Everything is heritable: "Paternal Antisocial Behavior and Sons' Cognitive Ability: A Population-Based Quasiexperimental Study", Latvala et al 2014 https://pdf.yt/d/UBIAUANGVNTT1gGg / https://www.dropbox.com/s/etg4rwspuycfpph/2014-latvala.pdf / http://sci-hub.org/downloads/a8ac/latvala2014.pdf ; excerpts:
"Parents' antisocial behavior is associated with developmental risks for their offspring, but its effects on their children's cognitive ability are unknown. We used linked Swedish register data for a large sample of adolescent men (N = 1,177,173) and their parents to estimate associations between fathers' criminal-conviction status and sons' cognitive ability assessed at compulsory military conscription. Mechanisms behind the association were tested in children-of-siblings models across three types of sibling fathers with increasing genetic relatedness (half-siblings, full siblings, and monozygotic twins) and in quantitative genetic models. Sons whose fathers had a criminal conviction had lower cognitive ability than sons whose fathers had no conviction (any crime: Cohen's d = −0.28; violent crime: Cohen's d = −0.49). As models adjusted for more genetic factors, the association was gradually reduced and eventually eliminated. Nuclear-family environmental factors did not contribute to the association. Our results suggest that the association between men's antisocial behavior and their children's cognitive ability is not causal but is due mostly to underlying genetic factors.
...Children of antisocial parents may experience various adversities that could compromise their cognitive development. Such factors include unstable rearing environments, poor parenting practices, limited cognitive stimulation, interparental conflict and violence, and even victimization from parental abuse or neglect (Eaves, Prom, & Silberg, 2010; Hanscombe, Haworth, Davis, Jaffee, & Plomin, 2011; Koenen, Moffitt, Caspi, Taylor, & Purcell, 2003; Perkins & Graham-Bermann, 2012; Schwartz & Beaver, 2013). Such adversities might influence cognitive development directly, or their effects could be mediated via other factors, such as the biological effects of chronic stress. For example, it has been suggested that exposure to violence-a factor potentially linked to parents' antisocial behaviors and criminality-can cause neurocognitive problems in children via neurological changes that result directly from the exposure or via problems in interpersonal communication between parents and children (Perkins & Graham-Bermann, 2012). Moreover, it has been speculated that parental involvement in the criminal-justice system might itself also have adverse effects on children's development (Murray et al., 2012).
However, antisocial behavior is not randomly distributed in the population. Antisocial parents are likely to differ from parents who have no tendency toward antisocial behavior in many important additional aspects that could influence the cognitive development of their children. Antisocial behaviors have increased incidence among parents with low levels of education and low socioeconomic status, both of which are associated with poor cognitive development in offspring (Lawlor et al., 2005; Neiss & Rowe, 2000). As a result, any association between parents' antisocial traits and cognitive development of their children could be noncausal and merely reflect confounding by factors such as parental cognitive abilities or genetic influences. Antisocial behaviors and cognitive abilities are well known to be significantly heritable; genetic factors explain 40% to 70% of their variation in the population (Burt, 2009; Frisell, Pawitan, Långström, & Lichtenstein, 2012; Haworth et al., 2010; Rhee & Waldman, 2002). Antisocial traits and cognitive abilities are also systematically negatively correlated (Frisell, Pawitan, & Långström, 2012; Isen, 2010), and twin studies have suggested that this correlation is due in part to common genetic influences (Koenen, Caspi, Moffitt, Rijsdijk, & Taylor, 2006). However, a meta-analysis of large-scale twin studies of cognitive ability confirmed that in addition to genetic influences, environmental factors shared by twins reared together explain a significant proportion of variation in cognitive development until young adulthood (Haworth et al., 2010). Thus, an association between parents' antisocial behavior and cognitive outcomes in their offspring might arise because of genetic confounding or because of a true causal influence of parental behavior.
...family-based quasiexperimental study designs, such as sibling-comparison and children-of-siblings designs, can be used to rigorously test for causal effects of the family environment and parental characteristics (D'Onofrio, Lahey, Turkheimer, & Lichtenstein, 2013; Rutter, Pickles, Murray, & Eaves, 2001). Such family-based, quasiexperimental designs compare biologically related individuals who differ with regard to the presence of a particular risk factor, such as parental antisocial behavior, and are thus able to partly control for unmeasured genetic and environmental factors shared by family members. The potential of quasiexperimental methods for testing causal hypotheses has been emphasized in the literature; so far, however, they have only rarely been applied to testing environmental influences on traits such as cognitive ability (D'Onofrio et al., 2014; Ellingson, Goodnight, Van Hulle, Waldman, & D'Onofrio, 2014; Lundberg et al., 2010).
