”) (r = .50 between two items). In order to create an index of early sexual behavior, the two items were dichotomized so that scores of 1–5 were re-coded as 0, and scores of 6 (i.e., “already happened”) were re-scored as 1 to indicate presence of the sexual behavior. The items were summed together to create a scale of early sexual behavior ranging from 0 to 2.

Delinquency

At Wave 4, problem behavior was assessed by seven items from a well-validated adolescent problem behavior scale (Elliott, Huizinga, & Menard, 1989) indexing the adolescent’s problems with aggression, stealing, and other antisocial behaviors. Example items included, “In the last 6 months, how often have you hit someone because you didn’t like what they said?” or “taken something from a store without paying for it?” Scores were assessed on a scale of 0 (“never”) to 5 (“more than 20 times”). The items were averaged to form a scale of delinquency (α = .74).

Academic Performance

Academic performance was assessed at Wave 4 through adolescents’ self-reported grades in their first semester of 11^th grade. Due to school-level differences in how GPAs were calculated, adolescents were asked to report how many As, Bs, Cs, Ds and Fs they received on their report card. Letter grades were converted into numeric scores (A = 4; B = 3; C = 2; D = 1; F = 0), which were then used to calculate each individual’s GPA.

Educational Attainment

At Wave 6, three years after high school, participants were asked to report the maximum grade level they had completed. Educational attainment was assessed on a scale of 1 to 9 as follows: 1 = 10^th grade; 2 = 11^th grade; 3 = high school or GED diploma; 4 = 1 year of vocational training; 5 = 2 years of vocational training; 6 = 1 year of college; 7 = 2 years of college or a 2-year degree; 8 = 3 years of college; 9 = Bachelor’s degree.

Covariates

Due to established relations with parenting, adolescent adjustment, and school functioning, we included family income, race, and maternal education as covariates in all analyses. In addition, prior academic ability was controlled for with standardized tests scores from the California Achievement Test taken during 5^th grade.

Analytic Strategy

We examined our hypothesized cascade model using a multi-group path analysis model, where groups were defined by gender, in the AMOS statistical software program (Arbuckle, 2009). In order to provide a conservative test of our hypotheses, we conducted analyses in a fully identified model in which: (a) harsh parenting and all covariates at Wave 1 were specified as predictors of extreme peer orientation and academic values at Wave 3, delinquency, early sexual behavior and GPA at Wave 4, and educational attainment at Wave 6; (b) extreme peer orientation and academic values were regressed on delinquency, early sexual behavior and GPA at Wave 4, and educational attainment at Wave 6; (c) early sexual behavior, delinquency, and GPA predicted educational attainment at Wave 6; and (d) correlations were estimated between all covariates and harsh parenting at Wave 1 and between all constructs measured at the same subsequent waves (e.g., extreme peer orientation and academic values at Wave 3). Because the model was fully identified, with path coefficients estimated for all possible predictors from previous waves and correlations estimated between all variables assessed at the same wave, model fit indices reflected a perfect fit for both the male and female models (χ² (0, N = 1493) = .00, RMSEA = .00, CFI = 1.00, TLI = 1.00).

To retain the full sample size, missing data (Median = 7.0%, Range = .00 – 61.5%) were estimated using full information maximum likelihood (FIML) for all primary analyses. FIML procedures do not impute scores for missing data but rather utilize the raw data to establish parameter estimates (see Enders & Bandalos, 2001). FIML is a widely accepted technique and is considered superior to listwise and pairwise deletion procedures as well as multiple imputation because it optimally maintains the original structural association (i.e., variance-covariance matrix) between variables in the model (see Enders, 2001). Results from Monte Carlo studies indicate FIML procedures provide reasonably accurate estimates of standard errors at up to 75% of missing data (Newman, 2003). Finally, in order to test for mediational effects in our developmental cascade, we utilized bootstrapping tests of mediation to examine all indirect effects (Tofighi & MacKinnon, 2011).

