Measurement issues in race ethnicity and substance use

ABSTRACT

Objective

To investigate substance use patterns by race and ethnicity with more detail about racial and ethnic background. There are many studies of substance use and race-ethnicity, however few studies have examined a more diversified delineation of race and ethnicity. By looking more closely at racial and ethnic groups who are multiracial and multiethnic, we may gain more information about patterns of substance use among diverse groups of people in the United States.

Methods

We use national data sample from the United States and delineate racial and ethnic groups.

Results

Findings indicate that there are substantive differences in patterns of substance use among the race, ethnic and multiracial/ethnic populations in the United States. These patterns vary by substance used and strongly suggest that we need a careful assessment of the measurement used to classify individuals as racial and ethnic people in our studies.

Conclusions

The measurement of racial and ethnic groups should be more fully assessed since the patterns of substance vary considerably within traditionally measured racial and ethnic groups.

KEYWORDS:

• Measurement of race and ethnicity
• race and substance use
• ethnicity and substance use
• multiracial Americans and substance use
• multiracial Americans and health
• substance use measurement

Introduction

In this paper, we investigate multiracial and multiethnic differences in substance use. There are many studies of substance use and race-ethnicity; however, the studies fail to examine a more diverse delineation of race and ethnicity. This measurement issue may affect our ability to fully understand the patterns that exist in the study of race, ethnicity, and substance use. What we do know from previous research is that there are differences in the lifetime and current prevalence of substance use among people of different racial and ethnic groups (Kim, 2021; Substance Abuse and Mental Health Services Administration [SAMHSA], 2020; Terry McElrath & Patrick, 2020). For example, Asian Americans have very low rates of substance abuse over their lifetime and within the past year (SAMHSA, 2020), while Native Americans have the highest rates of use for marijuana and other illicit substances (SAMHSA, 2020). White adolescents have the highest rates of alcohol use overall, though Native American adolescents have the highest rates of heavy alcohol use and of binge alcohol use (SAMHSA, 2020). Studies have also found differences in substance use for multiracial and multiethnic people (Chavez & Sanchez, 2010; Choi et al., 2006; Clark et al., 2013; Fisher et al., 2017). Most of these studies reported that multiracial and multiethnic individuals have higher rates of substance use than monoracial and monoethnic individuals. However, there are two issues of importance to consider here.

First, there is the limitation in the research concerning the measurement for being multiracial and multiethnic. Most studies simply ask people if they are multiracial, with a yes-no question. This seems a reasonable procedure, but it lacks the detail of knowing which specific groups people self-identify with, and research has shown this is of great importance (C. Broman et al., 2008; Manly, 2006; Shih & Sanchez, 2009). A second limitation in this measurement is that the question usually also results in everyone who answers “Yes”’ to the question “Are you multiracial?” being grouped into one category and compared with all the people who answer “No” to the question. This lack of precise detail may be highly consequential and not allow us to see patterns of substance use that may vary within specific multiracial and/or multiethnic groups.

Given the paucity of research that measures this kind of detail in race and ethnicity, there is a need to examine the details of which specific groups people self-identify with and how that might affect the use of substances. Some research has suggested that a greater delineation of racial and ethnic groups reveals different patterns of use than what is seen in our usual more limited measurement of race and ethnicity (C. Broman et al., 2008; Manly, 2006; Shih & Sanchez, 2009). In this study, we investigate how mono and multiracial and ethnic identity matter in patterns of substance use.

