Psychological Reports, 2003, 93, 365-370.
A CASE-CONTROL ANALYSIS OF FELONY CONVICTIONS AMONG
RECREATIONAL DRUG USERS'
VENKAT S. CHILAKAPATI DAVID F. DUNCAN
University of Louisville Brown University
THOMAS NICHOLSON, JOHN B. WHITE, AND LISA LINDLEY
Western Kentucky University
Summary.—This study applied an epidemiologic case-control design to the identification
of variables predicting which subjects in a population of recreational drug users were
likely to have a felony conviction. A sample of 158 self-identified drug users responded to
the web-based DRUGNET survey from February, 1997 through June, 1998. In this
sample of 704 U.S. citizens who reported using illicit drugs, 79 (11.2%) reported having a
drug-related felony conviction. A further 52 (7.4%) subjects reported felony convictions
not related to drugs. Control subjects were drawn from the same sample and matched to
cases on sex, age, and ethnicity. Subjects with drug-related convictions differed significantly
from controls on 13 variables and those with nondrug-related convictions differed on 11
variables out of the 82 used in this study. Although there are major limitations given the self-
selection of the sample, the differences between the groups indicates that we should be wary
in generalizing from the minority of drug users who are arrested or incarcerated to the
population of drug users in general.
The purpose of this study is to challenge the premise that prohibition of
illicit drug use through incarceration is effective and to suggest that it represents
a fundamentally flawed policy. In the past, conclusions about this hidden
population have been based largely on dubious generalizations from two more
accessible populations, (1) drug abusers in treatment and (2) drug users or
abusers within criminal justice settings. Retrospective accounts of the pre-
clinical experience of drug abusers have commonly been treated as descriptive
of typical drug use. Data on drug users encountered in criminal justice settings
are often similarly generalized. The latter appears to rest on an assumption that
drug users as a group are frequent offenders or alternately that involvement
with the criminal justice system is a randomly occurring event in the lives of
drug users. Therefore, it seems relevant to examine both the frequency of
felony convictions among drug users and whether such convictions are
systematically or randomly distributed in this population.
DRUGNET (Nicholson, White, & Duncan, 1998, 1999) is a voluntary,
self-report, Internet-based study designed to collect data descriptive of users
'Address reprint requests to Prof. Thomas Nicholson, Department of Public Health, Western Kentucky University,
Bowling Green, KY 42101.
366 V. S. CHILAKAPATI, ET AL.
of illicit drugs who do not meet the DSM-IV criteria for a substance abuse
diagnosis (American Psychiatric Association, 1994). Duncan, White, and
Nicholson (in press) suggested that this is an important "hidden" popula tion
in the epidemiological study of drug use.
DRUGNET is a multiyear, cross-sectional, panel survey of adult recre-
ational drug users, i.e., not abusers, conducted via the World Wide Web. A
detailed description of the DRUGNET procedures can be found in Nicholson,
et al. (1998, 1999).
The sample consists of self-identified drug users who have responded to
solicitations posted on various WWW newsgroups, mailing lists, etc., as well
as to word-of-mouth referrals and an article in the online magazine Wired.2
Data for this study were collected from February, 1997 through June,
1998. The survey was completed by 906 persons during this period. This in-
cluded 704 (78%) U.S. citizens and 198 (22%) non-U.S. citizens (« = 4 missing
data). Because drug laws and criminal justice policies differ from country to
country this study was limited to U.S. citizens aged 18 years and older.
We applied an epidemiologic case-control design to the identification of
variables predicting which subjects in a population of recreational drug users
were likely to have a felony conviction. Separate analyses were conducted for
drug-related felony convictions and nondrug-related felony convictions. In
both analyses each case (subject with a felony conviction) was matched with a
control (subject with no felony conviction) of the same sex and similar age (plus
or minus two years). Chi-squared tests were applied to test for significant
differences between cases and controls.
Forward stepwise discriminant analyses were conducted to identify pre-
dictors of whether subjects had felony convictions—drug-related or non-drug-
related, respectively. At each step, the variable that minimized Wilks lambda
was entered as recommended by Huberty (1984). Minimum significance of
partial F to enter a new variable was 0.1. Maximum significance of partial F to
remove a variable from the equation was 0.2.
Prevalence of Felony Convictions
Of the 704 U.S. citizens in this sample who reported using illicit drugs, 79
(11.2%) reported having a drug-related felony conviction. A further 52 (7.4%)
subjects reported nondrug-related felony convictions.
RECREATIONAL DRUG USERS: FELONY CONVICTION 367
Criterion for Drug-related Felony
These analyses were conducted on a sample of 158 subjects, comprised of
79 cases and 79 controls. There were 69 (87.3%) males and 10 (12.6%) females
in each group. Among cases, 72 (92.3%) were White and 6 (7.7%) non-White.