Results

Preliminary Analyses

Table 1 provides descriptive statistics and the bivariate correlations among the variables included in the study. With the exception of early sexual behavior and delinquency, skewness and kurtosis values were all within acceptable range (i.e., < 2.0 for skewness, < 3.0 for kurtosis). Early sexual behavior and delinquency were log transformed in order to minimize skewness and kurtosis and bring the variables closer to a normal distribution. The log transformation was successful in reducing skewness below 2.0, while kurtosis was reduced to 2.55 and 3.59 for early sexual behavior and delinquency, respectively. Structural equation models are generally considered robust to minor deviations from normality, and when estimating missing data, parameter estimates can be considered acceptable with univariate kurtosis values of up to 15 (Patterson & Stoolmiller, 1991). Thus, the log-transformed variables of early sexual behavior and delinquency were used in the subsequent analyses.

Table 1.

Means, Standard Deviations, Skewness, Kurtosis, and Correlations

	M	SD	Skewness	Kurtosis	1	2	3	4	5	6	7	8	9	10	11
1. Child Gender	1.49	.50
2. Child Race	.65	.48			−.04
3. Family Income	10.03	4.23	−.14	−1.05	.01	−.18*
4. Maternal Education	13.85	2.45	.55	.41	−.02	−.10*	.43*
5. Prior Academic Ability	493.63	52.74	.42	−.07	.12*	−.32*	.32*	.32*
6. W1 Harsh Parenting	1.98	.61	.81	.89	−.04	.14*	−.08*	−.04	−.10*
7. W3 Extreme Peer Orientation	.00	.73	.70	.07	−.17*	−.13*	.03	.00	−.12*	.20*
8. W3 Academic Values	5.14	1.29	−.41	.00	.00	.13*	−.03	.01	.05	−.06*	−.30*
9. W4 GPA	2.90	.73	−1.05	1.48	.23*	−.16*	.16*	.22*	.37*	−.10*	−.16*	.15*
10. W4 Early Sexual Behavior	.21	.49	2.34	4.73	.08*	.05	−.14*	−.14*	−.11*	.07*	.04	−.01	−.17*
11. W4 Delinquency	.34	.52	2.98	10.54	−.18*	.04	−.04	−.07*	−.12*	.11*	.31*	−.13*	−.27*	.15*
12. W6 Educational Attainment	5.99	2.16	−.69	−1.05	.22*	−.11*	.12*	.15*	.20*	−.06*	−.06	.02	.21*	−.04	−.14*

Note. Child gender: 1 = Male; 2 = Female. Race: 0 = White; 1 = Black.

^*p < .05

Prior to conducting our primary analyses, we first examined gender differences by carrying out a series of independent sample t-tests (see Table 2). As would be expected, there were no gender differences in family- and maternal-level covariates (e.g., income, maternal education). In addition, harsh parenting did not differ in regard to male and female subjects. There was a significant gender effect on prior academic ability, t(1118) = −4.046, p < .001; d = .24, with girls showing higher average scores (M = 500.00) on the California Achievement Test than boys (M = 487.29). There were also significant gender differences in later measures of academic achievement. Girls reported higher GPAs (M = 3.06) than boys (M = 2.73) at Wave 4, t(909) = −6.90, p < .001; d = .46, as well as slightly higher educational attainment by Wave 6, t(392.79) = −2.01, p < .05; d = .18, but there were no differences in the perceived value of academics between boys and girls. Girls were slightly more likely to engage in early sexual behavior, t(940.13) = −2.58, p < .05; d = .16. Conversely, boys endorsed higher levels of extreme peer orientation, t(1024.41) = 5.52, p < .001; d = .34, and delinquency, t(921.35) = 5.92, p < .001; d = .37, than girls. While most gender differences were robust (i.e., p < .001; moderate effect sizes), caution should be applied in interpreting these results due to increased likelihood of Type I errors. Specifically, the differences in early sexual behavior and educational attainment were small and were only significant at the p < .05 level, which warrants caution considering the number of analyses conducted.

Table 2.