...Using a large, population-based data set that included Swedish men born over a 40-year period and their parents, we estimated associations between fathers' antisocial behavior and their sons' cognitive ability. To test for potential genetic confounding, we examined this association in fathers with and without criminal convictions and of three sibling types with different degrees of genetic relatedness (i.e., half-siblings, full siblings, and MZ twins). We refer to these as sibling-father pairs. Sources of familial confounding were further investigated with quantitative genetic structural equation modeling in extended families.
...We performed a national cohort study by linking several Swedish longitudinal population-based registries maintained by governmental agencies. The link between the registries was the unique personal identification numbers given to all Swedish citizens at birth and to immigrants on arrival to Sweden. Data from cognitive-ability testing of males born in Sweden from 1952 until 1991 were obtained from the Conscript Register...The personal identification numbers in the MGR were used to link to data in other government records. Parents' antisocial behavior was indexed by data on criminal convictions obtained from the Crime Register, which documents all convictions in lower courts for individuals ages 15 (age of criminal responsibility) and older from 1973 onward. The register includes detailed information about the timing, nature, and number of offenses leading to court convictions. The MGR was used to identify biological parents of conscripted men and to construct extended families for the children-of-siblings analyses (by identifying paternal grandparents). The Swedish Twin Register was used to identify pairs of fathers who were MZ or DZ twins.
...We identified 2,207,631 men born in Sweden between 1952 and 1991...After exclusions, the population analyses included 1,177,173 men for whom cognitive-ability data were available. For the within-family analyses, 111,784 extended families with paternal cousins were identified, and we randomly selected one son per nuclear family. There were 102,133 pairs of full-sibling fathers (i.e., full brothers or DZ twins; n = 1,003), 8,977 pairs of half-sibling fathers, and 674 pairs of MZ-twin fathers. For the quantitative genetic structural equation modeling, data from all families were used; we selected a maximum of two brothers per nuclear family (n = 867,439).
...Four different models were fitted: Model 1 predicted sons' cognitive ability from fathers' criminal-conviction status without any covariates. Model 2 adjusted for fathers' and sons' birth years, family SES, and fathers' educational levels. Model 3 added mothers' birth years, criminal-conviction status, and educational levels as covariates. Model 4 further adjusted for parents' existing psychiatric disorder and immigration status. We conducted a separate population regression model for each crime subtype as well as for crime of any subtype. To account for confounding by fathers' cognitive ability, we conducted separate analyses for the subset of sons whose fathers had cognitive-ability data available. In these models, fathers' cognitive ability was used instead of educational level.
Using the xtreg modeling command with the fixed effects (fe) option in Stata (Version 12; StataCorp, 2011), we fitted conditional linear regression models (i.e., fixed-effects regression models; Gunasekara, Richardson, Carter, & Blakely, 2014) to study the association between fathers' criminal-conviction status and sons' cognitive ability within sibling-father pairs. These models use a within-family estimator to compare the cognitive-ability scores of cousins whose fathers' criminality data were discordant (i.e., one cousin's father had committed crimes but the other one's father had not) and thus adjust for all unmeasured factors shared by cousins. A separate model was fitted to the data for each type of sibling father: half-siblings, who share, on average, 25% of their segregating genes; full siblings and DZ twins, who share, on average, 50% of their genetic makeup; and MZ twins, who share 100% of their genes. In this order, these within-family analyses increasingly controlled for genetic confounding in the association between fathers' criminality and sons' cognitive ability. If genetic factors confounded the association, the regression coefficient would be reduced as the control for genetic factors increased (i.e., coefficient in the population > coefficient for offspring of half-siblings > coefficient for offspring of full siblings > coefficient for offspring of MZ twins). In contrast, a causal effect of fathers' criminality would be supported if the observed association with son's cognitive ability was equal across the population-based and within-family models.
