Comparison of Parent versus Child-Report of Child Impulsivity Traits and Prediction of Outcome Variables

Abstract

Five personality traits that dispose individuals to rash or ill-advised action (i.e., sensation seeking, negative urgency, positive urgency, lack of planning, and lack of perseverance), can be reliably and validly assessed in children. This paper reports on the first test of parental reports of these traits. In a sample of 94 children (ages 7–13, mean age 10.6), the authors found the following. First, parental reports of the five traits in their children appeared to be reliable. Second, there was moderate convergent validity: parent and child reports of the same traits had a median correlation of r = .30. Third, there was adequate discriminant validity: within-parent reports on different traits had a median correlation of r = .11. Fourth, concurrent prediction of child behavior from parental reports generally was inconsistent with prior findings. Fifth, discrepancies between the two reporters did predict dysfunctional child behavior. There are advantages to securing both child self-report and parental report of personality dispositions to rash action, although there is limited evidence for the concurrent validity of parental reports.

In recent years, researchers have paid increasing attention to how best to integrate information from multiple informants in the domain of child personality (Tackett, 2011). Because personality traits early in life can influence subsequent developmental trajectories, including the emergence of psychopathology (Caspi & Roberts, 2001; Dodge & Pettit, 2003; Ge et al., 1996; Rutter et al., 1997), the assessment of child personality, using either child self-report (Measelle, John, Ablow, Cowan, & Cowan, 2005; Roberts & Del Vecchio, 2000; Zapolski, Stairs, Settles, Combs, & Smith, 2010) or parent report (Achenbach, McConaughy, & Howell, 1987; Fruyt et al., 2006; Funder, Kolar, & Blackman, 1995; Kraemer et al., 2003; Roberts & Del Vecchio, 2000; Tackett, 2011) has become a focus for clinical scientists. To date, child personality is rarely assessed using both child self-report and parental report (Klonsky, Oltmanns, & Turkheimer, 2002). Of course, in the domains of childhood behavior and childhood psychopathology, the gold standard is the use of multi-informant, multi-setting, multi-method, and multi-occasion assessment (e.g., Achenbach et al.,1987; Frick, 2000; Kamphaus & Frick, 1996; Mash & Hunsley, 2005; Mash & Terdal, 1997; Myers & Winters, 2002; Ollendick & Hersen, 1993; Sattler, 2002).

As we describe next, this paper is the first to study parental reports of a set of traits reflecting their children’s personality dispositions toward rash or impulsive action. The specific questions addressed by the current research are as follows: (1) Are parental reports of selected high risk child personality traits reliable? (2) Do parental reports have convergent validity with child self-reports? (3) Is there discriminant validity between parental reports of different traits? (4) Do parental reports of their children’s traits have concurrent validity in the prediction of dysfunctional child behavior? (5) Do discrepancies between child self-report and parental report predict child dysfunction? Answers to these questions will help clarify the role of parental reports of child personality in the domain of personality-based risk for rash or impulsive behavior.

Personality Traits that Dispose Individuals to Rash, Impulsive Action

One model concerning the personality underpinnings to rash or impulsive action was developed following factor analyses of numerous existing measures of impulsivity-related traits (Whiteside & Lynam, 2001). This line of research led to the identification of five different personality traits that dispose individuals to rash, impulsive action (Cyders & Smith, 2007; Whiteside & Lynam, 2001); each trait can be understood within the framework of the five factor model of personality. One prominent version of the five factor model, represented by the NEO PI-R personality measure (Costa & McCrae, 1992), identifies the five broad personality domains as Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness (Costa & McCrae, 1995).

Two of the traits focus on emotion: negative urgency and positive urgency reflect the tendency to act rashly when experiencing intense negative or intense positive emotion, respectively (Cyders & Smith, 2007; 2008a). These two traits appear to be interstitial, in that they are associated with high levels of Neuroticism, low levels of Conscientiousness, and low levels of Agreeableness (Cyders & Smith, 2008a). Two other traits reflect deficits in conscientiousness: lack of planning reflects a tendency to act without forethought and appears analogous to low deliberation within the Conscientiousness domain of personality, and lack of perseverance reflects difficulty maintaining one’s focus on a task and appears analogous to low levels of self-discipline within the Conscientiousness domain (Costa & McCrae, 1995; Smith et al., 2007; Whiteside & Lynam, 2001). The last trait is sensation seeking, which refers, instead, to the disposition to pursue novel, thrilling, or exciting stimulation and overlaps with excitement seeking on the Extraversion dimension of personality (Whiteside & Lynam, 2001; Zuckerman, 1994).¹

The five traits appear to represent different pathways to risky behavior with different external correlates, both concurrently (Cyders & Smith, 2008a; Dick et al., 2010; Smith et al., 2007) and prospectively (Cyders, Flory, Rainer, & Smith, 2009; Cyders & Smith, 2008b, 2010; Pearson, Combs, Zapolski, & Smith, 2012; Settles, Cyders, & Smith, 2010; Zapolski, Cyders, & Smith, 2009). When all the traits are studied together, sensation seeking is uniquely associated with engaging in highly stimulating activities that do not necessarily involve immediate risk (e.g., riding roller coasters; Fischer & Smith, 2004). The affect-based traits, negative and positive urgency, predict problematic involvement in risky behaviors likely undertaken when experiencing intense mood states (Cyders & Smith, 2007, 2010), and negative urgency predicts aggression (Derefinko, DeWall, Metze, Walsh, & Lynam, 2011). Lack of planning and lack of perseverance appear uniquely associated with school performance (Smith et al., 2007), and lack of planning sometimes explains unique variance in some risky behaviors (Smith et al., 2007).

