risk factors for which there is within-subject change over time (variable risk factors) from those that do not change (e.g., gender, ethnicity, genotype—fixed markers) (Kraemer, Kazdin, et al., 1997). Causal risk factors are those that are modifiable by an intervention and for which modification is associated with change in outcomes. A risk factor that cannot be changed by an intervention or for which change in the factor has not been demonstrated to lead to a change in an outcome is considered a variable marker.
Protective factors are defined as characteristics at the individual, family, or community level that are associated with a lower likelihood of problem outcomes. The distinctions between risk factors discussed above can also be applied to protective factors. The term “protective factors” has also been used to refer to interactive factors that reduce the negative impact of a risk factor on a problem outcome, or resilience (Luthar, 2003). It is often difficult to distinguish the effect of protective factors from that of risk factors, because the same variable may be labeled as either depending on the direction in which it is scored (e.g., good parenting versus poor parenting, high self-esteem versus low self-esteem—Masten, 2001; Luthar, 2003). For example, in a meta-analytic review of studies of risk and protective factors, Crews et al. (2007) reported that low academic achievement was a risk factor for externalizing problems, whereas adequate academic performance was a protective factor.
One approach to distinguishing the effect of protective factors from risk factors is to consider them as the extreme ends of a continuous variable (Luthar, 2003; Rutter, 2003; Stouthamer-Loeber, Loeber, et al., 1993). For example Stouthamer-Loeber, Loeber, and colleagues (1993), in a study of schoolchildren in grades 1, 4, and 7, trichotomized 35 predictors of delinquency at the 25th and 75th percentile to refer to risk and protection, respectively. They found that for 43 percent of the variables both the risk and protective effects were significant predictors, whereas for 11 percent of the variables only the protective effect was significant. Similarly, Luthar and Lantendresse (2005), in a study of wealthy and poor preadolescents, assessed the relations of mental health problems with high or low scores on each of seven aspects of the parent–child relationship. They found that some dimensions (e.g., perceived closeness to parents) had both risk and protective effects (both high scores and low scores were related to symptoms), whereas for other parenting variables (e.g., criticism) there was only a risk effect. Thus, for many commonly studied predictor variables, it is likely that both the risk and protective effects may contribute to children’s mental, emotional, and behavioral problems, although studies that carefully differentiate whether the effects of these variables are at the risk or protective pole are not common.