Descriptive statistics for males and females and significance level of gender effects on primary variables

	Males		Females		T-test Result
	N	M (SD)	N	M (SD)
Family Income	705	9.98 (4.29)	684	10.07 (4.17)	ns
Maternal Education	753	13.90 (2.49)	722	13.80 (2.41)	ns
Prior Academic Ability	560	487.30 (53.56)	560	500.00 (51.17)	p < .001
W1 Harsh Parenting	748	2.00 (.60)	723	1.96 (.61)	ns
W3 Extreme Peer Orientation	532	.12 (.77)	512	−.12 (.65)	p < .001
W3 Academic Values	534	5.14 (1.29)	511	5.15 (1.29)	ns
W4 GPA	446	2.73 (.73)	465	3.06 (.70)	p < .001
W4 Early Sexual Behavior	487	.16 (.40)	512	.24 (.55)	p < .05
W4 Delinquency	513	.40 (.21)	534	.21 (.39)	p < .001
W6 Educational Attainment	205	5.73 (2.27)	372	6.13 (2.09)	p < .05

Developmental Cascade Model: Males

Direct effects

Results for the developmental cascade model for males are depicted in Figure 2. As explained above, the model was fully identified, with all possible pathways and correlations between variables estimated (χ² (0, N = 1493) = .00, RMSEA = .00, CFI = 1.00). Only significant paths are shown in Figure 2, and for clarity, pathway coefficient estimates between covariates and later adolescent outcomes are included in Table 3.

Figure 2.

Standardized path coefficients in developmental cascade from harsh parenting to educational achievement for boys. Only significant pathways are shown. *p < .05.

Table 3.

Estimated Path Coefficients Between Covariates and Child Outcomes in Path Analysis Models

	Family Income	Parental Education	Race	Prior Academic Ability
Boys
W3 Extreme Peer Orientation	−.04	.02	−.23*	−.19*
W3 Academic Values	−.01	.00	.17*	.19*
W4 GPA	.05	.09	−.11*	.19*
W4 Early Sexual Behavior	.00	−.12*	.03	−.01
W4 Delinquency	.03	−.06	.02	−.11*
W6 Educational Attainment	.18*	.10	.16*	.17*
Girls
W3 Extreme Peer Orientation	.04	.04	−.19*	−.15*
W3 Academic Values	−.11*	.04	.16*	.03
W4 GPA	−.01	.15*	−.04	.34*
W4 Early Sexual Behavior	−.13*	−.07	−.02	−.12*
W4 Delinquency	−.05	−.05	.10*	.05
W6 Educational Attainment	.08	.17*	.05	.21*

Note. Race: 0 = White; 1 = Black.

^*p < .05

As predicted, harsh parenting during 7^th grade was associated with greater extreme peer orientation in 8^th grade, β = .25, p < .001. Extreme peer orientation, in turn, was related to increased delinquency, β = .24, p < .001, and lower GPA, β = −.12, p < .05, in 11^th grade. Both delinquency and GPA during 11^th grade predicted ultimate educational achievement by Wave 6 (i.e., three years post-high school), β = −.22, p < .01 and β = .31, p < .001, respectively. Although harsh parenting did not significantly predict academic values in 8^th grade, higher academic values did predict higher GPA, β = .16, p < .01, and lower levels of delinquent behaviors, β = −.12, p < .05, during 11^th grade.

Indirect effects

These findings on the developmental cascade from harsh parenting to low educational achievement were further examined as two interlocking mediational chains. Specifically, in the first part of the developmental cascade, we found that extreme peer orientation in 8^th grade mediated the association between harsh parenting at 7^th grade and delinquency problems in 11^th grade (IE = .04, 95% CI = .02 to .06). In the second part of the developmental cascade, we found that problems with delinquency mediated the pathway between extreme peer orientation and low educational achievement in boys (IE = −.16, 95% CI = −.06 to −.28). Although not directly hypothesized, there was a significant indirect effect between 8^th grade extreme peer orientation, 11^th grade GPA, and educational achievement (IE = −.11, 95% CI = −.02 to −.23). Thus, it appears as though extreme peer orientation in 8^th grade influences ultimate educational achievement both through an increase in delinquent behavior and a decrease in academic grades in 11^th grade. There were also indirect effects of academic values on educational attainment through 11^th grade GPA (IE = .11, 95% CI = .04 to .21) and delinquency (IE = .06, 95% CI = .007 to .13).