Among controls, 75 (93.6%) were White and 4 were non-White. These
differences were not significant.
Cases differed significantly from controls (% 3 =18.3, /><.05) on education,
a majority having less than a bachelor's degree, while a majority of controls
had achieved at least a bachelor's degree. They also differed significantly
(%52=12.0, p<.05) on household income with a median of $30,000-49,999 for
cases and $50,000-69,999 for controls. Fewer cases (75.6%) than controls
(88.5%) reported voting regularly (X I2 = 4.3, p<.05). More cases (78.8%) than
controls (51.2%) reported that their children were aware of their drug use.
Discriminant analysis indicated 13 significant predictors of group mem-
bership, with Wilks lambda reduced to 0.527 Oc i;2 = 91.27, p<.01). Subjects
with drug-related felonies tended to have less education (median of two years of
college versus a bachelor's degree) and lower incomes (median of $30,000-
49,999 versus $50,000-69,999) than did controls. They reported using marijuana
less often (median of once per month versus at least once per week) and
depressants more often (median of once a month versus once a year) than did
controls. They reported experiencing less intoxication when they used
marijuana (10.4% very intoxicated versus 16%) but more when they used
hallucinogens (32.4% extremely intoxicated versus 23.3%). They were more
likely to report that their current use was about the same as their initial use of
both marijuana (50% versus 36.5%) and depressants (57.6% versus 77.8%).
They were more likely to report health problems due to alcohol use (20.5%
versus 5.3%) and less likely to report health problems due to marijuana use
(7.8% versus 12.3%). They were less likely to report having cut down on
hallucinogens given health problems (25% versus 66.7%) and reported a more
positive overall effect of hallucinogen use on their lives (median rating of 8
on a scale from 0 to 9 versus a median of 7). Use of these predictors resulted
in the correct classification of 83.9% of the subjects.
Criterion for Nondrug-related Felony
The sample for these analyses were 52 subjects, 26 cases and 26 matched
controls. There were 20 (76.9%) males and 6 (23.1%) females in each group.
Among cases, 25 (96.2%) were White and one (3.8%) was non -White, while
all of the controls were White. Chi-square analyses indicated no significant
differences between cases and controls on any of the demographic or lifestyle
368 V. S. CHILAKAPATI, ET AL.
The discriminant analysis for nondrug-related felonies identified 11 pre-
dictor variables, reducing Wilks lambda to 0.217 (%U2 = 65.03, p<.01). In this
analysis, case subjects assigned a greater importance to spirituality in their
lives, a median of 6 versus a median of 5 on a scale anchored by 0: no im-
portance and 9: central focus of life) and were more likely to attend religious
services regularly (24% versus 11.5%) than their matched controls. Case sub
jects were also more likely to report that they did not vote (30.8% versus
15.4%). They were less likely to report cutting down on hallucinogens due to
health problems (35% versus 42.9%). They reported a more positive effeet of
alcohol on their life (median of 6 versus 5 on a 10-point scale) than did
controls. They were more likely to report that cocaine had harmed their health
(36.8% versus 13.3%). They reported greater intoxication when they used
cocaine (median rating of 4 versus 3). They reported less hallucinogen use
(median of once a month versus once a year). More of them reported cocaine
use in the past year (38.1% versus 12.5%). While both cases and controls
reported relative freedom from worry or concern about their health, cases
showed small but significantly lower concern (mean of 11.4 versus 11.0 on
scale on which 15= no worries or concerns). Fewer cases reported alcohol use
in the past year (83.3% versus 88.5%). Using these predictors, 94.1% of the
subjects were correctly classified.
There are no comparative data available on the lifetime prevalence of
felony convictions in the general population. It is not possible, therefore, to
make any direct comparison between the DRUGNET sample's conviction
rates and the general population. It has been estimated that 2.1% of adult
U.S. citizens are deprived of the right to vote because they have a felony
conviction. Given that not all states deprive convicted felons of voting rights
and some do so only for a fixed period of years or only for certain felonies,
this percentage does not represent the full proportion of the population with
such convictions (Sengupta, 2000).
Likewise, the estimate that 5.1% of the U.S. population will be imprisoned
in a state or federal prison at some time during their lives (Bonczar & Beck,
1997) does not provide a direct basis for comparison. This estimate is a
projection of the future and is based on current incarceration rates that are
at an all-time high, thus inflating the rate for comparative purposes. On the
other hand, it does not include persons sentenced to probation or other
alternatives instead of prison. In comparison to either of the estimates
referred to above the prevalences of 11.2% and 7.4% for drug-related and
nondrug felonies, respectively, appear to be high. It is not surprising, of
course, that a population of drug users might have a high rate of drug-related
felony convictions. Lack-
RECREATIONAL DRUG USERS: FELONY CONVICTION 369
ing a direct basis for comparison, we can only tentatively conclude that illicit
drug users have a higher probability of felony conviction than do nonusers. It
should be noted, however, that the majority of this sample of drug users did
not report a felony conviction. Thus any association between nonabusive drug
use and crime appears to be weak at best.