Zapolski et al. (2010) reported on the development of measures to assess four of the five traits in children (all but positive urgency), based on child self-report. Both questionnaire and interview assessments of the traits were reliable, there was good convergent validity between questionnaire and interview reports of the same trait, and there was good discriminant validity between traits within assessment method. They also found that the four traits concurrently predicted different criteria, consistent with findings found among adult samples in previous studies (for example: Cyders et al., 2009; Cyders & Smith, 2008b, 2010): negative urgency predicted aggressive behavior and risky behavior undertaken while in an extremely negative mood; sensation seeking predicted the risky behaviors of riding roller coasters and jumping out of trees; lack of perseverance predicted academic problems, and lack of planning predicted academic problems prior to correction for its overlap with lack of perseverance (Zapolski et al., 2010). Gunn and Smith (2010), using a measure of all five traits, confirmed the five factor structure based on child self-report in a sample of 1,843 5^th graders (the last year of elementary school; approximately 90% of the children were age 10 or 11).

Multiple Informants and Childhood Personality Assessment

Klonsky and colleagues (2002) argued that two major limitations of sole reliance on self-report were that: (1) self-reports provided only one opinion/perspective, and (2) because individuals with personality dysfunction are frequently unable to view themselves realistically and are unaware of the effects of their behavior on others, their assessment may be flawed. On the other hand, assessing personality traits based on multiple informants may also be problematic. It may be the case that when, for example, parents provide assessments of their child, the parents’ assessment may be (1) an expression of characteristics of their own personalities rather than their child’s personality (Goldsmith, Losynoya, Bradshaw, & Campos, 1994), (2) providing answers based on preconceptions of typical childhood characteristics (Miller & Davis, 1992), or (3) simply inaccurately describing internal states of another person (Klonsky et al., 2002).

Studies investigating multiple informants for child personality found that self/informant agreement on DSM personality disorders was modest (r = .36, k = .14), with agreement slightly higher among non-DSM domains of personality (e.g., aggression: r = .47: see review by Klonsky et al., 2002). This finding appears to be consistent across studies (Barbaranelli, Vittorio Caprara, Rabasca, & Pastorelli, 2003; Costa &McCrae; 1992; Laidra, Allik, Harro, Merenakk, & Harro, 2006; Tromp & Koot, 2010). Agreement across informants improved as the child subject’s age increased (r = .69, p = .03), and informant agreement was comparable in clinical (r = .38, n = 8) and nonclinical samples (r = .40, n = 4). In addition, informant/informant agreement was higher than that between subject and informant (e.g., informant/informant: r = .81; subject/informant: r = .13: Klonsky et al., 2002).

This modest level of agreement across informants is consistent with what has been documented with respect to child psychopathology that is internalizing in nature, where parent-child agreement correlations tend to be as low as 0.25 (Achenbach et al., 1987; Rey, Schrader, & Morris-Yates, 1992; Seiffge-Krenke & Kollmar, 1998; Youngstrom, Loeber, & Stouthamer-Loeber, 2000). Parent-child agreement does appear to be larger when the behavior is observable, such as for externalizing behaviors (March, Parker, Sullivan, & Stallings, 1997; Jensen, Traylor, Xenakis, & Davis, 1988) than for non-observed, inferred states, such as internalizing symptoms (Rey et al., 1992; Sourander, Helstela, & Helenius, 1999; Tackett, 2011). Because personality traits are internal states, it is perhaps not surprising that parent-child agreement tends to be modest.

An important issue in addition to the level of agreement between informants is that if discordance is found among informants, what is to be done with this information? Of course, information from each of multiple informants may be useful, to the extent that it provides different (but valid) information from the differing perspectives of the informants (Klonsky et al., 2002; Ready & Clark, 2002). More recently, researchers have also demonstrated that the informant discrepancy variable, itself, can be potentially informative in predicting behavioral outcomes, by highlighting sources of conflict for the target individual (Achenbach, 2006; De Los Reyes, Alfano, & Beidel., 2010; Grills & Ollendick, 2002; Tackett, 2011). Discrepancies among parent-child reports may be an artifact of communication problems within the family unit, which may be displayed by having more discrepancies in reports of internalizing symptoms compared to reports of externalized symptoms (e.g., Guion, Mrug, & Windle, 2009).

The Current Study

The purpose of this study is to address five questions with respect to parental reports of their children’s personality dispositions toward rash or impulsive action: (1) Are parental reports reliable? (2) Do they have convergent validity with child self-reports? (3) Is there discriminant validity between parental reports of different traits? (4) Do parental reports have concurrent validity in the prediction of dysfunctional child behavior? (5) Do discrepancies between child self-report and parental report predict child dysfunction? We anticipated that parental reports would be internally consistent, have modest convergent validity with child self-reports, and there would be discriminant validity of parental reports of different traits.

To assess concurrent prediction of dysfunctional behavior, we selected five forms of childhood dysfunction using the requirement that there have been consistent previous findings relating specific ones of the five traits to each form of dysfunction. On the basis of the existing literature, we formed the following hypotheses when the five traits are considered together. Positive and negative urgency would be the only predictors of aggressive behavior (Derefinko et al., 2011; Settles, Fischer, Cyders, Combs, Gunn, & Smith, 2012; Zapolski et al., 2010). Low perseverance would be the only concurrent predictor of attention problems (Zapolski et al., 2010). Sensation seeking would be the sole predictor of involvement in childhood behaviors associated with risk and high stimulation, such as riding roller coasters and jumping out of trees (Zapolski et al., 2010). Negative urgency would be the sole trait associated with rash action while in a negative mood and positive urgency would be the sole trait associated with rash action while in an unusually positive mood (Cyders et al., 2010). Lastly, the two low conscientiousness traits, lack of planning and lack of perseverance, would be the sole predictors of problematic academic performance. In addition, we tested whether the discrepancy between parental reports and child self-reports would relate to child dysfunction.