Data and methods

The data used in this study are from the US National Longitudinal Study of Adolescent to Adult Health (commonly referred to as Add Health), a nationally representative study of US individuals who participated in five waves of interviews, starting in adolescence in 1994–1995 (Wave 1) and ending with the most recent wave (data collection in 2016–2019), where respondents were aged 33–44 years (Wave 5) (Blum et al., 2000; Liao et al., 2019). A multistage, stratified, school-based cluster sampling design was used to collect data in schools in 1994–1995, which later determined who would be chosen to participate in in-home interviews. Approximately 120,000 adolescents were eligible for a school interview, and more than 90,000 completed the in-school questionnaire. Among those eligible for the school interview, some were selected for in-home interviews. The Wave 1 in-home interviews lasted between one and two hours and were conducted in person using a laptop computer and audio-enhanced, computer-assisted self-interviewing (audio-CASI) for sensitive questions. The topics covered in the Wave 1 in-home interviews included health status, nutrition, peer networks, decision-making processes, family composition and dynamics, educational aspirations and expectations, employment experience, involvement in sexual activity, substance use, and criminal activity.

The Wave 1 data consisted of responses from 20,745 adolescents and were collected during 1994–1995 when the respondents were in grades 7–12. Wave 2 surveyed the same sample of adolescents (excluding the high school seniors from Wave 1) a year later in 1996 via in-home interviews and included 14,736 respondents. The questions in the Wave 2 interview were almost identical to those asked in Wave 1. The Wave 2 data were excluded from our study, as high school seniors were not interviewed, and the data was not missing at random. All the respondents from Wave 1 were again sought for interviews in Waves 3 to 5. The Wave 3 data were collected in 2001–2002 from 15,197 respondents; the Wave 4 data were collected in 2007–2008 from 15,701 respondents; and the Wave 5 data were collected in 2016–2019 from 12,300 respondents. The current study uses the full dataset from waves 1, 3, 4, and 5. We excluded the missing values of our dependent variables. We used STATA 16 (StataCorp., 2019) to conduct the analyses and used cluster and weight variables to account for non-independence of observations, unequal probability of selection, and complex survey design. Thus, data accurately represent the US population of adolescents at Wave 1.

Measures

The measure of race-ethnicity is taken from Wave 3 of the Add Health data. This measure was constructed from questions in the Add Health dataset concerning race and ethnicity. One of the first questions asked was about participants’ Hispanic or Latino origin. Like the US census, the question asked, “Are you of Hispanic or Latino origin?” A second question then asked, “What is your race? You may give more than one answer.” Following this were other questions that tried to ascertain specific ethnic groups for those who identified as Latino or Asian. From these questions, we created a 17-category self-identification variable. Respondents in the first seven categories self-identified in single racial categories, while the others were categorized as multiracial or multiethnic, based on respondents choosing more than one racial or ethnic group.

Thus, the categories were 1) White; 2) Black; 3) Native American (AIAN); 4) Chinese American; 5) Filipino American; 6) East Asian (this consists of respondents identifying as Japanese, Korean, or Vietnamese and is constructed in this manner due to the small number of people who identified as such); 7) Hispanic or Latino; 8) Black and White; 9) White and Native American; 10) multiracial Native American; 11) multiracial Asian; 12) other Asian; 13) White Latino; 14) Black Latino; 15) Native American and Latino; 16) other Latino; and 17) another race and ethnicity, not elsewhere classified (NEC) (respondents listed in this category are those in any “other” group, based on the small sample size of people in the group, or such people choosing the category of “other”).

Again, respondents in the first seven categories chose only one racial group. Anyone choosing more than one (including respondents choosing “Latino” for the ethnicity question) was then coded as multiracial or multiethnic. People who chose “Asian” during the interview were then directed to a subsequent question which asked, “What is your Asian background? You may give more than one answer.” Any respondent who answered these questions was categorized as primarily Asian, and therefore someone who may have chosen more than one race or ethnicity would be categorized as “multiracial Asian.” Again, anyone who only chose “Asian” was categorized as one of the Asian groups above from 4) Chinese American; 5) Filipino American; or 6) East Asian.

