The National Household Survey on Drug Abuse (NHSDA) provides estimates of prevalence, incidence, demographic and geographic distribution, and correlates of use of illicit drugs, alcohol, and tobacco in the civilian, noninstitutionalized U.S. population 12 years of age or older. The survey gives particular emphasis to collecting information on adolescents by oversampling 12 to 17 year olds and by using questionnaire modules designed exclusively for adolescents. In the 1997 NHSDA, a new module was added for 12 to 17 year olds to examine risk and protective factors related to substance use. Risk factors include those individual characteristics or social environments associated with an increased likelihood of substance use, while protective factors are related to decreased likelihood of substance use or of nonuse.

The role of risk and protective factors in social interaction and substance use has been investigated for about 20 years. Reviews of this literature are presented by Hawkins, Catalano, and Miller (1992) and by Petraitis, Flay, Miller, Torpy, and Greiner (1998). Botvin, Botvin, and Ruchlin (1998) also reviewed the effectiveness of selected substance use prevention programs, classifying the programs into four types of approaches: information dissemination, affective education, social influence, and comprehensive or expanded social influences. Information dissemination approaches provide information about the risks of substance use, and affective education approaches focus on personal and social development. Both of these approaches have been shown to have little or no effect in reducing substance use due to their narrow focus. Social influence and integrated social influence approaches have, however, been shown to be effective. Social influence approaches involve persuasive messages from peers and the media, and integrated social influence/competence enhancement approaches that teach self-management, social, cognitive, self-esteem enhancing, adaptive coping, and general assertiveness strategies and skills. Each of these latter two approaches has been linked to substantial reductions in the use of cigarettes, alcohol, and illicit drugs.

This report presents findings from the 1997 NHSDA on the relationship between risk and protective factors and substance use. The report is based on a large body of research regarding the types of factors that have been associated with reductions in substance use when implemented as part of a well-planned prevention program. However, although the specific factors investigated in this report have generally been thought of as risk and protective factors, not all have been shown to be independently associated with substance use.

Chapter 1 describes the NHSDA methodology and the prevention risk and protective factors included in the survey. Chapter 2 looks at the estimated prevalence of the various risk and protective factors in the U.S. population and how these vary by race/ethnicity, gender, and age. Chapter 3 examines the relationship of these factors to different levels of substance use. The main focus is on marijuana; however, cigarette use, alcohol use, and use of any illicit drug other than marijuana are discussed as well. Chapter 3 also discusses the relative "odds" of using a substance for different levels of risk. Chapter 4 introduces multivariate techniques to analyze the strength of association of each of the major domains of risk and protective factors and of demographic variables with youth substance use.

Risk and protective factors involve attitudes and behavior associated with the higher likelihood of use or of nonuse of substances. The classification approach used in this report combines factors into one of five domains: community, family, peer/individual, school, and general. A complete list of the questions and response categories included in the analyses, with each item mapped onto a particular domain, is presented in Appendix A. The 1997 NHSDA questionnaire included at least one, and in some cases as many as a dozen, specific items drawn from the research literature on prevention related to these domains. Community factors include availability and marketing of licit and illicit drugs. Family factors include parentaldisciplinary approach, family conflict, parental attitudes about substance use, and parental communication about drugs and alcohol. Peer/individual factors include perceptions of risk of substance use, delinquent behaviors, and friends' substance use and attitudes toward substance use. School factors include enrollment, grades achieved, and formal antidrug education programs. The general domain consists of social support, participation in activities, exposure to antidrug media messages, and intensity of religious beliefs and observance. Most of these items were designed for and asked only of the 12 to 17 year olds in the sample. Many of these items focused only on behavior in the past year or at the present time.

In developing this report, the prevalence and distribution of these risk and protective factors across the youth cohort was examined and correlated with past year use and intensity of use of cigarettes, alcohol, marijuana, and illicit drugs other than marijuana. A very striking result of these analyses was the uniformity, regardless of the substance, in the patterns of association between substance use and risk and protective factors. The factors most strongly correlated with past year use, intensity (frequency) of past year use, and past year nonuse of cigarettes were also the factors most strongly associated with the same measures of use, intensity, or nonuse of alcohol, marijuana, and illicit drugs other than marijuana although the strength of the relationships may have varied. Factors that were weakly or not discernibly associated with measures of cigarette use were also weakly or not discernibly associated with measures of alcohol, marijuana, and illicit drug use other than marijuana. It was apparent, in view of this high degree of commonality, that focusing particular attention on one substance would be an effective and efficient way to examine these factors. For this purpose, marijuana was selected as the "lead" substance to be discussed most extensively in this report. Note, as discussed in Chapter 3, that the substance use measures used in this report yield estimates of use that are slightly lower than those presented by the Office of Applied Studies (OAS, 1999b).

Detailed data tables are presented in Appendix B and standard errors for selected tables in Appendix C. Distributions of risk and protective factors and substance use by age are presented in Appendix D, and further analyses of unexpected findings on the relationship of marijuana use to exposure to prevention messages are presented in Appendix E.

Overview of the NHSDA Methodology

The NHSDA is the primary source of statistical information on the use and correlates of illicit drugs, alcohol, and tobacco in the United States. Conducted by the Federal Government since 1971, the NHSDA is administered to a representative sample of the civilian, noninstitutionalized population of the 50 States aged 12 or older at their place of residence. The NHSDA is directed by the Substance Abuse and Mental Health Services Administration (SAMHSA) of the U.S. Department of Health and Human Services. The target population includes residents of noninstitutional group quarters, such as those residing in college dormitories or group homes and civilians living on military installations. It also includes persons with temporary but not permanent residence at the time of the survey (i.e., homeless people in shelters and residents of single rooms in hotels). The sample excludes active-duty military personnel, U.S. citizens living abroad, residents of institutional settings (e.g., prisons and hospitals), and homeless persons not living in a shelter at the time of the survey.

The household interview takes approximately 1 hour to complete and incorporates procedures designed to maximize truthful responses to potentially sensitive questions about illicit drug use. Data are collected on the recency and frequency of use of various licit and illicit drugs, opinions about substances, problems associated with substance use, and substance abuse treatment experiences. In addition to detailed information about substance use, the NHSDA also collects basic demographic information on employment, race/ethnicity, age, education, income, marital status, health status, mental problems, health insurance, utilization of health services, and access to health care.

The 1997 NHSDA consisted of a first-stage clustered sample of counties (or groups of counties), a subsample of blocks (or block groups), and a sample of households and individuals in those households. Hispanics, blacks, and adolescents were oversampled to increase the precision of estimates for these groups.² The survey was also designed to examine seasonal variation, so data were collected throughout the calendar year; for more information, see the 1997 Main Findings report (Office of Applied Studies [OAS], 1999b). The 1997 NHSDA had 24,505 respondents, of whom 7,844 were 12 to 17 years of age on the day of interview. The household screening response rate for the year was 92.7 percent, and the personal interview response rate for this age group was 82.8 percent. Data presented in this report are weighted to obtain unbiased estimates of substance use in the population represented by the NHSDA. Additional information on survey methodology, sampling, and weighting is presented in the 1997 Main Findings report (OAS, 1999b).

This chapter provides estimates derived from the 1997 NHSDA of the levels of various risk and protective factors of substance use among adolescents. Because risk and protective factors are often correlated with demographic characteristics, estimates of these factors are presented separately by race/ethnicity, gender, and age. Distinctive differences in risk and protective factors by year of age can lead to misinterpretation of the relationship between risk and protective factors and the prevalence of substance use. This issue is addressed in greater detail at the end of this chapter and in Appendix D.

Estimates in this chapter are based on simple cross-tabulations and tests for statistical significance. More detailed multivariate analyses presented in Chapter 4 shed additional light on the underlying relationships.

Table 2.1 summarizes the prevalence rates of the risk and protective factors covered in this report. More detailed data about the distributions of risk and protective factors by race/ethnicity, gender, and age are presented in Tables B.5 to B.10 (Appendix B). In this chapter, risk and protective factors are grouped and presented in the domain classification introduced in Chapter 1, with the one exception that the three items related to exposure to prevention messages, which are part of the general domain, are presented separately.

Community Domain

Community risk and protective factors measured by the NHSDA include availability and marketing of licit and illicit drugs.

Drug availability. Drug availability was a primary community-level risk factor measured in the 1997 NHSDA. Respondents were asked whether it was difficult or easy to obtain each of the following types of drugs: marijuana, cocaine, crack, LSD, and heroin. The five available response options ("probably impossible," "very difficult," "fairly difficult," "fairly easy," and "very easy") were dichotomized as "difficult" and "easy." Data were also available on whether anyone had ever offered or attempted to sell marijuana/hashish or cocaine to the respondent.

Figure 2.1 and Table B.5 show adolescent perceptions of drug availability. Marijuana was the only drug a majority of youths aged 12 to 17 (58 percent) indicated was easy to obtain. Almost 30 percent of youths thought it would be easy to obtain cocaine, crack, or LSD. Respondents were less likely to see heroin as easy to obtain (21 percent) compared to the other drugs. More than one third (35 percent) of those surveyed reported someone offered to give or sell them marijuana, and about 10 percent indicated this for cocaine (see Table 2.1).

Factors, by race/ethnicity, gender, and age. About 60 percent of whites and blacks, but only 52 percent of Hispanics, said marijuana was easy to obtain (Table B.5). However, 35 percent of each racial/ethnic group reported that someone had offered to give or sell them marijuana. More black than white or Hispanic adolescents reported that it would be easy for them to get cocaine (41 vs. 29 and 32 percent, respectively), crack (45 vs. 26 and 29 percent), or heroin (31 vs. 20 and 23 percent). However, a smaller percentage of black adolescents than white or Hispanic adolescents was actually offered cocaine (6 vs. 10 and 13 percent), consonant with their lower rates of past year use. In general, females were more likely than males and older youths were more likely than younger adolescents to perceive illicit substances as easy to get. Older youths were also more likely than their younger counterparts to have been offered marijuana or cocaine.

Family Domain

Family risk and protective factors measured by the NHSDA include parental disciplinary approach, family conflict, parental attitudes about substance use, and parental communication about drugs and alcohol.

Family management and conflict. Three items measured family management: how strict adolescents said their parents were about the way the youths dressed, how late they stayed out at night (curfew), and how much time they spent doing homework (Table B.6a). The available response options for these items were "not at all strict," "just strict enough," and "too strict." Family conflict was measured as the frequency of arguing with parents in the past 12 months, ranging from "at least several times per week" to "rarely or never."