Developmental Cascade Model: Females

Direct effects

Results for the developmental cascade model for females are depicted in Figure 3. Again, estimated pathways between covariates and later outcomes are not depicted in the figure but are included in Table 3. As predicted, harsh parenting at 7^th grade was associated with greater self-reported extreme peer orientation one year later, β = .17, p < .001. Extreme peer orientation, in turn, predicted greater early sexual behavior, β = .10, p < .05, as well as increased delinquency, β = .38, p < .001, in 11^th grade. However, only early sexual behavior, and not delinquency, predicted ultimate educational attainment three years post-high school, β = −.23, p < .001. In contrast to the developmental cascade model for boys, harsh parenting at 7^th grade also predicted lower academic values in 8^th grade for girls, β = −.12, p < .01. Academic values also predicted 11^th grade GPA, β = .10, p < .05, which in turn predicted educational attainment, β = .12, p < .05. However, academic values did not have a significant direct effect on 11^th grade GPA, early sexual behavior, or delinquency in the model.

Figure 3.

Standardized path coefficients in developmental cascade from harsh parenting to educational achievement for girls. Only significant pathways are shown. *p < .05.

Indirect effects

Similar to the model for boys, the developmental cascade model for girls was examined as a series of mediational chains. In the first mediational chain, we found that 7^th grade harsh parenting predicted later early sexual behavior through the mediating influence of extreme peer orientation (IE = .004, 95% CI = .0001 to .009). In addition, although not hypothesized, there was a significant indirect effect of W1 harsh parenting on 11^th grade delinquency through 8^th grade extreme peer orientation (IE = .03, 95% CI = .01 to .04). In the second link of the mediational chain, there was a significant indirect effect of extreme peer orientation on educational attainment through the mediating influence of early sexual behavior (IE= −.08, 95% CI = −.006 to −.17).

Results also revealed a potential mediating cascade from harsh parenting to educational attainment through diminished academic values and GPA. In order to test this possibility, we again examined our findings as a series of mediational chains. In the first mediational chain, harsh parenting was indirectly related to lower GPAs during 11^th grade through lower academic values (IE= −.01, 95% CI = −.0001 to −.04). However, in the second link of the mediational chain, the indirect effect of academic values on educational attainment was nonsignificant, as evidenced by a 95% confidence interval that encompassed zero (95% CI = −.002, .07).

Discussion

Utilizing life history theory, this study tested a developmental cascade model linking harsh parenting to low educational attainment by using a longitudinal data set that spanned nine years, beginning with youth in the 7^th grade and ending three years after the students’ expected high school graduation year. Results supported indirect links between harsh parenting and low educational attainment through the development of extreme peer orientations, early sexual behavior, and delinquency, with these developmental pathways varying by gender. Among males, harsh parenting predicted the development of an extreme peer orientation, which further led to increased delinquency, and subsequently lower educational attainment. Among females, harsh parenting predicted extreme peer orientations, which increased both delinquency and early sexual behavior. Early sexual behavior, but not delinquency, predicted lower educational attainment in females.

Harsh Parenting and Extreme Peer Orientation

In the first link of our developmental cascade model, the experience of harsh parenting in 7^th grade predicted the development of an extreme peer orientation in 8^th grade for both males and females. Evolutionary theories of attachment and reproductive strategies (e.g., life history theory) have suggested that harsh and insensitive caregiving engenders an insecure attachment style that may be reflected in the way children learn to relate to peers (Belsky et al., 1991; Del Giudice, 2015). In the present study, extreme peer orientation reflected a construct that specifically examined an unhealthy dependency on peers, characterized by seeking closeness to peers at the expense of others, rejecting parental norms and rules, and willingly acquiescing to their peers’ desires rather than their own (Fuligni & Eccles, 1993; Fuligni et al., 2001). Hence, the finding that harsh parenting related to later extreme peer orientation may suggest that children whose needs are not met by their primary attachment figures may attempt to seek validation elsewhere, perhaps in an unhealthy manner that undermines their own autonomy and makes them susceptible to negative peer influences.

