The Report of the National Commission on Marihuana and Drug Abuse

Marihuana and (Non-Violent) Crime

VI. MARIJUANA USE AND CRIME: CAUSAL OR SPURIOUS

Our first clue as to the nature of the underlying relationship between marijuana use and crime is obtained by examining Table 5, which documents the association between number of offenses and marijuana use, taking into account the use of other drugs as well. Table 5 shows that "marijuana only" marijuana users are only very slightly more likely to commit crimes than non-users. Tire original nine percentage point difference between users and non-users in committing four or more offenses has shrunk to only two percentage points (in fact, only one percentage point--we are comparing 24.4% with 25.5%, and the original 19 percentage point difference between users and non-users in committing no offenses has been reduced to 13 percentage points. In fact, the "marijuana only" user is far more similar to the non-user in number of offenses committed than he is to the user of marijuana plus two or more other drugs. The application of the control involving other drugs clearly attenuates the marijuana-crime relationship, and much of the user-non-user differences in offenses committed can be traced to the fact that marijuana users are significantly more likely to use other drugs, rather than the use of marijuana per se.*

*The exploration of the "escalation" hypothesis-the question of whether or not marijuana "leads to" the use of other more dangerous drugs-would take us far afield in this report. The issue is dealt with in a separate report by the author prepared for the National Commission on Marihuana and Drug Abuse, however.

Our confidence in the lack of basic association between marijuana use itself and offenses--and our ability to attribute the simple original relationship to the use of drugs other than marijuana-is strengthened when we examine the relationship between frequency of use and offenses, holding the use of other drugs constant. Table 5A presents these data. Among "marijuana only" users, there is no statistically significant correlation between frequency of use and the commission of crimes--the differences are small, insignificant, and in no consistent direction. Among users of one other drug, likewise, frequency of marijuana use is extremely loosely associated with committing offenses-the differences which are observed are small, statistically insignificant, and point in no particular direction. And lastly, among users of two or more other drugs aside from marijuana, the association between frequency of marijuana use and the commission of the offenses described earlier is loose and not at all significant. Table 5A very strongly indicates the validity of the "spurious" model.

Closely related to the use of other drugs is a control which should be explored: the respondent's involvement with a drug-using subculture-with others who also use drugs. Naturally, as we saw in the last section, individuals who use marijuana tend to have friends who also use marijuana. What we would like to know is whether the social patterns they exhibit can be traced to their use of the drug. or to the fact that they have friends who use drugs. Thus, we should be very interested in looking at the individual who uses marijuana, but who does riot have drug-using friends, as well as the individual who does riot rise marijuana, but has drug-using friends. If interesting changes take place in the original marijuana-crime correlation, then we have indications that it is the individual's social relations, and not his rise of marijuana, per se, that determine marijuana use's correlation with crime. Tables 6 and 7 deal with this issue; they should be examined in conjunction with one another.

Among respondents with no marijuana-using friends (out of their 10 closest friends). as well as those with no drug-using friends at all. almost an identical relationship between marijuana use and crime obtains as in the original Table 3. Marijuana users are, significantly (but not strikingly) more likely to commit four or more offenses, and significantly less likely to commit none. The percentage differences in Tables 6 and 7 are almost the same (even slightly more) as those in Table. 3-about 25% at the "none" end vs. 19% for the original table, and about 18% at the "four or more" end vs. 9% for the original table (Table 3). If we were to rely only on these segments of the table, we would be led to the, inference that marijuana is meaningfully related to the commission of crimes.

However, an extraordinary thing happens to the other segments of the table: among those with drug-using friends, there is no relationship at all between marijuana use and committing offenses. The differences are so small as to be statistically insignificant. (And they are actually in the opposite direction-non-users are slightly more likely to commit four or more offenses than users are.) In other words the marijuana-crime relationship is completely wiped out by the application of controls-in this case, integration into the drug-using subculture. We are led overwhelmingly to the conclusion that marijuana users tend to be somewhat more likely to commit crimes solely because they are part of a drug-using subculture; the actual properties of marijuana appear to be completely unassociated with criminal behavior. Anyone (whether he uses marijuana or not) who makes friends and becomes involved with others who use drugs-especially others who use drugs in addition to and aside from marijuana stands a higher likelihood of committing offenses, simply because this segment of the population tends to be more lax about obeying the law. It is merely because marijuana users tend to associate with others who are part of this subculture that their crime rate is somewhat higher. In other words, the marijuana-crime relationship-in terms of the causal or effects model is completely spurious.