Adolescents were most likely to say that their parents were lenient about the way they dressed compared to the two other topics. Almost 45 percent of youths indicated that their parents were not at all strict about the way they dressed, 25 percent said their parents were not strict about homework, and only 13 percent felt the same way about curfew. Nearly 3 out of 10 (30 percent) adolescents rarely or never argued with their parents in the year prior to the interview.

Parents' attitudes toward substance use. Youth perceptions of parental feelings about substance use were measured for various substances and frequencies of use by asking whether youths thought their parents would be "not at all upset," "somewhat upset," or "very upset" if their child used the substance named in the example (Tables B.6b and B.6c). Youths thought their parents would be relatively more upset about their use of cocaine and heroin than about marijuana and cigarettes (see Figures 2.2a and 2.2b). Close to 81 percent of adolescents thought their parents would be very upset if they smoked one or more packs of cigarettes a day or if they tried marijuana once or twice. In regard to binge drinking (having five or more drinks) one to two times a week, smoking marijuana once a month, and trying inhalants once or twice, 89 percent of adolescents thought their parents would be very upset. These percentages increased to 94, 95, and 97 percent of adolescents perceiving that their parents would be very upset if they smoked marijuana one to two times a week, tried heroin once or twice, and used cocaine once a month, in that order. The percentage of adolescents indicating their parents would be not at all upset about use of a particular substance was generally quite low, ranging from 2 to 5 percent.

Factors, by race/ethnicity, gender, and age. Whites were more likely than Hispanics who, in turn, were more likely than blacks to report arguing with parents in the past year (76 vs. 64 vs. 51 percent, respectively) (Table B.6a). However, whites were less likely than either of the other two groups to report that their parents were too strict about dress (6 vs. 10 percent for blacks and 11 percent for Hispanics), homework (15 vs. 28 percent for blacks and 20 percent for Hispanics), or curfew (21 vs. 35 percent for blacks and Hispanics). One possible explanation is that white adolescents' greater tendency for arguing with parents may include arguing about these parental policies, leading to the higher level of parental latitude. Females were also significantly more likely than males (27 vs. 23 percent) to report arguing with their parents several times per week. In general, younger adolescents were more likely to believe their parents were too strict in their family management policies.

Black and Hispanic parents were more often perceived as being very upset if the teen used marijuana once or twice (85 percent for blacks and 89 percent for Hispanics vs. 79 percent for whites), smoked at least one pack of cigarettes per day (84 percent for blacks and Hispanics vs. 80 percent for whites), or tried inhalants once or twice (93 percent for blacks and Hispanics vs. 88 percent for whites) (Table B.6b). In general, older adolescents were less likely than younger adolescents to perceive their parents as getting very upset at substance use; however, both younger and older adolescents perceived their parents as getting very upset if they used heroin or cocaine.

Peer/Individual Domain

Peer and individual risk and protective factors measured by the NHSDA include perceptions of risk of substance use, delinquent behaviors, friends' substance use, and friends' attitudes toward substance use.

Friends' attitudes toward substance use. Respondents were asked questions about the attitudes of close friends regarding substance use. These questions were similar to those dealing with parental attitudes. Youths were more likely to think that their parents rather than their friends would get very upset at their substance use (see Figures 2.3a and 2.3b, Tables B.7a and B.7b). Only 40 percent of youths thought their friends would get very upset if they smoked at least one pack of cigarettes a day. Percentages of youths perceiving friends as very upset varied with the substance as follows: trying marijuana once or twice (46 percent), smoking marijuana once a month (51 percent), and binge drinking one to two times a week (53 percent). Almost 58 percent of adolescents thought their friends would be very upset if they smoked marijuana one to two times a week or tried inhalants once or twice. The percentage of adolescents who thought their friends would get very upset about their substance use jumped to 69 percent for heroin use and 72 percent for monthly cocaine use. Although the percentage of adolescents who perceived that their parents would be not at all upset about their use of various substances ranged from only 2 to 5 percent, the percentage who felt similarly about their friends ranged from 9 to 28 percent.

Friends' use of substances. In addition to questions about the attitudes of close friends toward substance use, the 1997 NHSDA asked how many close friends had used substances (Table B.7c). The majority of youths indicated they had some close friends who were pack-a-day smokers (53 percent) and some who tried or used marijuana in the past year (55 percent). More than 4 out of 10 respondents (44 percent) had a few close friends who were binge drinkers on one or two occasions per week. Also, 3 out of 10 respondents had a few friends who had tried heroin or inhalants or used cocaine once a month.

Perceptions of risk of substance use. The survey also examined how risky adolescents found the following behaviors to be: monthly marijuana use, smoking one or more packs of cigarettes per day, binge drinking one to two times per week, and using cocaine once a month (Table B.7d). Respondents indicated "no risk," "slight risk," "moderate risk," or "great risk" for each of these substance use behaviors. About one in three youths (31 percent) found monthly marijuana use to be a great risk, a smaller percentage than for monthly cocaine use (54 percent), smoking at least one pack of cigarettes per day (54 percent), and binge drinking one to two times per week (47 percent).

Delinquency. Almost 8 percent of adolescents aged 12 to 17 reported having been involved in the delinquent activity of gang fighting in the past 12 months (Table B.7e). More than 17 percent of adolescents had engaged in past year shoplifting.

Factors, by race/ethnicity, gender, and age. More often than close friends of white or Hispanic teens, black teens' close friends were perceived as tolerating (not becoming very upset at) high levels of marijuana use, such as one to two times a week (49 vs. 42 percent, whites, and 40 percent, Hispanics) (Table B.7a). Fewer blacks than whites or Hispanics had close friends who binge drank one to two times per week (37 percent) or who had tried inhalants or heroin or used cocaine monthly (19 percent) (Table B.7c). Across the board, females were more likely than males to perceive that their friends would disapprove of their substance use, a pattern that was not found in relation to parents' substance use attitudes (Tables B.7a and B.7b). The pattern of increased age being associated with increased tolerance, among friends, of substance use was similar to that found when looking at parental substance use attitudes. The likelihood of having a few close friends who engaged in substance use in the past year also increased with age.

There also were racial/ethnic, gender, and age differences in the perceptions of risk and delinquent behavior, such as shoplifting and gang fighting (Tables B.7d and B.7e).

School Domain

School risk and protective factors measured by the NHSDA include enrollment, grades achieved, and formal antidrug education programs.

Commitment to school and academic performance. The two key school-level risk factors measured in the 1997 NHSDA were low commitment to school, as measured by current enrollment status, and academic performance level, as measured by last semester grades (Table B.8). Nearly 98 percent ofrespondents were enrolled in school, and of those respondents reporting last semester grades, over half received mostly A's or B's (51 percent) and only 3 percent made mostly D's or below.

Factors, by race/ethnicity, gender, and age. Only small variations can be seen in the low rates of school nonenrollment (whites, 2 percent; blacks, 3 percent; Hispanics, 5 percent) (Table B.8). Very clear differences can be seen in attaining mostly A's or B's in school (whites, 55 percent; blacks, 35 percent; Hispanics, 38 percent).

General Domain

The general domain of risk and protective factors includes social support, participation in activities, exposure to antidrug media messages, intensity of religious beliefs and observance, and exposure to prevention messages.

Social support, activities, and religious beliefs and practices. Adolescents were asked to whom they would talk about a serious problem. This question served to assess the extent to which youths had access to socioemotional support and where that support would be sought. Possibilities included mother, father, siblings, other relatives, other adults, and nonadult friends. Having good access to parental support is a protective factor for substance use. For measures of involvement in past year activities, youths were asked whether they participated in various activities, including 4-H, private music lessons, and student government. Youths who are significantly involved in activities are associated with lower levels of substance use. The survey also asked about frequency of attendance at religious services, perceptions of importance of religious beliefs, whether beliefs influenced personal decisions, and importance of friends sharing their religious beliefs. Low religious commitment has been associated with higher levels of substance use in a number of studies (Petraitis et al., 1998).

The majority of adolescents would talk to a parent (80 percent) or a friend (83 percent) about a serious problem, and more youths would prefer to talk to a parent (51 percent) than anyone else (Figure 2.4, Table B.9a). Almost three quarters (74 percent) of youths had participated in an extracurricular activity in the past year, and more than 37 percent had been involved in at least three activities (Table B.9b). About 42 percent of adolescents indicated attending religious services weekly in the past year (Table B.9c). A large percentage of adolescents indicated their religious beliefs were important (84 percent) and influenced their decisions (76 percent); 32 percent reported it was important for their friends to share their religious beliefs.

Factors, by race/ethnicity, gender, and age. In terms of social communication, activities, and religious beliefs and practices, white adolescents seemed somewhat more attached to school and peers, and black adolescents to church and extended family. For example, whites were more likely than blacks or Hispanics to indicate a friend as the most likely person they would talk to about a serious problem (45 vs. 32 and 33 percent, respectively), while blacks were the most likely to report having a relative other than parent or sibling to talk to about serious problems (72 vs. 66 percent of whites and 64 percent of Hispanics) (Table B.9a). There also were differences in participation in extracurricular activities, with white, female, and older youths more involved in these activities (Table B.9b). Blacks were more likely to participate in church-related activities and to attend religious services at least monthly in the past year (Table B.9c).

General Domain: Exposure to Prevention Messages

The final type of protective factor is any antidrug prevention activity that seeks to increase the youth's perception of the risk or harm of substance use. Such protective factors include communication between youths and their parents or other adults specifically about the dangers of drug and alcohol use, alcohol and drug prevention education classes in school, and alcohol and drug prevention messages outside of school, such as on the radio and in television ads. These factors reflect the emphases of national initiatives, such as the Drug-Free Communities and National Youth Anti-Drug Media Campaign of the Office of National Drug Control Policy (ONDCP) and the Department of Education's Safe & Drug-Free Schools.

The majority of adolescents aged 12 to 17 indicated having spoken with a parent or other adult about drugs or alcohol (55 percent) or having received in-school alcohol/drug education (57 percent) in the past 12 months (see Figure 2.5 and Table B.10). This percentage was even higher for having seen or heard an alcohol or drug prevention message outside of school in the past year (85 percent).