Method

Participants

Participants were recruited from both the general community (n = 67) and from a clinical setting (n = 27). The intent of this recruitment strategy was to obtain a sample with a reasonably wide range of both impulsivity-related personality traits and externalizing behaviors. Clinical participants were recruited through their participation in anger management groups (n = 5), social skills groups (n = 8), and/or other treatment for a variety of disorders including Attention-deficit hyperactivity disorder (ADHD and Oppositional defiant disorder (ODD). The participants did not differ significantly on any of the demographic variables based on type of recruitment. For the clinical group, the mean age was 10 (SD = 1.15), mean grade was 5^th (SD = 1.34), and 61.5% were male, compared to the community sample, were the mean age was 10.7 (SD = 1.12), mean grade was 5^th grade (SD = 1.09), and 56.8% were male. For the overall sample of 94 children, ages ranged from 7 to 13 (M = 10.6 years, SD = 1.14; M grade = 5^th, SD = 1.14). A majority of the child sample was male (57.4%) and European American (83%). Seven children (7.4%) described themselves as African American, and other groups had small representation in this sample.

Participating parents included 83 mothers (mean age = 41.1 years) and 11 fathers (mean age = 47.5). Overall, the sample was relatively affluent: 61.7% reported total family incomes over $55,000; 21.3% reported incomes between $40,000 and $54,000; 9.6% reported incomes between $25,000 and $39,000; 4.3% reported incomes between $10,000 and $24,000; and 3.2% reported incomes under $10,000. Most participating fathers (45.5%) were in professional or managerial positions, and 43.3% of participating mothers were in such positions as well. In addition, 21.7% of participating mothers described themselves as homemakers. Fully 75.9% of mothers were college graduates, and 33.7% of those mothers had some post-college education. The same was true for fathers: 81.8% were college graduates, and 18.2% of those fathers had some post-college education. Concerning ethnicity, 84.3% of mothers described themselves as White; the largest other group (8.4%) described themselves as African American. Other groups had very small representation. Among participating fathers, all described themselves as White.

Measures

The UPPS-P-Child Version (Zapolski et al., 2010)

The UPPS-P-C is a modification of the original questionnaire by Whiteside and Lynam (2001): Minor modifications were made to the wording of the measure to be comparable to a 4^th grade reading level and the positive urgency scale was added to the measure. The UPPS-P-C is a 40 item Likert-type scale designed to assess lack of planning (e.g., “I like to stop and think about something before I do it”, reverse scored), lack of perseverance (e.g., “I almost always finish projects that I start,” reverse scored) negative urgency (e.g., “When I feel bad, I often do things I later regret in order to make myself feel better now”), positive urgency (e.g., “When I am in a great mood, I tend to do things that could cause me problems”), and sensation seeking (e.g., “I like new, thrilling things to happen”). Items are assessed from 1 (agree strongly) to 4 (disagree strongly). Past reliability estimates for the scales range from 0.80 to 0.94 (Smith et al., 2007). In the current sample, coefficient alpha estimates of reliability were .90 (sensation seeking), .87 (negative urgency), .84 (lack of planning), .81 (lack of perseverance), .89 (positive urgency).

The UPPS-P-Parent Version (Whiteside & Lynam, 2001)

The UPPS-P-Parent is similar to the UPPS-P-C, with the only modification being the source of the respondent. The UPPS-P-Parent is a 40 item scale designed to assess the same five traits, using the same response options as the child scale. In the current sample, coefficient alpha estimates of reliability for the scales were .91 (sensation seeking), .87 (negative urgency), .87 (lack of planning), .82 (lack of perseverance), .95 (positive urgency).

The Child Behavior Checklist (CBCL; Achenbach, 1991)

The CBCL is a 140 item Likert-type scale designed to assess problem behaviors and competencies in children. Numerous studies have supported the scale’s reliability and validity in the assessment of childhood dysfunction. We used it to assess both academic performance and behavioral/emotional problems. For academic performance, we used items assessing academic performance in language arts, history, mathematics, and science. Items are assessed from 1 (failing) to 4 (above average); we summed the items to provide a marker of overall performance. For behavioral and emotional problems, we used these standard parent report scales: Attention Problems and Aggressive Behavior. Items are assessed from 0 (not true) to 2 (very true or often true). Internal consistency reliabilities obtained for these scales were found to be 0.88 and 0.92, respectively. This measure was completed by the participant’s parent.

Risky Behaviors Scale (RBS: Fischer and Smith, 2004)

We used a two item measure of risky behaviors: on a five point Likert-type scale, children were asked to report on the frequency (from “never” to “often”) with which they“rode a roller coaster” and “jumped out of trees.” This measure has been used in the past with children this age (Zapolski et al., 2010), and is taken from a longer, 83 item measure that includes other topic domains, such as alcohol use, illegal drug use, gambling, investing in stocks, and sexual behavior (Fischer & Smith, 2004). The child participant completed this measure.