To complement the within-family regression analyses, we conducted quantitative genetic structural equation modeling (Kuja-Halkola, D'Onofrio, Larsson, & Lichtenstein, 2014) for the data from the extended families to estimate the magnitudes of genetic and environmental sources of familial confounding. The model we used is an extension of the standard model used in twin research, which decomposes variance into genetic influences (A), shared environmental influences (C), and nonshared environmental influences (E).
...As shown in Table 1, men whose fathers had any criminal convictions had lower cognitive-ability scores than men whose fathers had no criminal convictions (4.70, 95% confidence interval, or CI = [4.69, 4.70], vs. 5.22, 95% CI = [5.22, 5.23], Cohen's d = −0.28). Of the subtypes of paternal convictions, violent and sexual crimes were associated with the lowest cognitive ability among sons (4.28, 95% CI = [4.26, 4.30], Cohen's d = −0.49, and 4.32, 95% CI = [4.27, 4.38], Cohen's d = −0.47, respectively); traffic crimes were associated with the highest cognitive ability among sons (4.73, 95% CI = [4.72, 4.74], Cohen's d = −0.26). Similar mean differences were observed for paternal cognitive ability among the subset of fathers for whom cognitive-ability data were available (n = 223,433 fathers). Fathers' criminality was also associated with lower parental education, higher prevalence of parental psychiatric morbidity, parental immigration to Sweden, lower SES when the son was younger than 10, and mothers' criminal-conviction status (Table 1).
Results from regression models predicting sons' cognitive ability using fathers' criminal-conviction status are presented in Table 2. In unadjusted models, fathers' criminal convictions predicted a reduction of 0.43 to 0.86 stanine units in sons' cognitive-ability scores. For all crime types, adjustment for cohort effects, family SES, and paternal education significantly reduced the association (by between 41% and 47%; Model 2). Adjustments for maternal characteristics (Model 3) and for parental immigrant status and psychiatric morbidity further reduced the associations (by between 59% and 63% in total; Model 4) but did not completely eliminate them. Table 2 also presents results for the subsample of fathers with cognitive-ability data; the results paralleled those found in the whole population.
To investigate the mechanisms linking fathers' criminal-conviction status and sons' cognitive ability, we compared the association in the population and across fathers who were half-siblings, full siblings, and MZ twins. As shown in Table 3, the association in Model 1 was gradually reduced with increasing adjustment for unmeasured genetic factors-population: b = −0.53, 95% CI = [−0.54, −0.52]; sons of half-siblings, b = −0.38, 95% CI = [−0.46, −0.29]; sons of full siblings, b = −0.22, 95% CI = [−0.25, −0.19]; and sons of MZ twins, b = 0.14, 95% CI = [−0.18, 0.46]. The pattern of associations remained similar in models adjusting for measured parental covariates (Models 2-4).
...Results of the quantitative genetic model strongly supported the children-of-siblings regressions (Tables S3-S5 in the Supplemental Tables). Genetic factors explained 80% of the intergenerational association between fathers' criminal-conviction status and sons' cognitive ability. The rest of the association was due to environmental factors shared by all members of an extended family (C); nuclear-family environmental influences (F) did not contribute to the association.
...These within-family analyses showed that as more genetic factors were controlled for, the association between fathers' criminality and sons' cognitive ability gradually diminished. When genetic factors confounding the association were completely adjusted for, in sons of MZ twin brothers, no association between fathers' criminal-conviction status and sons' cognitive ability was observed. This analysis was followed up with quantitative genetic modeling using the extended-family data, the results of which supported the main analysis: Genetic influences explained most of the association. Extended-family environmental factors also contributed to the association, but nuclear-family environmental factors shared only between fathers and sons did not. Our findings, therefore, do not support a causal effect but rather suggest that genetic effects and nongenetic factors shared by members of extended families and, potentially, reflecting wider socioeconomic differences are sufficient to explain the intergenerational link between parents' antisocial behavior and lower cognitive ability in their offspring.
...Finally, it should be noted that the children-of-siblings design, like any nonexperimental research design, -cannot conclusively prove or rule out a causal association. Each quasiexperimental design is based on assumptions that need to be carefully considered when interpreting the results (D'Onofrio et al., 2013). Among the assumptions of the children-of-siblings analyses are that there are no carryover effects, meaning that the exposure of one cousin has no effect on the outcome of the other cousin and that the results generalize to extended families with no cousins. Note, however, that the results of our children-of-siblings analyses were supported by compatible findings from quantitative genetic structural equation modeling in which data from all families were used."