Outcome measures: substance use

The dependent variables of substance use are measured from the Wave 5 data. We used several measures, three pertaining to alcohol use, and one each for prescription drug misuse, tobacco use, marijuana use, and the use of other illegal drugs, apart from marijuana. Our dependent variables are as follows. The first measure of alcohol use was that of the number of days of drinking in the past 12 months. The question asked, “During the past 12 months, on how many days did you drink alcohol (beer, wine, or liquor)?” The measure was coded as 0 = “none in the past 12 months” through to 6 = “every day or almost every day.” The second measure of alcohol use was that of the number of days of drinking in the past 30 days. The question asked, “During the past 30 days, on how many days did you drink alcohol (beer, wine, or liquor)?” The measure was coded as 0 = “none in the past 30 days” through to 6 = “every day or almost every day.” Heavy episodic alcohol use is measured using a question which asked, “During the past 12 months, on how many days did you drink [5 or 4, for men and women, respectively] drinks in a row?” The measure was coded as 0 = “none in the past 12 months” through to 6 = “every day or almost every day.”

The marijuana use measure asked, “During the past 30 days, on how many days did you use marijuana?” The measure was coded as 0 = “never in the in past 30 days” through to 6 = “every day or almost every day.” We recoded the measure due to the small numbers of cases at the higher end into 4 = “2 or more days a week.” The “smoking” measure was based on responses to three questions which asked, “During the past 30 days, on how many days did you a) smoke cigarettes; b) smoke a cigar or pipe, use chewing tobacco or snuff; c) use an e-cigarette?” Respondents who answered “From 1 to 30 days” were coded as 1, while those who answered “0 days” were coded as 0.

Prescription drug misuse was measured the same way as in Wave 4: that is, from a question asking, “In the past 30 days, which of the following types of prescription drugs have you taken that were not prescribed for you, taken prescription drugs in larger amounts than prescribed, more often than prescribed, for longer periods than prescribed, or taken prescription drugs that you took only for the feeling or experience they caused?” The specific drugs asked about are 1) sedatives or downers such as sleeping pills, barbiturates, Seconal; 2) tranquilizers, such as Librium, Valium, or Xanax; 3) stimulants or uppers, such as amphetamines, prescription diet pills, Ritalin, Preludin, or speed; and 4) painkillers or opioids, such as Vicodin, OxyContin, Percocet, Demerol, Percodan, or Tylenol with codeine. As with the measure used in Wave 4, responses were dichotomous, with 0 = “No” and 1 = “Yes.” Lastly, the measure of other illegal drugs was based on questions which asked, “In the past 30 days, have you used any of the following drugs?” The follow-up questions asked about crystal meth, cocaine, heroin, and others, such as LSD, PCP, ecstasy, or mushrooms or inhalants. We recoded the measures into one indicator that ranged from 0 to 1, where 0 = never and 1 = any use of illegal drugs other than marijuana.

Control measures

The control variables of prior substance use and others were measured using data from Waves 1 and 4. Alcohol and marijuana use from Wave 1 were measured based on questions which asked about both the quantity and frequency of use. The measures were coded into a single measure of the quantity and frequency of use. The alcohol use measure was on a four-point scale ranging from 0 (never) to 3 (roughly corresponds to drinking at least once a week and drinking at least four drinks each time). The alcohol use measure was coded based on measures used in prior studies (C. L. Broman et al., 2020). Marijuana use was similarly coded but is truncated at the value of 2 (corresponding to about once monthly and moderate use), owing to the infrequency of responses. From Wave 4, we used a measure of marijuana use and prescription drug misuse. The marijuana use measure asked, “How many days in the past year was marijuana used?” As most people said “None,” we coded the measure as a dichotomy, where 1= any marijuana use in the past year, and 0 = no use in the past year.