Mood-based Questionnaire (MBQ-C: Cyders and Smith, 2007)

The MBQ is a self-report measured aimed to assess endorsement of 23 risky behaviors engaged in while being in either an unusually positive mood or an unusually negative mood. Child participants reported, using a checklist, which of several risky behaviors they have engaged in while in an unusually positive mood. They then completed the same checklist concerning their behavior while in an unusually negative mood. Items include “drank alcohol,” “gotten sick from alcohol,” “used drugs,” “got arrested,” “broke the law,” and “started a fight.” The MBQ-C is a modification of the MBQ, designed to remove developmentally inappropriate items and simplify item language. Six items were deleted (e.g., “played drinking games”) and others were modified (e.g., “drove a car while intoxicated” was modified to “drove a car”), due to not being age appropriate. In previous research, good evidence was reported for the MBQ’s reliability and validity (Cyders & Smith, 2007); in the current sample, α = .73 for negative mood based rash action and α = .68 for positive mood based rash action. This measure was completed by the child participant.

Procedure

Participants were recruited through two sources. A community based sample of participants was gathered through advertisements in an urban Midwestern newspaper. Participants were informed that the study would examine personality traits in children, that the children would be asked to complete several questionnaires, and that the accompanying parent would also be asked to complete questionnaires as well. A second sample of participants was recruited through the psychological services center affiliated with the university’s clinical psychology program. Parents were provided the same information as the community sample as to the purpose of the study. All families volunteered to participate in the study by contacting the researchers via phone. One parent was required to accompany the child participant to the study and remain present throughout the session. The accompanying parent provided informed consent and the children provided informed assent. If the child had difficulty reading the assent form, the researcher read the form to the child. After signing the consent and assent form, the child participant completed the questionnaire packet. After reading and signing the consent form the parent participant completed a demographic form and the questionnaire packet. Participating families received $20 for participation.

Data Analysis

Missing data treatment

Only 1.6% of all values were missing. Individuals who provided complete data were compared to individuals who did not provide complete information on each study variable. The two groups did not differ on any variables. We therefore concluded that data were missing at random, and we imputed missing values using the Expectation Maximization (EM) method; monte carlo studies indicate that this method of data imputation yields more unbiased estimates of population parameters than do either dropping missing cases or mean substitution procedures (Enders, 2006).

Sample recruitment

Because we recruited participants both from the community and from a clinical site, there is the possibility that the associations we observed were inflated due to contrasts between dysfunction and healthy children. To investigate this possibility, we (a) examined the sample distributions closely for indications of bimodality and (b) re-ran all analyses, covarying out clinical versus community recruitment. All findings were the same when we included recruiting source, so we present the original analyses below.

Prediction of maladaptive behaviors based on informant report

For each of six maladaptive behaviors, we tested whether child self-reports or parental reports predicted in accord with hypothesis and previous research as described above. We used multiple regression to test these hypotheses. Gender was controlled in all the regressions. At step 1, we entered the set of traits not expected to predict a given criterion. At step 2, we entered the traits hypothesized to predict that criterion. This procedure enabled us to report whether non-hypothesized relationships were present or not, following step 1, and whether hypothesized relationships were present at step 2, net the predictive effects included in step 1. We did this separately for child self-reports and parental reports. There appeared to be adequate power to detect significant and meaningful effects: power to detect an effect when R² = .10 was .94; for a smaller effect of R² = .05, power was .73.

Incremental prediction of maladaptive behaviors

For each criterion for which there was evidence of concurrent validity for both child self-reports and parental reports, we tested the incremental validity of one reporter over the other using hierarchical multiple regression. In the first set of analyses, we entered the child self-report predictors at step 1 and, at step 2, tested whether parental reports added incremental validity in concurrent prediction. In the second set of analyses, we entered the parent reports as predictors at step 1 and, at step 2, tested whether child self-reports added incremental validity beyond them.

Prediction from Discrepancy between Parental Report and Child Self-Report

To investigate whether disagreement between the parent and child reports of child personality predicted the outcome variables examined, standard difference scores (SDS) were computed for each of the impulsivity-like traits according to procedures used by De Los Reyes et al. (2010). Specifically, all scores for parent and child were converted to z scores, and then parent’s standardized scores were subtracted from child’s standardized score. Positive SDS scores reflect the child reporting stronger endorsement of the personality trait than the child’s parent, whereas negative SDS scores reflect parents reporting stronger endorsement of the personality trait than their child. SDS scores were then used to predict each maladaptive behavior concurrently.

Results

Distributions of Study Variables

The distributions of scores on both child and parent measures of all five personality traits approximated normality, with very little skew or kurtosis; there was no indication of bimodality in the distributions of any study variables. Five of the six criterion variables (aggressive behavior, attention problems, general risky behavior, risky behavior while distressed, risky behavior while in an unusually good mood) were positively skewed. Square root transformations were performed on each; the transformed variables all had skew and kurtosis values less than 1.0. The sixth criterion variable, academic problems, had little skew or kurtosis.

Parental Report Reliability, Convergence with Child Self-report, and Discrimination among Traits

As noted above, parental reports were reliable. To compare parental reports with children’s self-reports, we calculated a multitrait, multimethod matrix. The matrix is made up of the five traits as assessed by the two methods (child and parent); the results are given in Table 1. The evidence of convergent validity between parent and child was very similar to what has been reported in the past, particularly for internalizing dysfunction: the average correlation between two methods of assessing the same trait was r = .30, and the range of correlations was from .20 (lack of planning) to .38 (sensation seeking and lack of perseverance). Parents did discriminate between the traits: the average heterotrait, monomethod correlation from parental reports was r = .11 (range from -.05 to .23), which was very similar to the average heterotrait, monomethod correlation using child self-reports (r = .09, range from -.13 to .60).