Factors, by race/ethnicity, gender, and age. There were few differences by race/ethnicity and gender in terms of exposure to prevention messages. Age had a more dramatic relationship with in-school alcohol and drug education than with the other prevention messages, with increased age associated with a decline in the receipt of in-school alcohol/drug education classes in the past year (e.g., 65 percent for 12 and 13 year olds vs. 42 percent for 17 year olds) (Table B.10).

Comparisons of Estimates with Different Age Distributions

Adolescence is a period of very rapid behavioral change, especially with regard to substance use, and it can be misleading to treat 12 to 17 year olds as an undifferentiated age group. Hence, the adolescent sample represented in the 1997 NHSDA can be thought of as six consecutive 1-year age groups. For all substances, a pattern of steadily increasing substance use at each age emerged. This strong positive correlation can often hide the true nature of underlying relationships among risk and protective factors and between these factors and substance use. For example, students were asked to mark each type of in-school alcohol/drug education class they had taken in the past year. The types of classes and the percentage taking each class were (1) a special course about alcohol or drugs taught by a regular teacher (39 percent); (2) a special course taught by someone other than a regular teacher (38 percent); (3) special classes or experience (like a field trip) outside of regular classes (18 percent); or (4) some other school-based drug or alcohol education experience, which was usually described either as a module within a regular class, such as health or physical education, or as a special assembly, or speakers (16 percent). We tabulated the percentage of youths aged 12 to 17 who had used marijuana in the past year by whether or not they had taken in the past year a special course about alcohol or drugs taught by someone other than a regular teacher. Youths who had taken such a course were less likely to have used marijuana in the past year (11 percent) than those who had not taken one (19 percent). Therefore, one might conclude that special courses appeared to be effective for youths aged 12 to 17.

However, looking at the results by single year of age, the special courses may be effective for ages 12 to 15, but not for ages 16 and 17 (Figure 2.6). One reason for this result is that those courses are typically aimed more at the younger ages. In fact, Figure 2.7 shows by age the participation in each of the above types of classes. For students reporting taking a special course about alcohol or drugs taught by someone other than a regular teacher, the participation was highest for age 12 and decreased consistently through age 17.

The focus of such programs on the younger ages reflects the targeting of this type of drug prevention programming to middle schoolers, which is to say, that type of program is offered largely to groups who are below the age that most adolescents first try drugs. Overall, about 57 percent of all adolescents received one or more of these drug education exposures during the past year (see also Table B.10). Except for the miscellaneous "other" category, which was at about the same level each year, the exposure to school-based drug education activities declined with age.

The effect of age and other demographic variables is controlled for in the multivariate models described in Chapter 4. More detailed analyses of the risk and protective factors and substance use by age can be found in Appendix D. Tables B.1 to B.4 in Appendix B present the prevalence and level of substance use by age and other demographic characteristics.

Table 2.1 Percentage of the U.S. Civilian, Noninstitutionalized Population Aged 12 to 17 Reporting Risk and Protective Factors: 1997
 

Domain: Risk/ProtectiveFactor	N (in 1,000s)	Total		N (in 1,000s)	Total
Community: Drug Availability (Table B.5)
Marijuana			Cocaine
Fairly easy-very easy to get	12,972	57.9	Fairly easy-very easy to get	6,787	30.4
Fairly difficult- probably impossible to get	9,422	42.1	Fairly difficult- probably impossible to get	15,542	69.6
Crack			LSD
Fairly easy-very easy to get	6,456	28.9	Fairly easy-very easy to get	6,215	28.0
Fairly difficult- probably impossible to get	15,889	71.1	Fairly difficult- probably impossible to get	16,003	72.0
Heroin
Fairly easy-very easy to get	4,780	21.4
Fairly difficult- probably impossible to get	17,579	78.6
Anyone offered/tried to sell marijuana			Anyone offered/tried to sell cocaine
Yes	7,517	35.0	Yes	2,075	9.7
No	13,948	65.0	No	19,228	90.3
Family: Family Management (Table B.6a)
Parental strictness about dress			Parental strictness about homework
Not at all strict	9,783	44.5	Not at all strict	5,513	25.2
Just strict enough	10,684	48.6	Just strict enough	12,519	57.2
Too strict	1,519	6.9	Too strict	3,868	17.7
Parental strictness about curfew
Not at all strict	2,778	12.7
Just strict enough	13,565	61.9
Too strict	5,565	25.4
Family: Family Conflict (Table B.6a)
Argued with parents in past year
At least several times per week	5,556	25.2
Once per week-once per month	9,909	45.0
Rarely or never	6,575	29.8

Table 2.1 (continued)
 

Domain: Risk/ProtectiveFactor	N(in 1,000s)	Total		N(in 1,000s)	Total
Family: Prevention Message (Table B.10)
Spoke with parent/other adult about drugs/alcohol in past year
Yes	12,059	54.5
No	10,064	45.5
Family: Parents' Attitudes Toward Drug Use (Tables B.6b and B.6c)
If smoked 1+ packs of cigarettes per day			If tried marijuana once or twice
Not at all upset	1,017	4.6	Not at all upset	893	4.0
Somewhat upset	3,148	14.2	Somewhat upset	3,404	15.4
Very upset	17,995	81.2	Very upset	17,854	80.6
If had 5+ drinks 1-2 times a week			If smoked marijuana once a month
Not at all upset	665	3.0	Not at all upset	662	3.0
Somewhat upset	1,846	8.3	Somewhat upset	1,731	7.8
Very upset	19,625	88.7	Very upset	19,737	89.2
If tried inhalants once or twice			If smoked marijuana 1-2 times a week
Not at all upset	522	2.4	Not at all upset	537	2.4
Somewhat upset	1,858	8.4	Somewhat upset	792	3.6
Very upset	19,772	89.3	Very upset	20,800	94.0
If tried heroin once or twice			If used cocaine once a month
Not at all upset	404	1.8	Not at all upset	404	1.8
Somewhat upset	632	2.9	Somewhat upset	323	1.5
Very upset	21,104	95.3	Very upset	21,411	96.7
Peer/Individual: Friends' Attitudes Toward Drug Use (Tables B.7a and B.7b)
If smoked 1+ packs of cigarettes per day			If tried marijuana once or twice
Not at all upset	6,226	28.1	Not at all upset	5,956	26.9
Somewhat upset	7,132	32.2	Somewhat upset	5,988	27.1
Very upset	8,759	39.6	Very upset	10,161	46.0
If had 5+ drinks 1-2 times a week			If smoked marijuana once a month
Not at all upset	4,906	22.2	Not at all upset	5,351	24.2
Somewhat upset	5,516	25.0	Somewhat upset	5,476	24.8
Very upset	11,678	52.8	Very upset	11,254	51.0
If tried inhalants once or twice			If smoked marijuana 1-2 times a week
Not at all upset	3,377	15.3	Not at all upset	4,276	19.3
Somewhat upset	5,996	27.1	Somewhat upset	5,090	23.0
Very upset	12,712	57.6	Very upset	12,744	57.6

Table 2.1 (continued)
 

Domain: Risk/ ProtectiveFactor	N (in 1,000s)	Total		N (in 1,000s)	Total
If tried heroin once or twice			If used cocaine once a month
Not at all upset	2,023	9.2	Not at all upset	1,915	8.7
Somewhat upset	4,890	22.1	Somewhat upset	4,310	19.5
Very upset	15,188	68.7	Very upset	15,862	71.8
Peer/Individual: Friends' Use of Drugs; At Least a Few Close Friends Have... (Table B.7c)
Smoked 1+ packs of cigarettes per day			Had 5+ drinks 1-2 times a week
Yes	11,748	53.4	Yes	9,643	43.8
No	10,272	46.6	No	12,378	56.2
Tried or used marijuana			Tried inhalants or heroin once or twice or used cocaine monthly
Yes	12,045	54.7	Yes	6,597	30.0
No	9,991	45.3	No	15,376	70.0
Peer/Individual: Perception of Risk of Drug Use (Table B.7d)
Risk if smoke 1+ packs of cigarettes per day			Risk if smoke marijuana once a month
Great risk	12,060	53.6	Great risk	6,925	30.9
No risk-moderate risk	10,421	46.4	No risk-moderate risk	15,479	69.1
Risk if have 5+ drinks 1-2 times a week			Risk if use cocaine once a month
Great risk	10,421	46.5	Great risk	12,173	54.4
No risk-moderate risk	11,978	53.5	No risk-moderate risk	10,213	45.6
Peer/Individual: Delinquency (Table B.7e)
Gang fight in past year			Shoplifted in past year
Yes	1,743	7.9	Yes	3,838	17.4
No	20,292	92.1	No	18,203	82.6
School: Commitment to School and Academic Performance (Table B.8)
Enrolled in school			Last semester grades
Yes	21,985	97.6	Mostly A's or B's	9,844	50.6
No	547	2.4	Mostly B's or C's	6,727	34.6
			Mostly C's or D's	2,347	12.1
			Mostly D's or below	529	2.7

Table 2.1 (continued)
 

Domain: Risk/ProtectiveFactor	N (in 1,000s)	Total		N (in 1,000s)	Total
School: Prevention Message (Table B.10)
In-school alcohol/drug education class in past year
Yes	12,825	56.9
No	9,722	43.1
General: Social Support; Who Would Talk to About Serious Problem... (Table B.9a)
Parent(s)			Other relative(s)
Yes	17,581	79.8	Yes	14,577	66.3
No	4,445	20.2	No	7,422	33.7
Friend(s)			Other person(s)
Yes	18,285	83.1	Yes	12,212	55.5
No	3,718	16.9	No	9,784	44.5
Most likely parent(s)			Most likely friend(s)
Yes	11,166	50.8	Yes	9,106	41.5
No	10,794	49.2	No	12,854	58.5
General: Activities (Table B.9b)
Participated in past year			Number of activities in past year
Yes	16,714	74.1	3+ activities	8,476	37.6
No	5,833	25.9	2 activities	3,800	16.9
			1 activity	4,438	19.7
			0 activities	5,833	25.9
Sports/physical activities			Church-related activities
Yes	11,142	57.0	Yes	2,930	15.5
No	8,389	43.0	No	15,939	84.5
Music/art/performing arts			Club/youth group
Yes	7,187	37.7	Yes	9,028	46.7
No	11,893	62.3	No	10,311	53.3
Student government/ROTC/other civic activity
Yes	2,119	11.2
No	16,788	88.8

Table 2.1 (continued)
 

Domain: Risk/ProtectiveFactor	N (in 1,000s)	Total		N (in 1,000s)	Total
General: Religious Beliefs and Practices (Table B.9e)
Attended religious services past year			My religious beliefs are very important
At least once a week	9,334	41.9	Agree	18,735	84.3
Once or twice a	4,981	22.4	Disagree	3,493	15.7
0-2 times	7,961	35.7
My religious beliefs influence my decisions			It is important for my friends to share my religious beliefs
Agree	16,831	75.8	Agree	7,188	32.3
Disagree	5,369	24.2	Disagree	15,042	67.7
General: Prevention Message (Table B.10)
Saw/heard alcohol/drug prevention messages outside of school in past year
Yes	18,786	84.6
No	3,417	15.4

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

In this chapter, we present relationships between reported levels of risk and protective factors and past year marijuana use among adolescents in the 1997 NHSDA. We report those data here in the form of simple cross-tabulations. When we observe a statistical relationship in cross-sectional data such as these, inferences about cause and effect are not inherent in the results. However, findings may be informative about the mutual association between certain risk factors and substance use. For example, if a person is offered drugs by someone, that may increase the chances that the person will use drugs. At the same time, if a person is already using drugs, that may increase the chances of someone offering drugs to that person.