The measure of prescription drug use in Wave 4 is based on a question that asked, “Have you ever taken any prescription drugs that were not prescribed for you, taken prescription drugs in larger amounts than prescribed, more often than prescribed, for longer periods than prescribed, or taken prescription drugs that you took only for the feeling or experience they caused?” The specific drugs asked about were 1) sedatives or downers such as sleeping pills, barbiturates, Seconal; 2) tranquilizers, such as Librium, Valium, or Xanax; 3) stimulants or uppers, such as amphetamines, prescription diet pills, Ritalin, Preludin, or speed; and 4) painkillers or opioids, such as Vicodin, OxyContin, Percocet, Demerol, Percodan, or Tylenol with codeine. As with the use of marijuana in Wave 4, responses were dichotomous, with 0 = “No” and 1 = “yes.”

The control variables of depression, availability, and religious attendance are described next. Depression is measured from a Wave 4 measure and is based on a question that asked respondents if they had ever been diagnosed with depression by a doctor. If “Yes,” the response was coded as a “1;” otherwise, it was coded as a “0.” Availability of alcohol in the home (coded from Wave 1) was measured, as in prior research (C. L. Broman, 2016), from the respondent using a question that asked, “Is alcohol easily available to you in your home?” The measure was dummy coded in the original data where 1 = “Yes” (easily available) and 0 = “No.” Attendance at religious services was measured from Wave 4, from a question that asked, “How often have you attended church, synagogue, temple, mosque, or religious services in the past 12 months?” The measure was coded on a scale ranging from never (0) to once a week or more (5).

Other control measures used are sociodemographic measures of age, biological sex, and education completion, which were self-reported by respondents. The age measure was from Wave 5 and ranged from 33 to 44 years. Sex was a dummy coded variable, with male = 1. This measure was used from Wave 1. Education was used from Wave 4 of the data, when respondents’ average age was 29 years. Respondents’ education was measured using five categories: 1) 8^th grade or less and some high school; 2) high school graduate; 3) some vocational/technical training (after high school) and/or some college; 4) completed college (bachelor’s degree); and 5) graduate school, completed a master’s degree, some graduate training beyond a master’s degree, completed a doctoral degree, some post baccalaureate professional education (e.g., law school, medical school, nursing), and completed post baccalaureate professional education. We used Wave 4 education because it occurs at a time in the lives of many respondents when the bulk of formal education has been completed (average age of 29 years).

Results

Descriptive data are shown in Table 1. At Wave 5, the sample averaged about 38 years old, 43% male, with an average educational attainment within the “some post high school” training range. The next data are presented for the racial and ethnic variables. The numbers of observations presented are for the overall variable at Wave 5. For self-identified race-ethnicity, most of the sample self-identified as non-Hispanic White (at 56%), 19% of the sample self-identified as non-Hispanic Black, and approximately 10% of the sample self-identified as White Hispanics. These were the largest three groups of respondents. None of the other self-identified people comprised more than about 2% of the sample. Native-Hispanics, Filipino Americans, Other Asian Americans, and White-Natives each comprised about 2% of the sample. When rounded, the other racial-ethnic groups comprised about 1% of the sample.

Table 1. Descriptive data for the sample at wave 5.

	Mean	Range	N
Age w5	38.0	33–44	12,271
Sex (1 = male)	.43	0–1	12,273
Education W4	3.20	1–5	10,888
Self-identified White W3	.56	0–1	10,187
Self-identified Black W3	.19	0–1	10,187
Self-identified Black-White W3	.01	0–1	10,187
Self-identified Hispanic W3	.01	0–1	10,187
Self-identified White-Hispanic W3	.10	0–1	10,187
Self-identified Black-Hispanic W3	.01	0–1	10,187
Self-identified Native-Hispanic W3	.02	0–1	10,187
Self-identified Other-Hispanic W3	.01	0–1	10,187
Self-identified Chinese American W3	.01	0–1	10,187
Self-identified Filipino American W3	.02	0–1	10,187
Self-identified East Asian American W3	.01	0–1	10,187
Self-identified Multiracial Asian American W3	.01	0–1	10,187
Self-identified Other Asian American W3	.02	0–1	10,187
Self-identified AIAN W3	.01	0–1	10,187
Self-identified White AIAN W3	.02	0–1	10,187
Self-identified Multiracial AIAN W3	.01	0–1	10,187
Other, N.E.C.	.01	0–1	10,187
Alcohol use W1	1.25	0–4	12,093
Marijuana use W1	.39	0–3	11,909
Any marijuana use W4	.21	0–1	10,885
Prescription drug misuse W4	.16	0–1	10,846
Depression W4	.16	0–1	10,890
Alcohol easily available W1	.31	0–1	12,199
Religious attend W4	1.65	0–5	10,887
Alcohol use days per year	2.87	0–6	11,382
Alcohol days in last month	2.61	0–6	10,402
Heavy episodic alcohol use in last year	1.23	0–6	10,419
Marijuana use in past month	.50	0–4	12,283
Tobacco use in past month	.31	0–1	12,249
Prescription drug misuse in past month	.11	0–1	12,258
Illegal drug use in past month	.03	0–1	12,255