"Parents' antisocial behavior is associated with developmental risks for their offspring, but its effects on their children's cognitive ability are unknown. We used linked Swedish register data for a large sample of adolescent men (N = 1,177,173) and their parents to estimate associations between fathers' criminal-conviction status and sons' cognitive ability assessed at compulsory military conscription. Mechanisms behind the association were tested in children-of-siblings models across three types of sibling fathers with increasing genetic relatedness (half-siblings, full siblings, and monozygotic twins) and in quantitative genetic models. Sons whose fathers had a criminal conviction had lower cognitive ability than sons whose fathers had no conviction (any crime: Cohen's d = −0.28; violent crime: Cohen's d = −0.49). As models adjusted for more genetic factors, the association was gradually reduced and eventually eliminated. Nuclear-family environmental factors did not contribute to the association. Our results suggest that the association between men's antisocial behavior and their children's cognitive ability is not causal but is due mostly to underlying genetic factors.
...Children of antisocial parents may experience various adversities that could compromise their cognitive development. Such factors include unstable rearing environments, poor parenting practices, limited cognitive stimulation, interparental conflict and violence, and even victimization from parental abuse or neglect (Eaves, Prom, & Silberg, 2010; Hanscombe, Haworth, Davis, Jaffee, & Plomin, 2011; Koenen, Moffitt, Caspi, Taylor, & Purcell, 2003; Perkins & Graham-Bermann, 2012; Schwartz & Beaver, 2013). Such adversities might influence cognitive development directly, or their effects could be mediated via other factors, such as the biological effects of chronic stress. For example, it has been suggested that exposure to violence-a factor potentially linked to parents' antisocial behaviors and criminality-can cause neurocognitive problems in children via neurological changes that result directly from the exposure or via problems in interpersonal communication between parents and children (Perkins & Graham-Bermann, 2012). Moreover, it has been speculated that parental involvement in the criminal-justice system might itself also have adverse effects on children's development (Murray et al., 2012).
However, antisocial behavior is not randomly distributed in the population. Antisocial parents are likely to differ from parents who have no tendency toward antisocial behavior in many important additional aspects that could influence the cognitive development of their children. Antisocial behaviors have increased incidence among parents with low levels of education and low socioeconomic status, both of which are associated with poor cognitive development in offspring (Lawlor et al., 2005; Neiss & Rowe, 2000). As a result, any association between parents' antisocial traits and cognitive development of their children could be noncausal and merely reflect confounding by factors such as parental cognitive abilities or genetic influences. Antisocial behaviors and cognitive abilities are well known to be significantly heritable; genetic factors explain 40% to 70% of their variation in the population (Burt, 2009; Frisell, Pawitan, Långström, & Lichtenstein, 2012; Haworth et al., 2010; Rhee & Waldman, 2002). Antisocial traits and cognitive abilities are also systematically negatively correlated (Frisell, Pawitan, & Långström, 2012; Isen, 2010), and twin studies have suggested that this correlation is due in part to common genetic influences (Koenen, Caspi, Moffitt, Rijsdijk, & Taylor, 2006). However, a meta-analysis of large-scale twin studies of cognitive ability confirmed that in addition to genetic influences, environmental factors shared by twins reared together explain a significant proportion of variation in cognitive development until young adulthood (Haworth et al., 2010). Thus, an association between parents' antisocial behavior and cognitive outcomes in their offspring might arise because of genetic confounding or because of a true causal influence of parental behavior.
...family-based quasiexperimental study designs, such as sibling-comparison and children-of-siblings designs, can be used to rigorously test for causal effects of the family environment and parental characteristics (D'Onofrio, Lahey, Turkheimer, & Lichtenstein, 2013; Rutter, Pickles, Murray, & Eaves, 2001). Such family-based, quasiexperimental designs compare biologically related individuals who differ with regard to the presence of a particular risk factor, such as parental antisocial behavior, and are thus able to partly control for unmeasured genetic and environmental factors shared by family members. The potential of quasiexperimental methods for testing causal hypotheses has been emphasized in the literature; so far, however, they have only rarely been applied to testing environmental influences on traits such as cognitive ability (D'Onofrio et al., 2014; Ellingson, Goodnight, Van Hulle, Waldman, & D'Onofrio, 2014; Lundberg et al., 2010).