First, we explore associations of risk and protective factors with past year use of marijuana. We also analyze past year use of alcohol, tobacco/cigarettes, and illicit drugs other than marijuana and find associations with risk and protective factors similar to our findings for marijuana. Therefore, we limit our discussion in the text to marijuana but include tables in Appendix B for the other substances.

Second, we examine the issue of differences by race/ethnicity and gender to see whether any differences in levels of risk and protective factors discussed in Chapter 2 correlate with differences in prevalence of substance use by different racial/ethnic groups.

Finally, we present a summary of the relationships between risk and protective factors and substance use from strongest to weakest, at the level of simple statistical relationships, and note the patterns among them. The data that correspond to the results presented in this chapter can be found in Tables B.11 to B.34 in Appendix B. Appendix C contains standard error tables to accompany the tables of past year marijuana use. The fourth chapter expands on this material using multivariate statistical models to examine individual factors while controlling for the presence of other factors in the model.

To place the findings of this chapter in context, the past year prevalence rates among 12 to 17 year olds for the substances studied were marijuana, 15.2 percent, cigarettes, 23.7 percent, alcohol, 32.5 percent, and any illicit drugs other than marijuana, 9.5 percent. These rates were slightly lower than those presented in OAS (1999b) due to the creation of new past year usage measures. The substance use measures reported here were created by summing the percentages of users at various frequencies of use rather than utilizing the percentages of persons who reported use within the past year regardless of frequency of use (see Table B.1 for an example). For the measure of past year use of any illicit drugs other than marijuana, youths were first categorized into the highest frequency of use across five drugs that make up "any illicit drugs other than marijuana" (cocaine/crack, inhalants, hallucinogens, heroin, nonmedical use of any psychotherapeutic drug). The measure of past year use was then created by summing percentages over these frequency of use categories.

Community Domain

Drug availability. The data presented in Figures 3.1a and 3.1b reveal a strong relationship between perceived drug availability and past year marijuana use among adolescents aged 12 to 17. Youths who thought it would be easy to get drugs, except for heroin, or who had been offered marijuana or cocaine, were more likely to report past year marijuana use than those who indicated drugs were difficult to obtain or who were never offered drugs (Table B.11). The greatest contrast was between those who had ever been offered marijuana and those who had never been offered it. Only 2.4 percent of youths who had never been offered marijuana reported past year marijuana use, compared with 39.7 percent of youths who had ever been offered it. To put this in the statistical language of probabilities, one would say that the likelihood (or odds) of having used marijuana in the past year, for a youth who had never been offered marijuana, was 97.6 to 2.4-that is, about 41 to 1. On the other hand, the odds of having used marijuana in the past year, for a youth who had been offered marijuana, were 60.3 to 39.7, or about 1½ to 1. The ratio between these odds, the "odds ratio," is roughly 41 divided by 1½, or 27 to 1 (abbreviated as "odds ratio of 27").

The odds ratio is a good measure of the observed strength of the relationship between a dichotomous ("yes"/"no") risk factor and the behavior at risk-with the caveat that even a strong statistical relationship is not necessarily one of cause and effect. Nevertheless, it is worth noting that risk factors in epidemiological research in the general population rarely produce odds ratios above the low single digits. An odds ratio of 27 for the perceived availability of marijuana is in the range usually observed, for example, between decades of smoking more than a pack of cigarettes each day (versus never smoking) and developing lung cancer.

Similarly, youths who indicated that it was easy to get marijuana had odds of using marijuana that were 15 times higher than those who said it was difficult to obtain. We found very similar relationships between the availability of both marijuana and cocaine and youths' use of the other substances we measured: past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.12 to B.14).

Family Domain

Family management and conflict. Figures 3.2a and 3.2b illustrate the relationship between parental communication risk factors and past year marijuana use. Adolescents who argued with their parents at least several times a week were more likely to have used marijuana in the past year than those who argued with their parents once a week to once a month. This group, in turn, was more likely to have used marijuana in the past year than adolescents who rarely or never argued with their parents (22 vs. 16 vs. 8 percent) (Table B.15a in Appendix B). Adolescents who considered their parents "not at all strict" on such matters as how to dress, homework, and curfew were more likely to report past year marijuana use than those whose parents were either "just strict enough" or "too strict" (23 vs. 9 and 12 percent for dress; 21 vs. 14 and 11 percent for homework; and 23 vs. 14 and 15 percent for curfew). These findings also held for past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.16a, B.17a, and B.18a).

Parents' attitudes toward substance use. The data presented in Figures 3.3a and 3.3b present the relationship between past year marijuana use and the perceived level of parental antipathy toward substance use. The attitudes in each figure are ordered top to bottom by the approximate extent to which (in the youth's opinion) the parents would feel "very upset" if they thought the adolescent had used substances at a specified level. Although not shown in Figure 3.3b, fewer adolescents (about 18 million out of 22 million) thought their parents would be very upset about their smoking a pack of cigarettes daily than the number (about 21 million) who thought their parents would be very upset if the youths used cocaine once a month or tried heroin once or twice (see Tables B.15b and B.15c). The data displayed in Figures 3.3a and 3.3b are the percentages of youths who used marijuana for each level of perceived response by the parents.

For marijuana, cigarettes, and binge drinking, 12 to 17 year olds who perceived that their parents would be very upset reported the lowest prevalence of marijuana use in the past 12 months (Table B.15b). For inhalants, differences were in the expected direction (youths who thought that their parents would be very upset if the youths used the substances reported lower levels of marijuana use than youths who thought that their parents would be not at all upset), but they were smaller and not statistically significant (Table B.15c). The results for heroin and cocaine appear quite anomalous, with youths who thought that their parents would be very upset reporting higher prevalences, although not significantly higher, than those who thought that their parents would be not at all upset. These may be in part due to the small numbers of youths who did not think that their parents would be very upset and the resulting large sampling error. In addition, youths whose parents would be "somewhat upset" if their child smoked marijuana once or twice a week registered higher prevalence levels for past year marijuana use than youths whose parents would be "not at all upset."

Peer/Individual Domain

Friends' attitudes toward substance use. It is enlightening to compare Figures 3.3a and 3.3b with Figures 3.4a and 3.4b, in which data from questions about perceived friends' attitudes toward substances were asked. On the whole, the attitudes ascribed to close friends were more strongly related to drug use than the reported parental attitudes were. For example, for youths whose parents would be "not at all" upset about their child trying marijuana once or twice, the odds ratio of past year marijuana use relative to the "very upset" group was about 11 ([91.3/8.7]/[49.6/50.4]). But for youths whose close friends who would be "not at all" upset, the comparable odds ratio was nearly 35 ([(97.8/2.2]/[56.1/43.9]) (Tables B.15b and B.19a in Appendix B). In other words, peer attitudes had more than three times as much influence compared to parental attitudes about this particular substance use pattern. Moreover, youths who perceived only "somewhat upset" parents had high rates of marijuana use (20 to 59 percent), while youths who perceived only "somewhat upset" friends had low to middling rates of marijuana use (9 to 19 percent) (Tables B.15b, B.15c, B.19a, and B.19b). These overall patterns were also observed for past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.20a, B.20b, B.21a, B.21b, B.22a, and B.22b).

The data suggest the following conclusions:

A perception of strong antidrug attitudes (i.e., "very upset") appears to be a powerful inhibitor of drug use among close friends, either because it keeps close friends from using drugs or because it inhibits close friendships between drug users and adolescents with antidrug attitudes. On the other hand, an adolescent's use of drugs may in itself create the perception that one's close friends are less negative about drug use.

Strong antidrug attitudes by parents are also inhibitors of drug use, but not with nearly the same strength as peer attitudes. This finding may be explained by the fact that it is easier for a teenager to change friends than to change parents, or that it is easier for a teenager to change the attitudes of their friends than those of their parents.

Even mild antidrug attitudes (i.e., "somewhat upset") on the part of close friends are associated with lower levels of individual marijuana use, but mild antidrug attitudes by parents are no more inhibiting or incompatible with respect to youth drug use than parental dispassion (i.e., "not at all upset").

Perceptions of risk of substance use. With one exception, perception of great risk of substance use was associated with significantly lower past year substance use than perceptions of no or moderate risk of substance use (Tables B.19d, B.20d, B.21d, and B.22d). The chart of the relationship between perceptions of risk and marijuana use is displayed in Figure 3.5. Adolescents who thought there was no risk or moderate risk if one smoked marijuana monthly, for example, were more than nine times as likely as those who perceived great risk in this level of substance use to have used marijuana in the past 12 months. Past year marijuana use was unrelated only to perceptions of risk for monthly cocaine use (Table B.19d).

Delinquency. Shoplifting and gang fighting were also associated with higher rates of marijuana use (Table B.19e). About 40 percent of respondents who had been involved in such delinquent behavior had also used marijuana in the past year. Similar relationships were found with past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.20e, B.21e, and B.22e).