Extreme Peer Orientation and Problem Behaviors

Evolutionary theories of development have suggested that the peer group functions to pull individuals away from adult supervision and allow greater access to mating opportunities (Del Giudice, 2015; Ellis et al., 2012). Thus, in the second pathway of our cascade model, we proposed that individuals more heavily oriented toward the peer context at the expense of parental authority would be more susceptible to risky behaviors that signal desirability as mating partners. Consistent with research indicating that teenage women were more likely to engage in pre-marital sex, become pregnant, and have higher numbers of sexual partners when reporting that their friends were more influential on their beliefs than their parents (Shah & Zelnik, 1981), we found that extreme peer orientations were directly related to early sexual behavior among females.

Conversely, extreme peer orientation in males was related to delinquency, rather than to early sexual behavior. Evolutionary theories have proposed that because males have historically faced greater intrasex competition for mating partners, sexual selection processes have evolved to link heightened displays of aggression and social status in males with greater reproductive fitness (Campbell, 1999; Ellis et al., 2012). In support of these theories, we found that delinquency was positively correlated with early sexual behavior in males, suggesting that males who engaged in higher acts of aggression and delinquency were more likely to have had sex or gotten someone pregnant by the 11^th grade.

Although not included in our original hypotheses, we also found a link between extreme peer orientation and delinquency in females. Previously, researchers have posited that females are less likely to engage in risky and aggressive behavior because they have historically been the predominant caregiver of infants and children, making aggression risky for both the mother and the future of her offspring (Campbell, 1999; Kaplan & Gangestad, 2005). At first glance, results indicating that an extreme peer orientation was correlated with heightened delinquency in females appeared to counter the life history theory’s assertion that females generally refrain from risky and aggressive behavior. However, a closer analysis of our data reveals that females did in fact have significantly lower mean levels of delinquency than males. In addition, Campbell (1999) argued that females in fact engage in aggressive behaviors to gain or maintain resources; however, this aggression is more likely to use more indirect, covert, or subtle methods of aggression or criminal behavior (e.g., relational or verbal aggression, theft, fraud) than their male counterparts. In this study, only two items in the delinquency scale assessed direct physical aggression, while the rest involved indirect acts (e.g., stealing from a store). Future research should attempt to address whether unhealthy peer relationships foster direct aggression in females and males as well as broader forms of delinquency.

Alternatively, aggression in females, while risky, may also be a form of mate attraction. Olthof and Goossens (2008) suggested that the desire to be accepted by the opposite sex may induce girls to behave in similar behavioral styles, as they found that middle school girls who engaged in anti-social bullying practices endorsed a higher desire to be accepted by male classmates who engaged in the same type of bullying behaviors. From an evolutionary standpoint, females with a fast life strategy are particularly attracted to males who engage in risky and aggressive behavior (Jackson & Ellis, 2009); therefore, it is possible that these same females adopt similar behavioral patterns in an effort to attract these particular mates.

Problem Behaviors and Educational Attainment

In accord with the deleterious effects that early sexual behavior and delinquency have for adolescent adjustment and school outcomes, our hypothesis that early sexual behavior would predict low educational attainment for females was supported. Per life history theory, females’ increased energy spent on mating and relationships likely detracted the attention paid to educational pursuits, resulting in poorer academic performance and earlier termination of formal education (Richardson et al., 2015). Furthermore, we confirmed the hypothesized predictive link between heightened delinquency and lower educational achievement for males only. Again, this result was couched in life history theory’s assertion that individuals engaging in more delinquent behaviors are less likely to invest energy and resources toward academics and are also at-risk for school discipline issues that further erode academic performance and educational achievements (Richardson et al., 2015). Interestingly, unlike adolescent males, female delinquent behavior did not have a direct effect on educational attainment. Instead, females experienced an indirect effect on ultimate educational attainment via lower academic performance (i.e., GPA). Thus, it appears as if the effect of delinquency on educational attainment might operate through different mechanisms for males and females.