Table 1.

Multi-trait Multi-method correlation matrix

	SS-C	LPL-C	LPV-C	NU-C	PU-C	SS-P	LPL-P	LPV-P	NU-P	PU-P
SS-C		.16	−.13	.02	.14	.38**	.23*	−.05	.03	.01
LPL-C			.45**	.22*	.31**	.18*	.20*	.19*	−.04	.01
LPV-C				.15	.05	.12	.33**	.38**	.14	.19*
NU-C					.60**	.15	.24*	.15	.30**	.19*
PU-C						.18*	.27**	.29**	.18*	.24*
SS-P							.30**	−.07	.13	.10
LPL-P								.51**	.56**	.64**
LPV-P									.29**	.40**
NU-P										.69**
PU-P

Methods: SS-C: child sensation seeking; LPL-C: child lack of planning; LPV-C: child lack of perseverance; NU-C: child negative urgency; PU-C: child positive urgency; SS-P: parent sensation seeking; LPL-P: parent lack of planning; LPV-P: parent lack of perseverance; NU-P: parent negative urgency; PU-P: parent positive urgency. Correlations in bold are monotrait, heteromethod correlations.

^*p < 0.05

^**p < 0.01

Prediction of Maladaptive Behaviors based on Informant Report

Bivariate correlations between predictor and criterion variables

Table 2 presents the bivariate correlations between the five traits and the criterion variables of aggressive behavior (measured by the CBCL), attentional problems (CBCL), highly stimulating risky behaviors (RBS), risky behaviors undertaken while in an extremely negative mood (MBQ), risky behaviors undertaken while in an extremely positive mood (MBQ), and academic performance (CBCL). We note that we correlated race and sex with both child and parent reports of children’s personalities and with each of the six criterion variables. Of those 32 correlations, only 1 was significant at p < .05. Because the two demographic variables did not correlate with the study variables at a rate greater than would be expected by Type I error alone, we did not include them further in the regression analyses reported below. The following multiple regression analyses are presented in Table 3 for child self-report and Table 4 for parental report.

Table 2.

Bivariate Correlations between Traits and Criterion Variables

	AGG	ATN	RSB	NMR	PMR	ACD
Child-Report
SS	0.11	0.10	0.27**	0.00	0.14	−0.05
LPL	0.05	0.24*	0.23*	0.37**	0.27**	−0.25*
LPV	0.18	0.37**	0.09	0.18	0.07	−0.32**
NU	0.28**	0.18	0.19	0.52**	0.38**	−0.26*
PU	0.24*	0.24*	0.21*	0.28**	0.44**	−0.27**
Parent-Report
SS	0.19	0.14	0.21*	0.04	0.20	−0.09
LPL	0.65**	0.65**	0.05	0.04	0.15	−0.40**
LPV	0.32**	0.42**	0.04	0.03	0.09	−0.46**
NU	0.70**	0.58**	−.10	−0.07	0.17	−0.24*
PU	0.66**	0.69**	−0.05	−0.02	0.10	−0.26*

Criterion: AGG: Aggression, ATN: Attentional Problems, RSB: Risky Behaviors, NMR: Negative Mood Rash Action, PMR: Positive Mood Rash Action, ACD: Academic Performance. Traits: SS: Sensation Seeking, LPL: Lack of Planning, LPV: Lack of Perseverance, NU: Negative Urgency, PU: Positive Urgency.

^*p < 0.05;

^**p < 0.01; n = 94.

Table 3.

Prediction of Aggressive Behavior from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report
Variable	B	SE B	β	R2
Step One
Sensation Seeking	0.26	0.18	0.15
Lack of Planning	−0.19	0.26	−0.09
Lack of Perseverance	0.62	0.30	0.24*
				0.06
Step Two
Lack of Perseverance	0.63	0.30	0.24*
Negative Urgency	0.31	0.21	0.18
Positive Urgency	0.26	0.22	0.15
				0.14*

Prediction of Attentional Problems from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report
Variable	B	SE B	β	R2
Step One
Sensation Seeking	0.08	0.16	0.05
Negative Urgency	0.06	0.19	0.04
Positive Urgency	0.24	0.19	0.16
Lack of Planning	0.33	0.21	0.17
				0.09
Step Two
Lack of Perseverance	0.87	0.25	0.38**
				0.20**

Prediction of Non-harmful Risky Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report of Personality
Variable	B	SE B	β	R2
Step One
Lack of Planning	0.13	0.09	0.18
Lack of Perseverance	−0.01	0.10	−0.01
Negative Urgency	0.05	0.07	0.09
Positive Urgency	0.05	0.07	0.10
				0.08
Step Two
Sensation Seeking	0.14	0.06	0.25*
				0.14*

Prediction of Negative Mood-Based Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report
Variable	B	SE B	β	R2
Step One
Sensation Seeking	−0.08	0.12	−0.07
Lack of Planning	0.47	0.18	0.30*
Lack of Perseverance	0.05	0.21	0.03
Positive Urgency	0.23	0.12	0.20
				0.17**
Step Two
Lack of Planning	0.49	0.16	0.31**
Negative Urgency	0.64	0.13	0.53***
				0.35***

Prediction of Positive Mood-Based Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report of Personality
Variable	B	SE B	β	R2
Step One
Sensation Seeking	0.10	0.11	0.09
Lack of Planning	0.31	0.16	0.21
Lack of Perseverance	−0.12	0.19	−0.07
Negative Urgency	0.38	0.11	0.34**
				0.19**
Step Two
Positive Urgency	0.31	0.13	0.28*
				0.24***

Prediction of Academic Performance from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Child Self-Report
Variable	B	SE B	β	R2
Step One
Sensation Seeking	−0.01	0.04	−0.03
Negative Urgency	−0.06	0.05	−0.15
Positive Urgency	−0.07	0.05	−0.17
				0.09*
Step Two
Lack of Planning	−0.02	0.06	−0.04
Lack of Perseverance	−0.17	0.07	0.29*
				0.18**

Note: When entered together, neither of the two urgency traits provided significant incremental validity over the other, although the two, as a set, provided incremental validity over the other three traits. When entered separately, each provided incremental validity over the three traits entered at step 1: negative urgency, b = .27, p < .01; positive urgency, b = .26, p < .01.