School Domain

Commitment to school and academic performance. The relationships between the academic risk factors (enrollment status and last semester's grades) with past year marijuana use are presented in Figure 3.6. Dropping out of school or having low grades has been associated with a lower commitment and attachment to school and to a higher risk of substance use. Only 10 percent of adolescents who made mostly A's or B's used marijuana in the past year, but this prevalence rate was 46 percent for students who made mostly D's or below (Table B.23). Adolescents who were not currently enrolled in school were more likely to report past year marijuana use than those who were enrolled in school (42 vs. 15 percent). Very similar findings held for past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.24, B.25, and B.26).

General Domain

Social support. A summary of the results of the relationship between marijuana use and social support is displayed in Figures 3.7a and 3.7b. Adolescents who would talk to a parent about a serious problem were less likely to report past year marijuana use than those who would not (11 vs. 30 percent) (Table B.27a). This difference was in the neighborhood of an odds ratio of 3.5-not as large as observed for some other factors, but statistically significant and by no means negligible. Moreover, adolescents who could talk to some other relative (e.g., a sibling) or some other adult (e.g., a teacher) were also less likely to have used marijuana in the past year than those who could not (14 vs. 18 percent, some other relative; 11 vs. 20 percent, some other adult). In contrast, adolescents who said they would talk to a friend, especially those who were most likely to do so, were more likely to report past year marijuana use than those who would not (with an odds ratio in this reverse direction of about 2.6 for youths most likely to talk to a friend). These same patterns held for cigarette, alcohol, and illicit drug use other than marijuana (Tables B.28a, B.29a, and B.30a). These data suggest that the availability and use of family and other adult support operate as a protective factor, while a primary reliance on peer support is, in contrast, a risk factor for substance use.

Activities. The data presented in Figure 3.8 indicate that for any given activity, those adolescents who participated were significantly less likely to report past year marijuana use than those who did not (with odds ratios ranging from about 1.4 to 2.7) (Table B.27b). There was no significant difference between the prevalence of past year marijuana use among youths with no past year activities and those with only one past year activity (18 and 20 percent, respectively). However, youths who participated in more than one type of activity reported lower prevalence rates than those who participated in one type only. In addition, youths who participated in three or more activities reported a lower prevalence rate than youths who were less involved (11 vs. 20 percent).

Youthful participation in activities correlated somewhat differently with substances other than marijuana (Tables B.28b, B.29b, and B.30b). We obtained statistically significant differences in the prevalence of past year alcohol and illicit drug use other than marijuana for participants of certain types of activities-but not all types. Youths who participated in church-related and arts activities were less likely to report past year alcohol use than youths who did not participate in these activities. Youths who participated in sports, church-related, and arts activities were less likely to report past year illicit drug use, not including marijuana, than youths who did not participate in these activities.

Religious beliefs and practices. Figure 3.9 presents data for several items that define, in broad terms, the depth of religious commitment and association between religious beliefs and practices and past year marijuana use. The items are attendance at religious services weekly or more often, indication that one's religious beliefs are personally very important, indication that such beliefs influence one's personal decisions, and affirmation that it is important for the youths' friends to share their religious beliefs. In all instances, those reporting higher religious commitments were less likely to use marijuana, with odds ratios between 2 and 3 (Table B.27c). These same overall relationships were found for past year cigarette, alcohol, and illicit drug use other than marijuana (Tables B.28c, B.29c, and B.30c).

General Domain: Exposure to Prevention Messages

We expected that exposure to the prevention messages of in-school drug education, talking with a parent or other adult about the dangers of drugs and alcohol, and hearing prevention messages on the radio or television would all be associated with lower levels of drug use. However, as indicated in Figure 3.10, the results were mixed.

As expected, adolescents who had received no in-school drug/alcohol education in the past year reported higher past year marijuana use (the odds ratio is about 1.4), when compared with those who had received in-school drug/alcohol education (Table B.31). On the other hand, we found no significant difference in the prevalence of past year marijuana use between youths who had spoken with a parent or other adult in the past year about the dangers of alcohol and drug use and those who had not. In addition, youths who had seen or heard drug/alcohol prevention messages outside of school in the past year were somewhat more likely (with an odds ratio of almost 1.3) to report past year marijuana use than those who had not been exposed.

However, the lack of detail regarding these items makes us cautious in interpreting our results. For example, an adult might have initiated a discussion about the hazards of drug use, not as part of a general effort toward preventing drug use, but in specific reaction to learning that the youth was using drugs or spending time with other youths known or reputed to be drug users (in other words, at higher risk of using). Moreover, the youth rather than the adult may have initiated the discussion.

In an effort to obtain a better understanding of these somewhat anomalous findings, we looked for variables that might have confounded the expected relationship. We examined each of the prevention messages among four different variables:

detailed age subgroups (12 to 14 and 15 to 17 years of age),

gender (male and female),

onset of drug use (onset of drug use 2 or more years ago vs. onset less than 2 years ago or no drug use), and

amount of arguing with parents (argued with parents at least several times per week vs. argued with parents less than weekly).

Race/Ethnicity and Gender Differences

As discussed in Chapter 2, different racial/ethnic or gender groups reported high and low levels of different risk and protective factors. For example, whites and blacks were more likely than Hispanics to say that marijuana was easy to get. Females were more likely than males to perceive that their friends woulddisapprove of substance use. Despite these differences, the prevalence of risk and protective factors within a racial/ethnic or gender group does not appear to have much impact on the level of use for some factors, while for other factors, the impact on prevalence is more pronounced. In the interest of brevity, we only present results for two of the factors that were discussed in Chapter 2. Tables B.35 to B.40 in Appendix B present the complete results for all of the risk and protective measures by race/ethnicity and gender and by past year marijuana use.

Availability. As noted in Chapter 2 and Table B.5, a larger percentage of whites and blacks than Hispanics indicated that marijuana was easy to get (60 vs. 52 percent). For white youths who considered marijuana was easy to get, Table B.35 shows that 26 percent of them had used marijuana in the past year. For white youths who perceived marijuana was difficult to get, only 2 percent of them used marijuana in the past year. Therefore, odds of using marijuana in the past year for white youths were more than 21 times higher if they perceived marijuana as easy to obtain versus difficult. For blacks and Hispanics, the odds ratios were around 6 and 9, respectively. In a similar vein, although females were more likely than males to perceive that marijuana was easy to get, they were no more likely to use it.

Parental attitudes. As noted in Chapter 2 and Table B.6b, the percentage of youths who perceived their parents would be "very upset" if the youths tried marijuana once or twice was slightly smaller for whites (79 percent) than blacks or Hispanics (85 and 89 percent, respectively). Table B.36c shows that similar percentages of white, black, and Hispanic youths who perceived that their parents would be "very upset" reported use of marijuana in the past year: 8, 9, and 9 percent, respectively. On the other hand, for white youths, 62 percent of those who thought that their parents would be "not at all upset" used marijuana in the past year, a percentage much elevated compared to that for blacks and Hispanics (23 and 30 percent, respectively); however, the latter percentages have low precision, as noted in Table B.36c.

Summary Measures for the Risk and Protective Factors

The statistical results for past year marijuana use, relative to the risk and protective factors measured in the 1997 NHSDA, are summarized in Table 3.2 by the size of their odds ratios. Only those factors with an odds ratio greater or equal to 2.5 are displayed. Measures of peer substance use behavior, peer attitudes toward substance use, and availability of drugs dominate the picture. Because drug offers probably occur within peer networks, even the availability measures may reflect the strength of peer ties as much as they do marketing. The lesson here may be that adolescent substance use needs to be viewed more as a collective, rather than individual, behavior. Parent-child relationships, including parental attitudes toward drugs, are by no means insignificant, and they appear to be more prominent than school or religious influences in providing protection or inducing risk.

Note: Data in table read "6.4 percent of 12-14 year olds who had spoken with a parent or other adult about drugs or alcohol in the past year used marijuana in the past year."

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Table 3.2 Odds Ratios of Past Year Marijuana Prevalence, by Risk Factors: 1997

Risk Factor	Odds Ratio	Lower Limit	Upper Limit
At least a few close friends tried or used marijuana	38.5	24.9	59.6
Was ever offered marijuana	26.7	20.0	35.6
Friends not very upset if tried marijuana 1-2 times	15.9	12.8	19.7
Marijuana easy to get	15.2	11.0	21.0
Friends not very upset if smoked marijuana once a month	14.2	11.7	17.3
Friends not very upset if smoked marijuana 1-2 times a week	11.4	9.3	13.9
Perception of not great risk for monthly marijuana use	9.6	6.8	13.6
Parents not very upset if smoked marijuana once a month	9.6	7.4	12.4
Was ever offered cocaine	7.7	6.3	9.5
Parents not very upset if tried marijuana 1-2 times	7.5	6.0	9.5
Friends not very upset if smoked 1+ packs of cigarettes per day	7.1	5.9	8.6
At least a few close friends smoked 1+ packs of cigarettes per day	6.8	5.2	8.8
Parents not very upset if smoked marijuana 1-2 times a week	6.7	5.1	8.9
Shoplifted in past year	6.1	4.9	7.7
Parents not very upset if smoked 1+ packs of cigarettes per day	5.9	4.9	7.1
Friends not very upset if had 5+ drinks 1-2 times a week	5.5	4.5	6.8
At least a few close friends had 5+ drinks 1-2 times a week	5.2	4.1	6.7
At least a few close friends tried inhalants or heroin 1-2 times or used cocaine monthly	4.8	4.0	5.8
Gang fight in past year	4.6	3.6	5.8
Not enrolled in school	4.3	2.9	6.5
Parents not very upset if had 5+ drinks 1-2 times a week	3.5	2.9	4.3
Friends not very upset if tried inhalants once or twice	3.5	2.9	4.3
Would not talk to parent(s) if serious problem	3.5	2.9	4.2
LSD easy to get	3.2	2.7	3.8
Not most likely to talk to parent about problem	3.1	2.5	3.8
Parents not at all strict regarding dress	3.0	2.5	3.6
Perception of not great risk for having 5+ drinks 1-2 times a week	2.9	2.3	3.7
Does not attend religious services once a week	2.9	2.4	3.5
No music/arts activities	2.7	2.0	3.5
Most likely talk to friend if serious problem	2.6	2.1	3.2
Grades not mostly A's or B's	2.5	2.1	3.1
Not important for friends to share religious beliefs	2.5	1.9	3.2

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Earlier chapters explored the prevalence of risk and protective factors and the relationship of those factors to substance use, especially marijuana, using simple cross-tabulations and odds ratios. This chapter analyzes the strength of the relationship between risk and protective factors and substance use using multivariate methods. We seek answers to five questions:

How important are each of the risk and protective factors domains in predicting substance use?