Totals for race-ethnicity add up to more than 100% due to rounding.

Next, Table 1 presents data on the other variables used in our analyses. Many teens were not alcohol or marijuana users, as indicated by the low values for prior drug use in Wave 1. Use of alcohol and marijuana was higher in Wave 4, when the average age was 29 years. In Wave 4, 21% used marijuana in the previous year, and 16% misused prescription drugs. The value for depression shows that 16% had been told by a health professional that they (the respondents) suffer from depression. About 31% of adolescents reported in Wave 1 that alcohol was easily available to them, while there was some level of religious attendance in Wave 4. Values for the dependent variables are presented next. The value for alcohol days per year indicates that respondents had drunk alcohol almost 2–3 days per month over the past year. The value for alcohol days per month indicates that respondents had drunk alcohol between 2–3 days per month to one day per week over the past 30 days. Few people were found to drink in a heavy episodic manner. The meaning corresponds to a value indicating 1–2 days of heavy episodic drinking in the past year. Marijuana use is similarly low. The value indicates less than one day of marijuana use in the past 30 days (the percentages which indicate that 70% say they “never” used marijuana in the past 30 days is not shown). It was found that 30% of respondents had used tobacco in the past month, 11% misused prescription drugs in the past month, and a very small percentage (3%) had used illegal drugs in the past month.

Multivariate analyses

Table 2 presents results of OLS regression analyses of Wave 5 substance use for the continuous variables of alcohol and marijuana use. Other variables in the equation are self-identified race/ethnicity, age, biological sex, education at Wave 4, prior substance use, depression, availability of alcohol in the home as a teen, and religious attendance in Wave 4. Looking at the results in Table 2, we can see that our prediction is validated. Namely, there are significant differences in substance use for self-identified race/ethnicity. Both monoracial-ethnic and multiracial-ethnic populations differ. Most results show that the monoracial and multiracial-ethnic population uses the substances examined in Table 2 less than the white non-Hispanic population (the reference group) does. The self-identified black population is less likely to engage in heavy episodic alcohol use but more likely to use marijuana. The pattern for marijuana use is the same for the self-identified black-Hispanic population, as well as the native-Hispanic population; these cohorts are more likely to use marijuana than are white-non-Hispanics, the reference group. The major finding of this table is that there are many differences among the multiracial-multiethnic population. These results make clear that we cannot simply group the people in this population together. Rather, we need to account for the complexities of racial-ethnic identity that make up who people are.

Table 2. Regression of substance use in wave 5 on race-Ethnicity.