While further research is needed, we may better explain this result by investigating the gender-based differences in the frequency and type of delinquent behavior. As discussed earlier, females did have lower mean levels of delinquency in this study; thus, it is possible that females were engaging in relatively minor or less frequent delinquent behavior than their male counterparts, making them less susceptible to the negative effects of delinquency on educational pursuits. Another possibility is that, as proposed by Campbell (1999), females engage in more covert forms of behavior problems (e.g., shoplifting). These types of behaviors may escape attention and notice, thus diminishing school discipline issues that might affect overall educational attainment. Further research could examine whether the pathway between delinquency and educational attainment can be further explained by school discipline issues.

Bridging Systems via Developmental Cascade Model

The results of our study underscore the importance of examining predictors of educational achievement as a series of developmental processes that unfold over time. Developmental cascade models allow for the examination of how early predictors cumulatively influence a developmental outcome through a series of developmental courses and systems at multiple levels of analysis (e.g., family, peer, academic, behavioral) (Masten & Cicchetti, 2010). Often, examinations of predictors for educational attainment focus on identifying factors with a unique, single influence after controlling for other factors (e.g., Ou & Reynolds, 2008). While this approach is important for identifying risks associated with early school dropout and low educational achievement, it can ignore the developmental nature and course of multiple influences on adjustment. In the present study, we showed that harsh parenting was indirectly related to lower educational achievement through a cascading series of processes, including the development of an extreme orientation to peer relationships, delinquent behavior for males, and early sexual initiation for females.

In addition, these pathways held even after controlling for the influence of harsh parenting and extreme peer orientations on academic values and school performance. Among male adolescents, academic values predicted later grades and, indirectly, educational attainment. However, harsh parenting did not predict academic values. Conversely, harsh parenting negatively predicted academic values for female adolescents, which in turn predicted subsequent academic performance in 11^th grade. Lower high school GPAs, meanwhile, predicted lower educational attainment. Yet, mediational analyses revealed that the indirect pathway between academic values and educational attainment through GPA was not significant. Thus, while grades clearly impacted educational attainment in both genders and academic values influenced ultimate grade level completion for males, previous school performance and academic motivations could not explain the indirect link between harsh parenting and educational attainment. Instead, and in line with our proposed developmental model informed by life history theory, results suggested that harsh parenting influenced formal educational attainment through subsequent influences on peer relationships and adolescent sexual and delinquent behavior.

Implications

Prevention programs aimed solely at educating individuals of the negative effects of attractive but harmful behaviors (e.g., delinquency, sexual activity) are minimally successful and may even increase the undesirable behavior (Kirby, 2008; Shaffer et al., 1990; Tobler, Lessard, Marshall, Ochshorn, & Roona, 1999; Werch & Owen, 2002). Thus, a fuller understanding of the complex issues involved in why adolescents engage in certain behaviors is needed in order to inform more comprehensive prevention and intervention efforts (Ellis et al., 2012; Richardson et al., 2015).

In contrast to more conventional theories of development, evolutionary theories do not consider the development of risky behaviors to be maladaptive or negative in form; rather, they emphasize that risky behaviors are, in fact, adaptive reactions to the rearing environment. However, behaviors that were historically adaptive (e.g., heightened aggression and dominance, early mating) can also have damaging effects on long-term development in modern society. By understanding the underlying function and adaptive quality of these fast-life strategy behaviors, we can start to identify alternative means for engaging at-risk individuals in education (Richardson et al., 2015). For example, individuals with a fast life history strategy tend to discount the future and look for immediate rewards (Del Giudice, 2014). As humans, we are intrinsically motivated to learn through exploration and play, and we use peer groups and social relations (i.e., caregivers) to inform further knowledge acquisition of the natural world (Geary, 2012). Learning environments that create more rewarding aspects of education closer to our intrinsic approaches to learning (e.g., experiential learning, social group learning) may be particularly beneficial for engaging at-risk youth in modern school settings (Richardson et al., 2015).