How important are the demographic factors in predicting substance use?

How much do risk and protective variables add to the prediction of substance use beyond the demographic factors?

Are the demographic variables that seemed important based on earlier cross-tabulations still important in the multivariate analysis?

How important are all of the risk and protective factors together in explaining the variation in past year substance use?

The word "prediction" is used not to imply that we have actually observed a sequence of events in time, but to describe a statistical question, namely, "How well does statistical information about one thing improve our ability to guess what happened to something else?" For example, if knowing the employment status of each person in a group would improve how well we could correctly guess early initiation of alcohol use, we would call employment status a "predictor," without necessarily meaning that employment status came first. Moreover, there are statistical methods for determining just how strong a predictor employment status may prove to be in any given group of people. When we use a number of predictors together in a statistical analysis of this kind, the combination of predictors is referred to as a "prediction model."

Because of the complex survey design of the NHSDA, we performed regression analyses using the LOGISTIC procedure in SUrvey DAta ANalysis (SUDAAN), a statistical program employing variance estimation calculations that take into account this complexity (Shah, Barnwell, & Bieler, 1997). Note that we are using here simple individual (person-level) logistic regression models that adjust for the effects of clustering on the estimates but otherwise ignore the true hierarchical structure of the data, namely, the fact that youths aged 12 to 17 are nested within families that are, in turn, nested in neighborhoods. Therefore, this analysis treats variables at the higher levels of hierarchy as being individual (youth) variables.³ Logistic regression determines the importance of individual independent variables and groups of variables by measuring how much (additional) variation in the dependent variable the independent variable can explain. The lack of statistical significance of an independent variable does not imply that the variable is unimportantin the epidemiology of substance use. For example, the variable may have a significant indirect relationship to the dependent variable through another independent variable in a path analysis, and other techniques, such as structural equation modeling, may be more appropriate for analyzing those relationships.

Before performing regression analyses, for ease of interpretation, we dichotomized the multicategorical variables based on the strength and pattern of their relationship with substance use. For example, the three-level variable, "arguments with parents," was recoded as arguing with parents more than once a week versus arguing fewer times.

Following the lead of earlier chapters, in Chapter 4 we first present results from logistic regression modeling of past year marijuana use. We then briefly compare these results with the findings from modeling of past year cigarette use, alcohol use, and illicit drug use other than marijuana. Note, as discussed in Chapter 3, that the substance use measures used in this report yield estimates of use that are slightly lower than those presented in OAS (1999b).

Past Year Use of Marijuana

The first step in the analysis was to fit logistic regressions predicting past year marijuana use only from the set of major demographic variables: gender, race/ethnicity, age, four Census regions, and population density (urbanicity).⁴ We refer to these demographic-only regressions as Model 1. We grouped the sets of risk and protective factors into the five broad domains (community, family, peer/individual, school, and general) that were discussed in Chapter 1. The questions that are related to each of the domains are presented in Appendix A.

We then fitted second and third logistic regression models to each of these domains (as well as to a variety of other models not reported here, in order to test the robustness of these results). The second set of regressions (Model 2) consisted of all of the variables in a given domain. Because several items within the domains were highly correlated with each other (correlation $.50), in order to eliminate multicollinearity, we reran Model 2 regressions removing the extra variables and leaving in only one of the highly correlated variables to capture that construct variance. The amount of variation explained by the variables left in the domains remained essentially the same as that found in regressions that included all the variables in a domain. The third regression (Model 3) included both the specific domain variables and the demographic variables. In other words, for each domain of risk and protective factors, all the elements of Model 1 and Model 2 were combined in Model 3. For this chapter, we present results from Model 2 and Model 3 regressions that included all of the variables in each of the domains.

At the end of this chapter, Table 4.1 presents the results of three models. The first model (column 1) was based on a set of five demographic variables: gender, race/ethnicity, age, four Census regions, and population density (rural vs. nonrural). The second model (column 2) includes the community domain variables. The third model (column 3) combines the demographic variables with community risk and protective factors. Each model presents the coefficient for each variable, whether it is significant (p value

less than .05), and two measures that summarize the explanatory power for each model.⁵ The measure of the significance of each variable in a model is conditional on the fact that the other variables are in the model, too. For example, age (being 12 to 14 years old) is a significant variable in Model 1, after controlling for gender, race/ethnicity, region, and population density. The coefficient (-1.55) in Table 4.1 indicates that youths aged 12 to 14 are less likely to use marijuana than youth aged 15 to 17. To be more precise, the odds of using marijuana in the past 12 months for youths aged 12 to 14 are approximately .21 times the odds for youths aged 15 to 17, holding other variables in the equation constant. Other significant variables are race/ethnicity (Hispanic vs. white; black vs white), region (South vs. West), and population density (rural vs. nonrural).

The summary measures indicate that the community risk factors explain a significant amount of the variation in marijuana use-significantly more than do the demographic variables by themselves. It also shows that the addition of the demographic factors to the model with the community risk factors does not do much to improve the model. Analysis of the demographic factors reveals that, when added to the community risk factors, only age and population density are still significant. This indicates that the risk factors are capturing some of the same variation that was explained by the now nonsignificant demographic variables. Tables 4.2 to 4.5 show similar results for the other risk domains: family, peer/individual, school, and general. Of all of these factors, the peer/individual domain and the community domain explained the most variation in youth drug use. For these two domains, the measures of explained variation were R² =.30 (R_N²=.53) for peer/individual and R² =.25 (R_N²=.43) for community. Family influences (R²=.17; R_N²=.30) were next, then general factors (R²=.11; R_N²=.19), and finally, school factors (R²=.04; R_N²=.07). For demographics by themselves, the explained variation was R² =.07 (R_N²=.12). However, it should be noted that the estimate of relative contribution of the various domains is based on the limited set of risk and protective questions included in the 1997 NHSDA. The true relative importance of the various domains might be different if the full set of factors were included in each domain.

Tables 4.6 to 4.10 display models that combine the demographic variables together with risk and protective factors from each domain. These combined models included only those variables that were significant in the earlier domain-specific models. Table 4.6 presents the coefficients, odds ratios, and confidence intervals for each of the factors for past year use of marijuana. Due to limited space, the only variables listed are those that were still significant in the combined model. Collectively, the variables in the marijuana model, at this point, accounted for relatively more of the overall variation (R²=.38; R_N²=.64) in past year marijuana use than any of the domains individually.

Because this model included any variable that was significant in earlier domain-specific models, the number of variables was large. In an effort to obtain a more parsimonious model, we limited the final marijuana model to only the significant variables from the combined model. This resulted in a slight reduction in explanatory power and a model with 5 demographic variables and 14 risk and protective factors (see Table 4.7). In an absolute sense, the variables in the model accounted for 35 percent (R_N²=.61) of the total person-level variation in past year marijuana use. The significance of this measure is that the variables in this model accounted for a significant percentage of the total variation in whether a youth used marijuanaor not in the past year. To the extent that the model includes risk and protective factors that have been demonstrated in well-designed prevention programs to reduce marijuana use, application of such programs has the potential of reducing youth marijuana use. By contrast, if the prevention variables had only accounted for a small percentage of the total variation, then programs aimed at reducing the levels of the variables in the model could not be expected to lower usage of marijuana among youths in a significant way. It is worth emphasizing the fact that the NHSDA is an annual cross-sectional survey that provides a snapshot of the relationship between these risk and protective factors and marijuana use for youths who have been surveyed at some point during 1997. Based on their responses, some youths aged 12 to 17 reported that they used marijuana in the past year and indicated the presence of various risk factors. However, the use of marijuana may have preceded the presence of the risk factor for some youths, resulting in an "inflated" R_N square. Therefore, one should be cautious in drawing conclusions about potential changes in youth marijuana use.

The largest odds ratios in the final combined model were associated with factors involving drug availability, perceptions of risk of marijuana use, friends' attitudes toward drug use, and friends' use of marijuana. These odds ratios were conditional on all of the other variables in the model. Youths who had been offered marijuana or who were approached to ask if they would like to buy marijuana had odds 7.10 times higher than other youths. The odds for youths who thought their friends would be not at all upset if they smoked marijuana once a month were 5.23 times higher, and youths who did not see great risk in monthly marijuana use had odds 4.05 times higher. Many of the risk factors having high conditional odds ratios are the same risk factors that were cited as having high unconditional odds ratios at the end of Chapter 3.

In the combined model for marijuana, the demographic variables having conditional odds ratios significantly different from 0 are gender, race/ethnicity, and age. The model provides a more complex explanation for the observed national patterns. For example, the model suggests that when males and females have the same combination of risk and protective factors as identified here, males are actually only two thirds as likely as females to be users. When controlling for all of the risk and protective factors, Hispanics (odds ratio=0.71) and blacks (odds ratio=0.65) were less likely than whites to have used marijuana in the past year, but other races (odds ratio=2.17) were more likely than whites to have used marijuana in the past year. Younger youths aged 12 to 14 (odds ratio=0.54) were still less likely than older youths aged 15 to 17 to have used marijuana in the past year.

Past Year Use of Cigarettes, Alcohol, and Any Illicit Drug Other Than Marijuana

Tables 4.8 to 4.10 present the combined models for cigarettes, alcohol, and any illicit drug other than marijuana, respectively. The models that account for more of the variation (in terms of logistic r-square, Cox and Snell, and Nagelkerke, respectively) are the ones for marijuana (.35, .61) and alcohol (.40, .55), with cigarettes (.32, .48) being next, and any illicit drug other than marijuana last (.21, .44). The lesser statistical strength of the risk and protective factors in predicting illicit drug use other than marijuana may be, to some extent, a statistical artifact, due to the much lower prevalence of use of these other drugs, such as heroin and cocaine, among adolescents. However, these results may indicate that the pathways leading to these rarer drugs are different from those for marijuana, cigarettes, and alcohol, and that the pathways may not be as well covered by the items included in the 1997 NHSDA.

Note that the demographic variables that are significant vary by substance, but reflect many of the historic patterns for those substances, even when controlling for all of the other variables present in the model. For example, for cigarettes, blacks were less likely to smoke than whites, and youths in the South were more likely to smoke cigarettes than those in the West. For alcohol, younger youths were less likely to have drunk alcohol in the past year than older youths; blacks less likely to drink than whites; and North Central youths more likely to drink than youths in the West.