Variable	Alcohol Last Year	Alcohol Last Month	Heavy Episodic Use	Marijuana Use
Self-identified Black W3	.00	−.06	−.18***	.24***
Self-identified Black-White W3	−.33	−.22	−.43	.09
Self-identified Hispanic W3	−.37*	−.49**	−.08	.07
Self-identified White-Hispanic W3	−.16**	−.31***	−.18**	.02
Self-identified Black-Hispanic W3	−.02	−.14	−.08	.29*
Self-identified Native-Hispanic W3	−.24	−.59***	−.16	.21*
Self-identified Other-Hispanic W3	−.04	−.08	−.23	−.12
Self-identified Chinese American W3	−.58***	−.63***	−.57***	−.07
Self-identified Filipino American W3	−.42***	−.61***	−.25*	−.04
Self-identified East Asian American W3	−.12	.02	.25	.02
Self-identified Multirace Asian American W3	−.24	−.40	−.43*	−.13
Self-identified Other Asian American W3	−.41**	−.51**	−.08	−.04
Self-identified AIAN W3	.23	−.07	.25	.07
Self-identified White AIAN W3	−.28*	−.36*	.02	−.15
Self-identified Multirace AIAN W3	.35	.20	.06	.11
Self-identified Other, NEC W3	−.21	−.34	−.41	−.04
Age W5	−.04***	−.02*	−.03***	−.01
Sex (1 = male)	.46***	.57***	.46***	.10***
Education W4	.25***	.22***	−.06***	−.09***
Alcohol use W1	.11***	.09***	.12***	.01
Marijuana use W1	.03	.07*	.09**	.11***
Any marijuana use W4	.43***	.45***	.40***	1.26***
Prescription drug misuse W4	.04	.11*	.25***	.19***
Depression W4	−.23***	−.27***	−.06	.06*
Alcohol easily available W1	.24***	.26***	.12**	.01
Religious attend W4	−.12***	−.09***	−.09***	−.03***
Constant	3.21	2.52	2.40	.79
R2	.11	.10	.11	.27
N =	8,348	7,653	7,668	8,964

As examples of this point, note in Table 2 the significant differences for the White, Black, and Native Hispanic groups. This is an indication that delineating groups by race and ethnicity does matter in the patterns of substance use that we observe. Patterns differing across the Hispanic categories vary depending upon the respondent’s self-identified race. We also see that there are the same kinds of differences for the Asian population. Delineating the population reveals differences in the patterns of substance use among the Asian-American population.

The other variables in the table are important in ways that are consistent with prior research. In this sample, we see that the closer to the age of 44 years one is, the less likely one is to use alcohol. Men are more likely to be users than are women. The more highly educated are more likely to have used alcohol in the last year and last month but are less likely to engage in heavy episodic use or marijuana use. Prior use has an impact in that past use is associated with greater current use of substances. Depression is associated with less use of alcohol, as is religious attendance. However, depression is associated with more use of marijuana. The easy availability of alcohol in adolescence is associated with more current use, including heavy episodic use.

Table 3 presents the results of logistic regression analyses of Wave 5 substance use for the binary variables of tobacco use, prescription drug misuse and illegal drug use. Variables in the equation are the same as those in Table 2: self-identified race-ethnicity, age, sex, education at Wave 4, prior substance use, depression, availability of alcohol in the home as a teen, and religious attendance in Wave 4. The major result of Table 3 is similar to that of Table 2; namely, there are many differences among the multiracial-multiethnic population. These results again make clear that we cannot simply group the people in this population together. Looking at the results, we can see that self-identified Hispanics, white Hispanics, Native Hispanics, Chinese Americans, and Other Asian Americans are significantly less likely than the reference group of white non-Hispanics to engage in tobacco use. Self-identified white AIAN are more likely to engage in tobacco use behaviors. There are also significant differences for prescription drug misuse. Blacks, Native Hispanics, and Filipino Americans are more likely to engage in prescription drug misuse than the reference group of white non-Hispanics. Results for illegal drug use show that Native and other Hispanics, Filipino Americans, East and other Asian Americans are more likely to use than the reference group of white non-Hispanics.

Table 3. Logistic Regression of Substance Use in Wave 5 on Race-Ethnicity.