^*p < 0.05;

^**p < 0.01,

^***p < 0.001

Table 4.

Prediction of Aggressive Behavior from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	Β	R2
Step One
Sensation Seeking	−0.02	0.14	−0.10
Lack of Planning	1.56	0.24	0.66***
Lack of Perseverance	−0.04	0.24	−0.02
				0.42***
Step Two
Lack of Planning	0.73	0.24	0.31**
Negative Urgency	0.74	0.18	0.39***
Positive Urgency	0.35	0.17	0.21*
				0.61***

Prediction of Attentional Problems from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	Β	R2
Step One
Sensation Seeking	−0.01	0.11	−0.01
Negative Urgency	0.21	0.17	0.12
Positive Urgency	0.58	0.16	0.40***
Lack of Planning	0.68	0.21	0.32**
				0.55***
Step Two
Positive Urgency	0.57	0.16	0.39**
Lack of Planning	0.58	0.23	0.27*
Lack of Perseverance	0.21	0.19	0.09
				0.56

Prediction of Non-harmful Risky Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	β	R2
Step One
Lack of Planning	0.11	0.11	0.15
Lack of Perseverance	0.03	0.10	0.03
Negative Urgency	−0.10	0.09	−0.16
Positive Urgency	−0.02	0.09	−0.04
				0.03
Step Two
Sensation Seeking	0.12	0.06	0.23*
				0.07*

Prediction of Negative Mood-Based Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	β	R2
Step One
Sensation Seeking	0.05	0.14	0.04
Lack of Planning	0.09	0.27	0.05
Lack of Perseverance	0.06	0.23	0.04
Positive Urgency	−0.08	0.16	−0.07
				0.01
Step Two
Negative Urgency	−0.19	0.21	−0.14
				0.02

Prediction of Positive Mood-Based Behaviors from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	Β	R2
Step One
Sensation Seeking	0.21	0.12	0.19
Lack of Planning	−0.03	0.23	−0.02
Lack of Perseverance	0.12	0.20	0.07
Negative Urgency	0.17	0.16	0.13
				0.06
Step Two
Positive Urgency	−0.07	0.17	−0.06
				0.06

Prediction of Academic Performance from Sensation Seeking, Lack of Planning, Lack of Perseverance, Negative Urgency, and Positive Urgency Based on Parent Report of Child’s Personality
Variable	B	SE B	β	R2
Step One
Sensation Seeking	−0.02	0.04	−0.06
Negative Urgency	−0.04	0.06	−0.10
Positive Urgency	−0.08	0.05	−0.20
				0.08*
Step Two
Lack of Planning	−0.12	0.08	−0.22
Lack of Perseverance	−0.20	0.06	−0.36**
				0.26***

^*p < 0.05;

^**p < 0.01,

^***p < 0.001

Prediction of aggressive behaviors

In the prediction of aggressive behavior, we hypothesized that negative and positive urgency would be the only traits to concurrently predict aggressive behavior. Using the child scales as predictors, the three non-hypothesized scales (sensation seeking, lack of planning, and lack of perseverance) entered at step one did not explain significant variance in the CBCL measure of aggressive behavior, but the addition of the two urgency scales at step two resulted in a significant improvement in fit. The two urgency scales did not differ from each other in their predictive power. Thus, using children’s self-report, our hypothesis was confirmed.

Using the parent scales as predictors, of the three non-hypothesized predictors entered in step one, lack of planning did predict significantly. In the second step, negative urgency and positive urgency both had incremental validity over lack of planning. Using parental reports, our hypothesis was partly confirmed, in that the urgency scales predicted as anticipated but lack of planning, which was hypothesized not to predict aggression, was the strongest predictor.

Prediction of attentional problems

Consistent with our hypothesis, lack of perseverance (entered at step two) was the only significant predictor of attentional problems using child self-reports. Using parental reports, at step one two non-hypothesized predictors (lack of planning and positive urgency) did predict significantly and highly, and at step two the hypothesized predictor, lack of perseverance, did not have incremental validity beyond those predictors. Using parental reports, although significant variance in attentional problems was accounted for, our hypotheses were not confirmed.

Prediction of risky behaviors

Consistent with our hypothesis, for both child self-reports and parental reports, sensation seeking (entered at step two) was the only significant predictor of the risky behaviors of riding roller coasters and jumping out of trees.

Prediction of risky behaviors undertaken while in an extremely negative mood

We hypothesized that negative urgency would be the only predictor of risky behaviors undertaken while in an extremely negative mood. Using the child scales as predictors, counter to our hypothesis, lack of planning was a significant predictor at step one. However, at step two negative urgency did predict above lack of planning. Our hypothesis was thus partly confirmed using children’s self-reports. Using parental reports, no trait predicted negative mood based risky behaviors. Our hypothesis was not confirmed using parental reports.

