Interpretations and Public Policy
Although most social scientists recognize that there is an empirically demonstrated effect of media violence exposure on aggression, many scientists mistakenly believe that the size of the effect is trivially small. There probably are several reasons for this belief. Perhaps the most typical is statistical in nature. We were all taught in introductory statistics classes to square a correlation coefficient in order to determine how much variance can be accounted for. Thus, an effect size of 0.20 (in r terms) accounts for only 4% of the variance. Effects sizes in the 0.20 range are common in media violence studies. Many people may feel that this is not a large percentage of variance, and in absolute terms they may be right. Yet human aggression is a multicausal phenomenon. Why someone acts aggressively in a given situation is influenced by hundreds of factors, some proximal and some distal, some personal and some situational. If it is true that there are hundreds of variables that have some influence on whether an individual acts aggressively, then we should not expect any single variable to account for even 1% of the variance in aggression. Therefore, if any one variable can account for 1% or more of the variance, then that variable does not seem to us to be trivial. Indeed, among many empirically identified risk factors for aggression, media violence exposure accounts for at least as much variance as most others. Table 9.1 lists a number of factors that have been identified as risk factors for youth violence and aggression, as well as estimates of the magnitude of their longitudinal effect sizes. As shown in Table 9.1, most risk factors that many aggression scholars would call “important” do not account for a great deal of variance individually (e.g., being male accounts for 3.6%, having a low IQ accounts for 1.2%,
-142-