Table 4.1 Results of Logistic Regression Models Predicting Past Year Marijuana Use with DEMOGRAPHICS and COMMUNITY Risk Factors: 1997

Factor	Model 1: Demographics			Model 2:Community Risk Factors			Model 3: Demographics + Community Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Intercept	-0.74	.11	.0001	-4.57	.19	.0001	-4.12	.24	.0001
Gender-Male vs. Female	.01	.11	.8934	-			.04	.15	.7769
Race
Other Race vs. White	-.14	.33	.6634	-			.19	.35	.5776
Hispanic vs. White	-.41	.13	.0026	-			-.35	.15	.0231
Black vs. White	-.30	.15	.0434	-			-.14	.18	.4281
Age-12-14 vs. 15-17 Years	-1.55	.11	.0001	-			-.51	.14	.0004
Region
Northeast vs. West	-.36	.20	.0727	-			.10	.22	.6626
North Central vs. West	-.18	.19	.3421	-			.17	.23	.4518
South vs. West	-.31	.13	.0143	-			-.18	.15	.2498
Pop. Density-Rural vs. Urban	-.47	.16	.0029	-			-.35	.17	.0410
Someone Offered/Tried to Sell Marijuana	-			2.57	.19	.0001	2.47	.16	.0001
Someone Offered/Tried to Sell Cocaine	-			.95	.15	.0001	.95	.15	.0001
Marijuana Easy to Get	-			1.55	.21	.0001	1.43	.21	.0001
Cocaine Easy to Get	-			-.18	.20	.3809	-.17	.21	.4003
Crack Easy to Get	-			-.01	.21	.9558	.06	.21	.7888
LSD Easy to Get	-			.32	.13	.0166	.24	.14	.0865
Heroin Easy to Get	-			-.48	.21	.0261	-.43	.21	.0441
-2Log-Likelihood (df)	6065.23 (9)			4043.15 (7)			3981.47(16)ab
R2 c	.07			.25			.26
RN2 d	.12			.43			.44

Note: Significance of risk/protective factors in Model 3 is marked in bold font; these factors along with demographics and the three exposure to prevention message items were included in the overall model.

^aIndicates x² comparison -2Log-Likelihood of Model 1 vs. Model 3 is significant.
^bIndicates x² comparison -2Log-Likelihood of Model 2 vs. Model 3 is significant.
^cCox and Snell R² is a measure of the fit of the model, defined as 1-[L(0))/L(B^)]^2/N, where L(0) is the likelihood of the intercept-only
model, L(B^) is the likelihood of the full model, and N is the estimated population size.
^dRecognizing that the Cox and Snell R² reaches a maximum for discrete models that depends on the value of the estimated
percentage, Nagelkerke (1991) proposed dividing the Cox and Snell measure by the maximum. In this sense, R_N² measures the absolute percentage of variation explained by the model.

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Table 4.2 Results of Logistic Regression Models Predicting Past Year Marijuana Use with DEMOGRAPHICS and FAMILY Risk Factors: 1997

Factor	Model 1:Demographics			Model 2: Family Risk Factors			Model 3:Demographics + Family Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Intercept	-0.75	.11	.0001	-3.16	.14	.0001	-2.31	.18	.0001
Gender-Male vs. Female	.01	.11	.8934	-			.01	.14	.9340
Race
Other Race vs. White	-.14	.33	.6634	-			.15	.27	.5630
Hispanic vs. White	-.41	.13	.0026	-			-.10	.15	.5108
Black vs. White	-.30	.15	.0434	-			-.12	.17	.4724
Age-12-14 vs. 15-17 Years	-1.55	.11	.0001	-			-1.08	.11	.0001
Region
Northeast vs. West	-.36	.20	.0727	-			-.44	.21	.0370
North Central vs. West	-.18	.19	.3421	-			-.16	.21	.4506
South vs. West	-.31	.13	.0143	-			-.38	.13	.0055
Pop. Density-Rural vs. Urban	-.47	.16	.0029	-			-.37	.15	.0132
Parent Not at All Strict About Dress	-			.70	.10	.0001	.69	.09	.0001
Parent Not at All Strict About Homework	-			.10	.12	.4264	.02	.12	.8660
Parent Not at All Strict About Curfew	-			-.23	.15	.1337	-.27	.16	.0908
Argued with Parents More Than Once a Week	-			.50	.14	.0005	.60	.14	.0001
See notes at end of table.

Table 4.2 (continued)

Factor	Model 1: Demographics			Model 2: Family RiskFactors			Model 3: Demographics + Family Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Parents Not at All or Somewhat Upset if Found Out Smoke 1+ Packs of Cigarettes Per Day	-			1.06	.13	.0001	.93	.13	.0001
Parents Not at All or Somewhat Upset if Found Out Tried Marijuana Once or Twice	-			1.38	.16	.0001	1.20	.17	.0001
Parents Not at All or Somewhat Upset if Found Out Had 5+ Drinks 1-2 Times a Week	-			.09	.16	.5817	.03	.16	.8434
Parents Not at All or Somewhat Upset if Found Out Smoked Marijuana Once a Month	-			1.27	.23	.0001	1.25	.23	.0001
Parents Not at All or Somewhat Upset if Found Out Smoked Marijuana 1-2 Times a Week	-			.03	.29	.9049	.05	.29	.8666
Parents Not at All or Somewhat Upset if Found Out Tried Inhalants Once or Twice	-			-.30	.19	.1043	-.29	.20	.1364
Parents Not at All or Somewhat Upset if Found Out Tried Heroin Once or Twice	-			-1.42	.42	.0009	-1.28	.40	.0016
Parents Not at All or Somewhat Upset if Found Out Used Cocaine Once a Month	-			-.84	.41	.0439	-.71	.38	.0648
Spoken with Parent/Other Adult About Drugs/Alcohol	-			.23	.12	.0620	.20	.12	.1111
-2Log-Likelihood (df)	6273.52 (9)			4849.16 (13)			4648.09 (22)ab
R2 c	.07			.17			.20
RN2 d	.12			.30			.34

Note: Significance of risk/protective factors in Model 3 is marked in bold font; these factors along with demographics and the three exposure to prevention message items were included in the overall model.

^aIndicates x² comparison -2Log-Likelihood of Model 1 vs. Model 3 is significant.
^bIndicates x² comparison -2Log-Likelihood of Model 2 vs. Model 3 is significant.
^cCox and Snell R² is a measure of the fit of the model, defined as 1-[L(0))/L(B^)]^2/N, where L(0) is the likelihood of the intercept-only model, L(B^) is the likelihood of the full model, and N is the estimated population size.
^dRecognizing that the Cox and Snell R² reaches a maximum for discrete models that depends on the value of the estimated percentage, Nagelkerke (1991) proposed dividing the Cox and Snell measure by the maximum. In this sense, R_N² measures the absolute percentage of variation explained by the model.

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Table 4.3 Results of Logistic Regression Models Predicting Past Year Marijuana Use with DEMOGRAPHICS and PEER/INDIVIDUAL Risk Factors: 1997

Factor	Model 1: Demographics			Model 2: Peer/Individual Risk Factors			Model 3: Demographics + Peer/Individual Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Intercept	-0.75	.11	.0001	-6.32	.29	.0001	-5.32	.33	.0001
Gender-Male	.01	.11	.8934	-			-.38	.14	.0056
Race
Other Race vs. White	-.14	.33	.6634	-			.54	.38	.1565
Hispanic vs. White	-.41	.13	.0026	-			-.19	.20	.3478
Black vs. White	-.30	.15	.0434	-			-.48	.22	.0273
Age-12-14 vs. 15-17 Years	-1.55	.11	.0001	-			-1.05	.15	.0001
Region
Northeast vs. West	-.36	.20	.0727	-			-.32	.25	.1933
North Central vs. West	-.18	.19	.3421	-			-.20	.22	.3744
South vs. West	-.31	.13	.0143	-			-.17	.18	.3411
Pop. Density-Rural vs. Urban	-.47	.16	.0029	-			-.29	.18	.1061
At Least a Few Close Friends Smoked 1+ Packs of Cigarettes Per Day	-			.44	.17	.0097	.34	.16	.0313
At Least a Few Close Friends Had 5+ Drinks 1-2 Times a Week	-			-.08	.14	.5417	-.29	.14	.0465
At Least a Few Close Friends Tried or Used Marijuana	-			1.93	.27	.0001	1.85	.28	.0001
At Least One Close Friend Tried Inhalants or Heroin Once or Twice Or Used Cocaine Once a Month	-			.52	.13	.0001	.49	.14	.0005

See notes at end of table.

Table 4.3 (continued)

Factor	Model 1: Demographics			Model 2: Peer/Individual Risk Factors			Model 3:Demographics + Peer/Individual Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Gang Fight in Past Year				.69	.23	.0030	.90	.23	.0002
Shoplifted in Past Year	-			1.12	.14	.0001	1.22	.16	.0001
Friends Not at All Upset if Found Out Smoked 1+ Packs of Cigarettes Per Day	-			.18	.18	.3048	.20	.18	.2473
Friends Not at All Upset if Found Out Tried Marijuana Once or Twice	-			1.34	.24	.0001	1.15	.25	.0001
Friends Not at All Upset if Found Out Had 5+ Drinks 1-2 Times a Week	-			-.09	.19	.6484	-.05	.19	.8029
Friends Not at All Upset if Found Out Smoked Marijuana Once a Month	-			.63	.24	.0104	.60	.23	.0119
Friends Not at All Upset if Found Out Smoked Marijuana 1-2 Times a Week	-			.56	.24	.0199	.79	.24	.0014
Friends Not at All Upset if Found Out Tried Inhalants Once or Twice	-			-.30	.21	.1564	-.30	.23	.1850
Friends Not at All Upset if Found Out Tried Heroin Once or Twice	-			-.65	.32	.0436	-.54	.31	.0863
Friends Not at All Upset if Found Out Used Cocaine Once a Month	-			-.30	.36	.4026	-.30	.35	.3857

See notes at end of table.