...Using a large, population-based data set that included Swedish men born over a 40-year period and their parents, we estimated associations between fathers' antisocial behavior and their sons' cognitive ability. To test for potential genetic confounding, we examined this association in fathers with and without criminal convictions and of three sibling types with different degrees of genetic relatedness (i.e., half-siblings, full siblings, and MZ twins). We refer to these as sibling-father pairs. Sources of familial confounding were further investigated with quantitative genetic structural equation modeling in extended families.
...We performed a national cohort study by linking several Swedish longitudinal population-based registries maintained by governmental agencies. The link between the registries was the unique personal identification numbers given to all Swedish citizens at birth and to immigrants on arrival to Sweden. Data from cognitive-ability testing of males born in Sweden from 1952 until 1991 were obtained from the Conscript Register...The personal identification numbers in the MGR were used to link to data in other government records. Parents' antisocial behavior was indexed by data on criminal convictions obtained from the Crime Register, which documents all convictions in lower courts for individuals ages 15 (age of criminal responsibility) and older from 1973 onward. The register includes detailed information about the timing, nature, and number of offenses leading to court convictions. The MGR was used to identify biological parents of conscripted men and to construct extended families for the children-of-siblings analyses (by identifying paternal grandparents). The Swedish Twin Register was used to identify pairs of fathers who were MZ or DZ twins.
...We identified 2,207,631 men born in Sweden between 1952 and 1991...After exclusions, the population analyses included 1,177,173 men for whom cognitive-ability data were available. For the within-family analyses, 111,784 extended families with paternal cousins were identified, and we randomly selected one son per nuclear family. There were 102,133 pairs of full-sibling fathers (i.e., full brothers or DZ twins; n = 1,003), 8,977 pairs of half-sibling fathers, and 674 pairs of MZ-twin fathers. For the quantitative genetic structural equation modeling, data from all families were used; we selected a maximum of two brothers per nuclear family (n = 867,439).
...Four different models were fitted: Model 1 predicted sons' cognitive ability from fathers' criminal-conviction status without any covariates. Model 2 adjusted for fathers' and sons' birth years, family SES, and fathers' educational levels. Model 3 added mothers' birth years, criminal-conviction status, and educational levels as covariates. Model 4 further adjusted for parents' existing psychiatric disorder and immigration status. We conducted a separate population regression model for each crime subtype as well as for crime of any subtype. To account for confounding by fathers' cognitive ability, we conducted separate analyses for the subset of sons whose fathers had cognitive-ability data available. In these models, fathers' cognitive ability was used instead of educational level.
Using the xtreg modeling command with the fixed effects (fe) option in Stata (Version 12; StataCorp, 2011), we fitted conditional linear regression models (i.e., fixed-effects regression models; Gunasekara, Richardson, Carter, & Blakely, 2014) to study the association between fathers' criminal-conviction status and sons' cognitive ability within sibling-father pairs. These models use a within-family estimator to compare the cognitive-ability scores of cousins whose fathers' criminality data were discordant (i.e., one cousin's father had committed crimes but the other one's father had not) and thus adjust for all unmeasured factors shared by cousins. A separate model was fitted to the data for each type of sibling father: half-siblings, who share, on average, 25% of their segregating genes; full siblings and DZ twins, who share, on average, 50% of their genetic makeup; and MZ twins, who share 100% of their genes. In this order, these within-family analyses increasingly controlled for genetic confounding in the association between fathers' criminality and sons' cognitive ability. If genetic factors confounded the association, the regression coefficient would be reduced as the control for genetic factors increased (i.e., coefficient in the population > coefficient for offspring of half-siblings > coefficient for offspring of full siblings > coefficient for offspring of MZ twins). In contrast, a causal effect of fathers' criminality would be supported if the observed association with son's cognitive ability was equal across the population-based and within-family models.
To complement the within-family regression analyses, we conducted quantitative genetic structural equation modeling (Kuja-Halkola, D'Onofrio, Larsson, & Lichtenstein, 2014) for the data from the extended families to estimate the magnitudes of genetic and environmental sources of familial confounding. The model we used is an extension of the standard model used in twin research, which decomposes variance into genetic influences (A), shared environmental influences (C), and nonshared environmental influences (E).