Cases and controls differed little in terms of self-reported demographic
characteristics. Those with drug-related convictions reported significantly less
education and lower income on the average. The differences, however, were
not large, and there was substantial overlap of the distributions. There were no
demographic differences between subjects with nondrug-related convictions
and controls. This demographic similarity may reflect the somewhat limited
demographic information typical of an Internet-based sample. Such samples are
likely to be largely White, male, well educated, and prosperous.
Subjects with drug-related convictions differed from controls on 13 vari-
ables and those with nondrug-related convictions differed on 11 variables.
That they differed on these variables gives further reason to doubt the wisdom
of generalizing from the minority of drug users who are arrested or incarcerated
to the population of drug users in general.
Because this sample is self-selected it may not be representative of all
potential participants. Further, the potential participants were limited to
persons who use the Internet and among users to persons who participated in
the newsgroups, mailing lists, etc. on which DRUGNET was advertised or
who read the issue of Wired in which the article on DRUGNET appeared.
The sample thus generated may very possibly have differed in some unknown
way from the population of recreational drug users in the USA or the
subpopulation of such users who have a history of felony convictions.
White, Nicholson, Minors, and Duncan (2001) previously stated:
As with all survey research, the honesty of respondents when answering questions about highly
personal and in many instances, illegal, activity is of primary concern. However, no research
exists suggesting that persons responding on the Internet are any more or less truthful than in -
dividuals responding to surveys via phone or "anonymous" surveys sent through postal mail. It
may be that the perceived anonymity of the Internet allows people to be more frank in
responding to items.
The unique aspect of the DRUGNET survey is its use of an online survey form to access a hid -
den population and acquire sensitive information. Contrast this to the methods used in the
SAMHSA National Household Survey on Drug Abuse (1999), in which government contracted
workers visit individuals at their home, assure them that although the surveyor represents a
government agency, all the individual's responses are confidential, and proceed to ask them
whether they consume a variety of drugs which are illegal under federal law, and record their
responses on a paper and pencil coding sheet. Despite the level of confidence we may have in
the honesty of DRUGNET respondents, until research demonstrates whether systematic differ -
ences exist between the use of this methodology to administer surveys or other, more tradi -
tional methods, readers are cautioned against generalizing these r esults to the general popula-
tion, or to persons who utilize the Internet as a whole (p. 367).
370 V. S. CHILAKAPATI, ET AL.
Nonabusive drug users continue to be a hidden population about
which too little is known. The policy relevance of knowledge about this pop-
ulation should make it a priority target for research. The DRUGNET analyses
provide one source of data on this population but the inherent limitations of
an Internet-based survey must always be kept in mind. Additional studies
exploring different methodologies, including studies not reliant on self -report,
need to be developed and implemented. Continued generalization to this much
larger population from drug abusers or from drug users in the criminal justice
system can no longer be justified.
AMERICAN PSYCHIATRIC ASSOCIATION. (1994) Diagnostic and statistical manual of
mental disorders. (4th ed.) Washington, DC: Author.
BONCZAR, T. P., &BECK, A. J. (1997) Lifetime likelihood of going to state or federal
prison. (Bureau of Justice Statistics Special Report) Washington, DC: U.S.
Department or Justice.
DUNCAN , D. E, WHITE , J. B., & NICHOLSON, T. (in press) Using an Internet-based
survey to reach the hidden population of non-abusive illicit drug users.
American Journal of Health Behavior.
HUBERTY, C. J. (1984) Issues in the use and interpretation of discriminant analysis.
Psychological Bulletin, 95, 156-171.
NICHOLSON , T., WHITE, J., & DUNCAN, D. F. (1998) DRUGNET: a pilot study of
adult recreational drug use via the WWW. Substance Abuse, 19, 109-121.
NICHOLSON , T., WHITE, J., & DUNCAN, D. F. (1999) A survey of adult recreational
drug use via the WWW: the DRUGNET study. Journal of Psychoactive
Drugs, 31, 415-422.
SENGUPTA , S. (2000) Felony costs voting rights for a lifetime in 9 states. New York
Times, Nov. 3, A-18.
SUBSTANCE ABUSE AND MENTAL HEALTH SERVICES ADMINISTRATION. (1999)
Summary of findings from the 1998 National Household Survey on Drug
Abuse. [DHHS Publication No. (SMA) 99-3328] Rockville, MD: U.S.
Department of Health and Human Services.
WHITE, J. B., NICHOLSON, T, MINORS, P., & DUNCAN, D. (2001) A demographic
profile of employed users of illicit drugs. In M. A. Rahim, R. T.
Golembiewski, & K. D. MacKenzie (Eds.), Current Topics in Management,
Accepted July 17, 2003.