Additionally, evolutionary theories emphasize that acts of delinquency and aggression, which predict lower educational attainment for adolescent males, function to increase dominance in the social hierarchy (Ellis et al., 2012). However, social dominance goals are often at odds with academic aspirations, resulting in adolescents valuing social dominance being more likely to have lower school engagement and poor academic performance (Kiefer & Ryan, 2008). While this phenomenon might be an effect of resource distribution (i.e., more resources and energy spent on achieving social dominance versus attending to school work), research has also suggested that academic performance and school engagement negatively impact social status starting in the middle school years (Galván, Spatzier, & Juvonen, 2011). Consequently, individuals who consider social dominance a primary goal may undermine their academic pursuits to gain higher social status and prestige within the peer group. In order to combat the negative stereotypes around high academic performance, efforts could attempt to highlight high status role models (e.g., athletes, movie stars, musicians) who value education and academic pursuits.

Limitations and Future Research

Some limitations of the current study merit discussion. All primary constructs were derived from adolescent self-reports. Given the potential bias inherent to parent reports of harsh parenting and knowledge of their adolescents’ behavior, we believe that the adolescents themselves were best equipped to adequately report on the study’s constructs, but future studies should make use of multiple sources of information (e.g., adolescents, teachers, parents) and examine whether findings can be replicated across different informants. It is also worth noting that some of the constructs assessed in the current study were of peripheral interest to the primary research goals of the MADICS study. Thus, some of the scales (e.g., early sexual behavior) were limited and construed from a small number of items. Future research should attempt to replicate our findings with broader, widely used measures of sexual behavior.

In addition, while we did control for earlier academic performance and motivations, many of the constructs in our model did not have earlier assessments available (e.g., sexual behavior). Therefore, the current study cannot make definitive conclusions about the emergence and transactional nature of these constructs over time. For example, it is possible that early sexual behavior and delinquency also predict extreme peer orientation. Indeed, it is likely that there are reciprocal, dynamic processes unfolding between individual behavior and the peer group, as we know that peer groups influence adolescent behavior and that individuals are likely to self-select into peer groups that reinforce their attitudes and behavior (Dodge, Dishion, & Lansford, 2006). Nevertheless, the prospective, longitudinal nature of the study is a key strength and highlights the importance of these unfolding processes over key developmental transition periods. However, the longitudinal nature of the study also resulted in missing data, particularly in the last wave of data collection. Although FIML procedures were conducted to ensure adequate sample size and power for analyses, we cannot discount the effect that lower retention rates over time may have had on our results. Finally, while the sample was economically and racially diverse, this study was limited to a sample of adolescents from a single geographic location. As such, caution should be exercised in generalizing these findings to other samples.

In addition to addressing these limitations, future research of educational achievement and engagement can be aided by a greater understanding of individual differences in short-term versus long-term reward-based strategies. The current study examines a set of mediating pathways that promote lower educational attainment among adolescents exposed to harsh parental contexts. Future research should also attempt to address under what conditions these links might be ameliorated. For example, school contexts that promote hands-on learning and immediate, intrinsic benefits of learning might mitigate the effects of a fast life history strategy on low educational achievement. In addition, we were interested in developmental processes occurring in middle childhood and adolescence, but future research could attempt to determine at what point harsh environmental contexts (e.g., harsh parenting, dangerous neighborhoods) that promote the fast life strategy begin to impact educational outcomes. Consequently, research conducted with younger children will greatly aid in our understanding of this phenomenon.

Conclusion

Despite the noted limitations, this study possesses a number of key strengths to aid in advancing our understanding of the proximal and distal factors relating to low educational attainment over time. To our knowledge, this study is the first to integrate an evolutionary model of development into our understanding of how and why the rearing environment influences educational achievement. The prospective, longitudinal design allowed us to test a novel developmental model in which harsh parenting had both direct and indirect effects on peer relationships, sexual behavior, delinquency, and ultimately, educational attainment over a nine-year span. These results underscore how a broad range of factors and complex processes work together over time to undermine educational achievement in at-risk youth. The theoretical and empirical advancements in the current paper also provide important information for the development of prevention and intervention programs aimed at increasing school engagement and graduation rates, particularly among at-risk youth.