...As shown in Table 1, men whose fathers had any criminal convictions had lower cognitive-ability scores than men whose fathers had no criminal convictions (4.70, 95% confidence interval, or CI = [4.69, 4.70], vs. 5.22, 95% CI = [5.22, 5.23], Cohen's d = −0.28). Of the subtypes of paternal convictions, violent and sexual crimes were associated with the lowest cognitive ability among sons (4.28, 95% CI = [4.26, 4.30], Cohen's d = −0.49, and 4.32, 95% CI = [4.27, 4.38], Cohen's d = −0.47, respectively); traffic crimes were associated with the highest cognitive ability among sons (4.73, 95% CI = [4.72, 4.74], Cohen's d = −0.26). Similar mean differences were observed for paternal cognitive ability among the subset of fathers for whom cognitive-ability data were available (n = 223,433 fathers). Fathers' criminality was also associated with lower parental education, higher prevalence of parental psychiatric morbidity, parental immigration to Sweden, lower SES when the son was younger than 10, and mothers' criminal-conviction status (Table 1).
Results from regression models predicting sons' cognitive ability using fathers' criminal-conviction status are presented in Table 2. In unadjusted models, fathers' criminal convictions predicted a reduction of 0.43 to 0.86 stanine units in sons' cognitive-ability scores. For all crime types, adjustment for cohort effects, family SES, and paternal education significantly reduced the association (by between 41% and 47%; Model 2). Adjustments for maternal characteristics (Model 3) and for parental immigrant status and psychiatric morbidity further reduced the associations (by between 59% and 63% in total; Model 4) but did not completely eliminate them. Table 2 also presents results for the subsample of fathers with cognitive-ability data; the results paralleled those found in the whole population.
To investigate the mechanisms linking fathers' criminal-conviction status and sons' cognitive ability, we compared the association in the population and across fathers who were half-siblings, full siblings, and MZ twins. As shown in Table 3, the association in Model 1 was gradually reduced with increasing adjustment for unmeasured genetic factors-population: b = −0.53, 95% CI = [−0.54, −0.52]; sons of half-siblings, b = −0.38, 95% CI = [−0.46, −0.29]; sons of full siblings, b = −0.22, 95% CI = [−0.25, −0.19]; and sons of MZ twins, b = 0.14, 95% CI = [−0.18, 0.46]. The pattern of associations remained similar in models adjusting for measured parental covariates (Models 2-4).
...Results of the quantitative genetic model strongly supported the children-of-siblings regressions (Tables S3-S5 in the Supplemental Tables). Genetic factors explained 80% of the intergenerational association between fathers' criminal-conviction status and sons' cognitive ability. The rest of the association was due to environmental factors shared by all members of an extended family (C); nuclear-family environmental influences (F) did not contribute to the association.
...These within-family analyses showed that as more genetic factors were controlled for, the association between fathers' criminality and sons' cognitive ability gradually diminished. When genetic factors confounding the association were completely adjusted for, in sons of MZ twin brothers, no association between fathers' criminal-conviction status and sons' cognitive ability was observed. This analysis was followed up with quantitative genetic modeling using the extended-family data, the results of which supported the main analysis: Genetic influences explained most of the association. Extended-family environmental factors also contributed to the association, but nuclear-family environmental factors shared only between fathers and sons did not. Our findings, therefore, do not support a causal effect but rather suggest that genetic effects and nongenetic factors shared by members of extended families and, potentially, reflecting wider socioeconomic differences are sufficient to explain the intergenerational link between parents' antisocial behavior and lower cognitive ability in their offspring.
...Finally, it should be noted that the children-of-siblings design, like any nonexperimental research design, -cannot conclusively prove or rule out a causal association. Each quasiexperimental design is based on assumptions that need to be carefully considered when interpreting the results (D'Onofrio et al., 2013). Among the assumptions of the children-of-siblings analyses are that there are no carryover effects, meaning that the exposure of one cousin has no effect on the outcome of the other cousin and that the results generalize to extended families with no cousins. Note, however, that the results of our children-of-siblings analyses were supported by compatible findings from quantitative genetic structural equation modeling in which data from all families were used."