Variable	Tobacco use	Prescription misuse	Illegal drug use
Self-identified Black W3	.94	1.23*	1.08
Self-identified Black-White W3	.99	.63	.81
Self-identified Hispanic W3	.24***	.76	1.09
Self-identified White-Hispanic W3	.45***	1.02	1.42
Self-identified Black-Hispanic W3	.54	1.31	1.00
Self-identified Native-Hispanic W3	.37***	1.65*	2.19*
Self-identified Other-Hispanic W3	.56	1.78	3.27*
Self-identified Chinese American W3	.28**	68	1.57
Self-identified Filipino American W3	.70	1.61*	3.64***
Self-identified East Asian American W3	1.28	.45	3.38*
Self-identified Multirace Asian American W3	.55	.41	1.66
Self-identified Other Asian American W3	.52**	1.55	2.17*
Self-identified AIAN W3	.78	1.21	2.90
Self-identified White AIAN W3	1.50*	.69	1.08
Self-identified Multirace AIAN W3	.72	.85	.76
Self-identified Other, nec W3	.65	1.39	1.91
Age W5	.96*	1.01	.90**
Sex (1 = male)	1.87***	.86*	1.67***
Education W4	.57***	.78***	.77***
Alcohol use W1	1.22***	1.07*	1.02
Marijuana use W1	1.22***	1.10*	1.23*
Any marijuana use W4	2.24***	1.29**	2.87***
Prescription drug misuse W4	1.49***	2.31***	2.06***
Depression W4	1.38***	1.55***	1.23
Alcohol easily available W1	.83**	1.01	1.25
Religious attend W4	.90***	.99	.88**
Constant	5.44	.11	1.44
−2 log likelihood	−4566.9	−2964.9	−1157.4
N =	8,946	8,954	8,951

The other variables in the table show similar results as in Table 2. Again, we see that the closer to the age of 44 years one is, the less likely one is to use of tobacco and illegal drugs. Men are more likely to use tobacco products and illegal drugs than are women. However, women engage in more prescription drug misuse. The more highly educated are less likely to have used tobacco products and misused prescription drugs and illegal drugs than are the more poorly educated. Generally, prior use of substances is associated with greater current use of substances. A difference from the results in Table 2 concerns depression. Depression is associated with more use of tobacco products and prescription drug misuse. The easy availability of alcohol in adolescence is associated with less tobacco use, as is religious attendance. Religious attendance is also associated with less use of illegal drugs.

Discussion

The major finding of this paper is that the multiracial-ethnic population is diverse in their substance use. Studies that simply ask if people are multiracial err in combining all such people into just one group. The diversity of use has been shown in this research. We found that the monoracial-ethnic population shows differences in substance use, as does the multiracial-ethnic population. This variability is shown throughout our analysis. Among the Hispanic population, substance use patterns differ in terms of whether one identifies as Hispanic or multiracial-ethnic Hispanic in terms of alcohol use, marijuana use, tobacco use, prescription drug misuse, and illegal drug use. We see similar variations for the Asian-American population of people. There is not as much variation by multiracial-ethnic status among the AIAN populations, except for the white AIAN population.

The findings from our study relate to prior studies on racial-ethnic identification and substance use by helping us to understand more fully some of the patterns observed in prior research. For example, much prior research has shown that the multiracial-ethnic population has higher rates of substance use than monoracial-ethnic people (Chavez & Sanchez, 2010; Choi et al., 2006; Clark et al., 2013; Fisher et al., 2017). However, as we have shown here, this is not true for all multiracial-ethnic people. Our study reveals that black-white, black-Hispanic, multiracial Asian-American, and multiracial AIAN people generally do not differ from self-identified non-Hispanic whites in terms of their use of substances. This important variability is missed by a coding scheme that simply asks people if they are multiracial.