Table 4.3 (continued)

Factor	Model 1: Demographics	Model 2: Peer/Individual Risk Factors			Model 3:Demographics + Peer/Individual Risk Factors
	ß	SE	p value	ß	SE	p value	ß
Perceptions of No Risk to Moderate Risk for Smoking Marijuana Once a Month	-	1.35	.23	.0001	1.30	.24	.0001
Perceptions of No Risk to Moderate Risk for Smoking 1+ Packs of Cigarettes Per Day	-	.14	.12	.2574	.16	.13	.2080
Perceptions of No Risk to Moderate Risk for Having 5+ Drinks Once or Twice a Week	-	.22	.17	.1948	.31	.16	.0529
Perceptions of No Risk to Moderate Risk for Using Cocaine Once a Month	-	-.34	.12	.0072	-.32	.13	.0161
-2Log-Likelihood (df)	6273.52 (9)	3561.08 (18)			3404.37 (27) a, b
R2 c	.07	0.30			0.32
RN2 d	.12	0.53			0.55

Note: Significance of risk/protective factors in Model 3 is marked in bold font; these factors along with demographics and the three exposure to prevention message items were included in the overall model.

^aIndicates x² comparison -2Log-Likelihood of Model 1 vs. Model 3 is significant.
^bIndicates x² comparison -2Log-Likelihood of Model 2 vs. Model 3 is significant.
^cCox and Snell R² is a measure of the fit of the model, defined as 1-[L(0))/L(B^)]^2/N, where L(0) is the likelihood of the intercept-only model, L(B^) is the likelihood of the full model, and N is the estimated population size.
^dRecognizing that the Cox and Snell R² reaches a maximum for discrete models that depends on the value of the estimated percentage, Nagelkerke (1991) proposed dividing the Cox and Snell measure by the maximum. In this sense, R_N² measures the absolute percentage of variation explained by the model.

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Table 4.4 Results of Logistic Regression Models Predicting Past Year Marijuana Use with DEMOGRAPHICS and SCHOOL Risk Factors: 1997

Factor	Model 1: Demographics			Model 2: School RiskFactors			Model 3: Demographics + School Risk Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Intercept	-0.75	.11	.0001	-.75	.32	.0195	.19	.32	.5612
Gender-Male vs. Female	.01	.11	.8934	-			-.22	.12	.0780
Race
Other Race vs. White	-.14	.33	.6634	-			-.07	.33	.8197
Hispanic vs. White	-.41	.13	.0026	-			-.62	.15	.0001
Black vs. White	-.30	.15	.0434	-			-.50	.16	.0030
Age-12-14 vs. 15-17 Years	-1.55	.11	.0001	-			-1.42	.12	.0001
Region
Northeast vs. West	-.36	.20	.0727	-			-.44	.17	.0135
North Central vs. West	-.18	.19	.3421	-			-.21	.21	.3176
South vs. West	-.31	.13	.0143	-			-.41	.13	.0024
Pop. Density-Rural vs. Urban	-.47	.16	.0029	-			-.62	.16	.0325
Enrolled in School	-			- 1.17	.30	.0001	-1.10	.29	.0003
Last Semester Grades Other Than Mostly A's and B's	-			.87	.10	.0001	1.00	.11	.0001
In-School Alcohol/Drug Education Class in Past Year	-			-.53	.12	.0001	-.37	.12	.0018
-2Log-Likelihood (df)	6273.52 (9)			5392.06 (3)			4945.13 (12)ab
R2 c	.07			.04			.10
RN2 d	.12			.07			.18

Note: Significance of risk/protective factors in Model 3 is marked in bold font; these factors along with demographics and the two other exposure to prevention message items were included in the overall model.

^aIndicates x² comparison -2Log-Likelihood of Model 1 vs. Model 3 is significant.
^bIndicates x² comparison -2Log-Likelihood of Model 2 vs. Model 3 is significant.
^cCox and Snell R² is a measure of the fit of the model, defined as 1-[L(0))/L(B^)]^2/N, where L(0) is the likelihood of the intercept-only model, L(B^) is the likelihood of the full model, and N is the estimated population size.
^dRecognizing that the Cox and Snell R² reaches a maximum for discrete models that depends on the value of the estimated percentage, Nagelkerke (1991) proposed dividing the Cox and Snell measure by the maximum. In this sense, R_N² measures the absolute percentage of variation explained by the model.

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.

Table 4.5 Results of Logistic Regression Models Predicting Past Year Marijuana Use with DEMOGRAPHICS and GENERAL Risk/Protective Factors: 1997

Factor	Model 1: Demographics			Model 2: General Risk/Protective Factors			Model 3:Demographics + General Risk/Protective Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Intercept	-0.75	.11	.0001	-2.35	.23	.0001	-1.22	.33	.0004
Gender-Male vs. Female	.01	.11	.8934	-			-.13	.14	.3613
Race
Other Race vs. White	-.14	.33	.6634	-			.07	.39	.8523
Hispanic vs. White	-.41	.13	.0026	-			-.33	.16	.0354
Black vs. White	-.30	.15	.0434	-			-.28	.16	.0833
Age-12-14 vs. 15-17 Years	-1.55	.11	.0001	-			-1.30	.13	.0001
Region
Northeast vs. West	-.36	.20	.0727	-			-.34	.21	.1067
North Central vs. West	-.18	.19	.3421	-			-.10	.20	.6095
South vs. West	-.31	.13	.0143	-			-.16	.13	.2325
Pop. Density-Rural vs.  Urban	-.47	.16	.0029	-			-.58	.16	.0006
Would Talk to Parent(s) About Serious Problems	-			-.72	.13	.0001	-.74	.12	.0001
Would Talk to Some Other Relative(s) about Serious Problems	-			-.05	.12	.6453	.02	.12	.8611
Would Talk to Friend(s) About Serious Problems	-			.66	.18	.0003	.41	.19	.0302
Would Talk to Some Other Person About Serious Problems	-			-.27	.11	.0159	-.20	.11	.0700
Most Likely Talk to Parent(s) About Serious Problems	-			-.47	.15	.0023	-.41	.15	.0060
Most Likely Talk to Friend(s) About Serious Problems	-			.31	.14	.0300	.25	.14	.0735

See notes at end of table.

Table 4.5 (continued)

Factor	Model 1: Demographics			Model 2: General Risk/Protective Factors			Model 3: Demographics + General Risk/Protective Factors
	ß	SE	p value	ß	SE	p value	ß	SE	p value
Extracurricular Activity in Past Year	-			-.08	.19	.6656	-.09	.19	.6485
Sports/Physical Activities	-			-.14	.14	.3369	-.05	.14	.7375
Church-Related Activities	-			.32	.21	.1370	.15	.22	.5078
Music/Art/Performing Arts	-			-.76	.16	.0001	-.54	.17	.0017
Club/Youth Group	-			-.19	.14	.1716	-.34	.15	.0282
Student Govt./ROTC/Other Civic Activities	-			-.16	.21	.4565	-.19	.21	.3610
Attended Religious Services Less Than Once a Week in Past Year	-			.80	.12	.0001	.80	.12	.0001
My Religious Beliefs Are Not Very Important	-			-.01	.18	.9574	-.12	.19	.5325
My Religious Beliefs Do Not Influence My Decisions	-			.25	.16	.1089	.37	.17	.0289
It is Not Important for My Friends to Share My Religious Beliefs	-			.44	.17	.0138	.34	.16	.0360
Seen/Heard Alcohol/Drug Prevention Messages Outside of School in Past Year	-			.36	.17	.0353	.27	.17	.1096
-2Log-Likelihood (df)	6273.52 (9)			4708.70 (17)			4425.51 (26)ab
R2 c	.07			.11			.15
RN2 d	.12			.19			.25

Note: Significance of risk/protective factors in Model 3 is marked in bold font; these factors along with demographics and the three exposure to prevention message items were included in the overall model.

^aIndicates x² comparison -2Log-Likelihood of Model 1 vs. Model 3 is significant.
^bIndicates x² comparison -2Log-Likelihood of Model 2 vs. Model 3 is significant.
^cCox and Snell R² is a measure of the fit of the model, defined as 1-[L(0))/L(B^)]^2/N, where L(0) is the likelihood of the intercept-only model, L(B^) is the likelihood of the full model, and N is the estimated population size.
^dRecognizing that the Cox and Snell R² reaches a maximum for discrete models that depends on the value of the estimated percentage, Nagelkerke (1991) proposed dividing the Cox and Snell measure by the maximum. In this sense, R_N² measures the absolute percentage of variation explained by the model.

Source: Office of Applied Studies, SAMHSA, 1997 National Household Survey on Drug Abuse.
Table 4.6 Estimated Odds Ratios and 95% Confidence Intervals for Significant Associations of Risk and Protective Factors with Past Year MARIJUANA USE, Controlling for Demographics in COMBINED REDUCED MODEL: 1997

Factors	ß	Odds Ratio	Lower 95% Limit	Upper 95% Limit
Intercept	-7.68	-	-	-
Demographics
Male vs. Female	-.48	.62	.43	.90
Other Race vs. Whitea	.86	2.37	1.20	4.67
Hispanic vs. White	-.31	.73	.47	1.14
Black vs. White	-.33	.72	.45	1.15
Age-12-14 vs. 15-17 Years Old	-.55	.58	.38	.87
Factors b
Community
Anyone Offered/Tried to Sell Marijuana	1.84	6.31	4.00	9.95
Marijuana Easy to Get	.85	2.34	1.48	3.69
Family
Parents Would Be Not at All or Somewhat Upset If Found Out Smoked 1+ Packs of Cigarettes Per Day (vs. Very Upset)	.39	1.48	1.11	1.97
Parents Would Be Not at All or Somewhat Upset If Found Out Tried Heroin Once or Twice (vs. Very Upset)*	-1.09	.34	.15	.74
Spoken with Parent/Other Adult About Drugs/Alcohol	.39	1.48	1.07	2.06
Peer/Individual
At Least a Few Close Friends Have 5+ Drinks Once or Twice a Week*	-.49	.62	.45	.84
At Least a Few Close Friends Tried or Used Marijuana	1.33	3.80	1.99	7.26
Friends Would Be Not at All Upset If Found Out Smoked Marijuana Once a Month (vs. Somewhat or Very Upset)	.91	2.48	1.25	4.92
Perceptions of No Risk to Moderate Risk for Smoking Marijuana Once a Month	1.45	4.26	2.42	7.50
Gang Fight in Past Year	.66	1.93	1.15	3.24
Shoplifted in Past Year	1.18	3.24	2.34	4.49