Prediction of risky behaviors undertaken while in an extremely positive mood

We hypothesized that positive urgency would be the only predictor of risky behaviors undertaken while in an extremely positive mood. Using the child scales as predictors, negative urgency did predict significantly at step one. However, at step two positive urgency predicted above and beyond negative urgency. Moreover, once positive urgency was entered, negative urgency no longer predicted the criterion significantly. We conclude that our hypothesis was supported using children’s self-reports. Using parental reports, no scales predicted the criterion. Our hypothesis was not confirmed using parental reports.

Prediction of problematic academic performance

We hypothesized that the two low conscientiousness traits would predict academic problems, and the other three traits would not. Using the child scales as predictors, counter to our hypothesis, at step one the three nonhypothesized scales as a set did predict academic problems significantly, although none of the scales predicted significantly individually. At step two, lack of perseverance (but not lack of planning) added significant predictive power. Using children’s reports, our hypothesis were partially confirmed. Similar findings were found based from the parent reports: at step one, the three non-hypothesized scales were significant predictors, but again, no individual scale predicted significantly. At step two, lack of perseverance again had incremental validity. Using parental reports, our hypothesis was partially confirmed.

Incremental Prediction of Maladaptive Behaviors

For three of the six child maladaptive behavior criteria (aggressive behavior, risky behavior, academic problems), there was evidence of concurrent predictive validity for both child self-report and parental report. For those three behaviors, we tested whether each source of personality data provided incremental predictive power over the other. We first tested the incremental validity of parental reports, by entering all five child self-report scales in step 1 of the regression equation and then all five parental report scales in step 2. In predicting aggressive behavior, the set of five child scales predicted significantly: R² = .14, p < .05. Parental reports added significant and substantial incremental variance to prediction from child self-reports, with an increased R² = .47, p < .001. Consistent with the earlier regression analyses, the significant parental report predictors were lack of planning (b = .30, p < .01), negative urgency (b = .37, p < .001), and positive urgency (b = .22, p < .05). In predicting risky behaviors, the set of five child scales predicted significantly, R² = .14, p < .05. Parental reports did not add significant incremental variance to the prediction as a whole or by individual traits (R² = .17, ns). The only significant predictor of risky behaviors was the child report for sensation seeking (b = .25, p < .05). In predicting academic problems, the set of five child scales predicted significantly, R² = .18, p < .01. Parental reports added significant incremental variance to the prediction, with an increased R² = .12, p < .05. As was the case with the earlier regression analysis, there was one significant parental report predictor: lack of perseverance (b = .30, p < .05).

We next tested whether the child self-reports had incremental validity over parental reports for the same three criteria, by entering the set of parental trait reports in step 1 and the set of child trait self-reports in step 2. In predicting aggressive behavior, the set of parental reports predicted significantly, R² = .61, p < .001. In step 2, child self-reports did not provide any incremental validity over parental reports. In predicting risky behaviors, parental reports did not provide significant predictive power, R² = .07, ns. The addition of the set of child self-reports in step 2 also did not provide significant incremental validity over parental reports (R² = .17, ns). However at the individual trait level, both the parent report (b = .23, p < .05) and child report (b = .25, p < .05) for sensation seeking significantly predicted risky behaviors. In predicting academic problems, parental reports predicted significantly, R² = .25, p < .001. The addition of child self-reports in step 2 again provided no incremental validity over parental reports.

Prediction of Maladaptive Behavior from Discrepancies between Informants

Of the six maladaptive behaviors, SDS scores significantly predicted three. For aggressive behaviors, parent-child disagreement for lack of planning (b = .38, p < .001) was the only significant predictor: children reporting higher levels of lack of planning than their parents predicted aggressive behavior. For negative mood based risky behaviors, parent-child disagreement for negative urgency (b = -.55, p < .001) was the only significant predictor: parents reporting higher levels of negative urgency than did the children predicted negative mood based risky behaviors. For positive mood based risky behaviors, parent-child disagreement for positive urgency (b = .28, p < .05) was the only significant predictor: children reporting higher levels of positive urgency than did their parents predicted positive mood based risky behaviors.

Discussion

One clinically important domain of personality is the set of personality traits that dispose individuals to rash or impulsive action. Because early involvement in risky and maladaptive behaviors is associated with more pathological life trajectories (Caspi & Roberts, 2001; Dodge & Pettit, 2003; Ge et al., 1996; Rutter et al., 1997), it is important to be able to assess the personality contributors to such behaviors very early in life, even prior to adolescence. Following research identifying five personality dispositions to rash action, each of which can be understood from the framework of the five factor model of personality (Costa & McCrae, 1995; Cyders & Smith, 2008a; Whiteside & Lynam, 2001), and research confirming the presence and concurrent validity of the five traits assessed by self-report in children (Gunn & Smith, 2010; Zapolski et al., 2010), this paper evaluated parental reports of their children’s personalities, with respect to their reliability, convergent validity, discriminant validity, and concurrent validity in the prediction of dysfunctional child behavior. In addition, we tested whether discrepancies between parental report and child self-report were themselves predictive of child dysfunction.