These findings may also have other implications for studies of race-ethnicity and patterns of substance use. For example, previous research notes that Asian Americans have very low rates of substance abuse over their lifetime and within the past year (SAMHSA, 2020), while Native Americans have the highest rates of use for marijuana and other illicit substances (SAMHSA, 2020). But our study showcases that this is not true for all groups of Asian Americans and Native Americans. Self-identified AIAN do not differ from whites in substance use, but self-identified white-AIAN do. Self-identified East Asian Americans do not differ from the reference group of non-Hispanic whites in terms of alcohol and marijuana use, but self-identified Chinese and Filipino Americans do. These are just examples to demonstrate that if we can look more closely at racial-ethnic populations prioritizing their multitudinous self-identification, we can see different patterns than what is commonly thought to be the case.

While we recognize that our study also groups some multiracial-ethnic self-identities (due to issues of sample size, which is one limitation of our study), we argue that the delineation of monoracial and multiracial-ethnic identification into 17 categories went a long way in providing a fuller picture of the influences and variability of racial-ethnic identity on substance use. Race-ethnicity is a measure of lived experiences and is based on the varying combinations of these, individuals’ positionality and intersecting social identities are differentially structured, thereby affecting their social interactions and access to resources, social spaces, opportunities, and life outcomes (Bratter & Gorman, 2011; Roth, 2016; Saperstein, 2012). The variability of multiracial-ethnic identity may further impact phenotypic features that are salient in ascriptions of race-ethnicity and subsequent racial perceptions of others that may determine what racial-ethnic groups individuals identify with, or they think others see them as, or are accepted by (Feliciano, 2016; Roth, 2016). These complex issues of identity are all consequential. Therefore, the parsing that we facilitated with the data was valuable in untangling and highlighting the above consideration of identity and its relationship to substance use.

We did have little representation in certain race-ethnicity groups (e.g., Native Hispanics, Filipino Americans, Other Asian Americans, White-Natives), even though ours was a large sample. This is not just the case with the data used; this pattern is also observed within the US population. These race-ethnicity categories with little representation may increase over time as people simply learn more about themselves and their family history, or for other reasons. For example, it is well known that people are increasingly more likely to self-identify as AIAN over time (Cohn, 2015). The reasons for this are varied, but it illustrates the fact that self-identification can change. Therefore, at a different time, our results could differ. Relatedly, there is a growing body of work on the multiple dimensions of the race construct (Campbell et al., 2016; Roth, 2016; Saperstein, 2012; Saperstein et al., 2016). This means that differences in our findings may occur based on the measure of race used. As our aim was to facilitate a nuanced delineation of racial-ethnic identification, using another dimension of race such as interviewer-reported race and/or skin color (which, while present in Add Health data, both had only five categories) would not have allowed for such an examination. This, however, would be a point of further study to see how other dimensions of race may interact with self-identification to impact findings on substance use. Another limitation is that our use of logistic regression assumes linearity. This is an assumption that may not hold in all cases.

The control variables in our analyses were of importance in ways that were consistent with prior research. Older ages were associated with less use of alcohol. Men were more likely to be users than women, except for prescription drugs. Patterns for education showed that the more highly educated generally have less troublesome substance use. Prior use had an impact on current use of substances. Depression and religious attendance were associated with less use of alcohol. However, depression and the easy availability of alcohol in adolescence were associated with more current use, including heavy episodic use (see C. L. Broman, 2016, 2020; Chitwood et al., 2008; Choi et al., 2006; Fiellin et al., 2013; Fisher et al., 2017; Iwamoto et al., 2011; Jochman & Fromme, 2009).

Our analysis is the first step in understanding the variability that exists in the multiracial and multiethnic population in the use of substances. The measurement of race and ethnicity needs greater consideration in future research, since we have shown how there is great variation in the patterns of use, given how people self-identify with the various racial and ethnic groups. In addition, future research should focus on the factors that influence use of particular substances for the different multiracial and multiethnic populations. This will help us to better understand who uses what substance as we work toward gaining more knowledge of this issue.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14659891.2024.2374802.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.