What of the marijuana-using isolate? What of those who use marijuana but who have no friends who use drugs-either marijuana or other drugs? How can we explain the fact that the original difference in offense rate between users and non-users were not wiped out among those without drug-using friends-in contrast to being wiped out among those who had drug-using friends? Looking back at Tables 6 and 7, we are struck by the fact that the offense rate of the marijuana-using group without drug-using friends is as high as the groups with drug using friends. What does this indicate? First of all, involvement with a drug-rising subculture is clearly not the only determinant of a high rate of committing offenses. Probably something else is at work. The fact that the offense differences wash out among those involved with the drug subculture indicates that rise by itself (i.e., being high) cannot be a factor in the original marijuana-crime relationships difference cannot explain two things that are similar. The using isolate, however, is probably a (Qualitatively different social being from those who either do not rise, or those who use and have friends who also use. The isolate is probably deviant in a variety of ways crime being one of them. The fact that he uses in spite of having no friends who use (although lie may have friends who do not use) means that he uses most of the time alone. has not integrated his activities into a social life or a subculture, and probably pursues some activities which the dominant society would judge to be eccentric and unusual. In other words, it is possible that his involvement with crime is related to his social isolation from a group which pursues drug activities similar to his own, rather than his use of marijuana. However, these remarks must be regarded as speculation, since more complete information is not available.

A related process occurs when race is employed as the control variable. Among blacks, the original relationship remains intact; marijuana users are significantly (although not markedly) more likely to commit four or more offenses, and less likely to commit none, than is true of non-users. (The relationship is significant at the .004 level, using Chi-square.) However, among whites, the original relationship is completely washed out; users are not at all more likely to commit offenses. Table 8 presents these data. The minor differences can be completely accounted for by random fluctuations; the relationship is not significant at any level.

When education is treated as the control variable in this relationship, the marijuana-crime connection follows a pattern parallel to the two previous explorations-that is, it is completely washed out for some groups, but remains for others. In Table 9, we see that marijuana users who have at least attended college, but not graduate school, are not any more likely to commit offenses than college non-users. The percentages committing four or more crimes, and those committing none, are almost identical for users and non-users alike. (The differences are insignificant, and due to random fluctuation.) However, among respondents with only a high school education, the same basic relationship produced in Table 3 is upheld: marijuana users are significantly more likely to have committed offenses than non-users. However, at least two variables are compounded here, since most respondents with a college education are white. Thus, the retainment of the original marijuana-crime correlation among high school education respondents, and its wash-out among college educated respondents, should not come as any surprise. However, among respondents with at least some graduate school education, the original relationship asserts itself; users are more likely to commit offenses than non-users. (The differences observed are significant at the .05 level; in other words, the differences have a one in 20 chance of occurring purely at random.) Table 9 presents the marijuana-crime relationship, holding education constant.

Using age as a control variable also produces mixed results. Among the two youngest age groups (15--20 and 21-23), the differences obtaining between users and nonusers are statistically insignificant, and could have occurred by chance alone; in the youngest of these two groups, the non-user is slightly more likely to have committed four or more offenses, and in the next to youngest of these groups, the user is slightly more so--but in both, the differences are too small to be meaningful. In the 24-to-28-year-old group, the differences approach statistical significance, but they are not substantial (.07 level of significance, using Chi-square). However, among the oldest group (age 29 to 34), the same basic difference in offenses as obtained in the original relationship holds up here; users are about twice as likely to have committed four or more offenses as non-users (28% vs. 13%), and less than half as likely to have committed no offenses (21% vs. 58&). These data appear in Table 10.