Parental reports were reliable. Convergent validity between parent report and child self-report was modest and quite similar to what has been observed concerning parent-child convergence in reporting children’s internalizing dysfunction. Discriminant validity between traits as reported by parents was comparable to the discriminant validity evident in children’s self-reports. Concurrent prediction of children’s dysfunctional behaviors produced a more complex picture concerning the validity of parental reports. Based on extensive past research (refer to Cyders & Smith, 2010; Derefinko et al., 2011; Settles et al., 2012; Zapolski et al., 2010), we had formed hypotheses concerning which personality traits would predict which dysfunctional behaviors. For children’s self-reports, predictions confirmed hypotheses fully for four behaviors and partially for the other two. In contrast, for parental reports, predictions confirmed hypotheses fully in only one case and partially for two others; hypotheses were not supported for three of the six predictions. Although failure to confirm hypotheses does not constitute evidence that the hypotheses are false, the confirmation failures occurred despite substantial statistical power.

Of the six measures of childhood dysfunction, three were reported on by the children and three by their parents. The source of criterion reports appeared to have little impact on whether hypotheses were confirmed. Concerning child self-reports of their personalities, for the three criteria reported on by the children, predictive hypotheses using child self-reports were fully supported for two (general risky behavior and risky behavior undertaken while in an unusually positive mood) and partially supported for one (risky behavior undertaken while distressed). For the three criteria reported on by parents, predictive hypotheses using child self-reports were fully supported for two (aggression and attentional difficulties) and partially supported for one (academic problems). In contrast, considering parent reports of their children’s personalities, for the three criteria reported on by the children, parental reports resulted in hypotheses that were fully supported for one criterion (general risky behavior), but not supported for the other two (risky behavior while distressed or while in an unusually good mood). For the three criteria reported on by parents, parental personality reports resulted in hypotheses that were partially supported for two (aggressive behavior and academic problems), and not supported for one (attentional problems).

However, source of criterion report did appear to be related to percent of variance accounted for by personality reports. For the three criteria provided by parents, parental personality reports explained more variance than did child self-reports. For the three criteria provided by the children, child personality reports explained more variance than did parental personality reports. This finding probably reflects method bias in the source of the reports.

Discrepancies between child personality self-report and parental reports of their children’s personalities predicted three of the six measures of childhood dysfunction. In two cases, a discrepancy in which the child reported higher trait endorsement than their parent was predictive (lack of planning for the prediction of aggressive behavior and positive urgency for the prediction of positive mood risky behaviors). In the third case, parents reported higher levels of negative urgency than did the children, and this discrepancy concurrently predicted negative mood-based risky behavior. As discussed by Tackett (2011), there are important implications to findings concerning discrepant reports. Perhaps discrepancies between parental and child reports reflect characteristics of the family, such as communication or relationships problems (Bidaut-Russell et al., 1995; Jensen, Xenakis, Davis, & Degroot, 1988; Kolko & Kazdin, 1993; Treutler & Epkins, 2003; Van Roy, Groholt, Heyerdahl, & Clench-Aas, 2010). For example, Van Roy and colleagues (2010) found that among a large sample of 10–13 year old youth and their parents, higher symptom endorsement among children compared to their parents was associated with poor parental engagement, not living with both parents, and less communication between parent and child. When parents reported more symptom endorsement than their child, this was associated with low parental educational level, low income, and male gender of the child. Although the bulk of this prior research concerns behavioral reports, the finding that the discrepancy between informant personality reports is descriptive speaks to the likely clinical importance of these discrepancies.

Discrepant reports can, of course, lead to confusion for researchers or clinicians (Tackett, 2011). In the current study, the finding that child self-reports were in much better agreement with hypotheses based on prior research than were parent reports suggests the value of relying on child reports in such circumstances. However, it was also true that parental personality reports shared more variance with parent-reported dysfunction than did child self-reports. Overall, the findings of this study suggest that information from both types of informants can prove valuable. Our findings also clearly indicate the need for further investigation of the nature and meaning of reports from the different informants, as well as for the discrepancies between informants.

The findings of this study should be considered in the context of the study’s limitations. Although the sample size included a range of children, including children receiving clinical services and those not, it is of course true that larger samples would provide more stable results. In future research, gathering information on external criteria drawn from both parent and child reports, rather than one or the other, would be beneficial in many ways. Our index of risky behavior included only two items; measurement of that construct should be improved in future studies. For the current study, sex and ethnicity were found to be unrelated to most of the study variables. However, because our sample was primarily European American, we were unable to examine whether ethnic differences exist with respect to parent-child agreement; this possibility should be further investigated. The predictions were concurrent, so this study provides no data concerning the temporal precedence of these personality traits and the criteria studied (but see Pearson et al., 2012).

In addition, no questions were asked about the parent’s psychological functioning, and parental functioning may have contributed to the current findings. De Los Reyes and Kazdin (2005), proposed a model that focuses on the importance of examining both parental (e.g., parental psychopathology, parental stress, parental acceptance of the child) and child (e.g., age, gender, social desirability, problem type, perceived stress) characteristics and the interaction among these characteristics in order to understand underlying mechanisms related to informant discrepancies. The impact of these factors has been identified in several studies (for example: Smith, 2007; Stokes, Pogge, Wecksell, & Zaccario, 2011; Van Roy et al., 2010). Age of the child participants may also be an important factor to consider in future studies, given data that suggest that parent-child agreement appears to be stronger among older children compared to those of a younger age (Edelbrock, Costello, Dulcan, Conover, & Kala, 1986). Despite these limitations, the findings of the current study, when considered together with prior studies (e.g., Gunn & Smith, 2010; Pearson et al., 2012; Zapolski et al., 2010), provide good evidence for the construct validity of children’s self-reports on five personality traits that appear to underlie rash or impulsive action, good evidence for some forms of validity of parental reports of their children’s personalities, as well as clear evidence of the need for future research in this domain.