Content uploaded by Jonathan Howland
Author content
All content in this area was uploaded by Jonathan Howland
Content may be subject to copyright.
Employee Drinking Practices and Work Performance*
THOMAS W. MANGIONE, PH.D., JONATHAN HOWLAND, PH.D.,* BENJAMIN AMICK, PH.D.,* JENNIFER COTE, B.A., •
MARIANNE LEE, M.P.A.,* NICOLE BELL, SC.D.,* AND SOL LEVINE, PH.D.*
JSI Research & Training Institute, 44 Farnsworth Street, Boston, Massachusetts 02210
ABSTRACT. Objective: The purpose of this study was to examine the
independent effects of a variety of drinking indicators on self-reported
work performance. Method: Data from a cross-sectional mailed survey
(response rate = 71%) of managers, supervisors and workers (N = 6,540)
at 16 worksites were analyzed. Average daily volume was computed
from frequency and usual quantity reports. Drinking on the job included
drinking during any of six workday situations. The CAGE was used to
indicate alcohol dependence. Employees were also asked how frequently
they drank to get high or drunk. Work performance was measured
through a series of questions about work problems during the prior year.
The number of times respondents experienced work performance prob-
lems was regressed on the four drinking measures, and a variety of de-
mographic characteristics, job characteristics and life circumstances that
might also negatively affect work performance. Results: The frequency
of self-reported work performance problems increased, generally, with
all four drinking measures. In a multivariate model that controlled for
a number of demographics, job characteristics and life-situations, aver-
age daily volume was no longer significantly associated with work per-
formance but the other three drinking measures were. Interestingly,
although moderate-heavy and heavy drinkers reported more work per-
formance problems than very light, light, or moderate drinkers, the
lower-level-drinking employees, since they were more plentiful, ac-
counted for a larger proportion of work performance problems than did
the heavier drinking groups. Conclusions: Employers should develop
clear policies limiting drinking on the job and, in addition to employee
assistance programs for problem drinkers, should develop worksite edu-
cational interventions aimed at informing all employees about the rela-
tionship between drinking behaviors and work performance. (J. Stud.
Alcohol 60: 261-270, 1999)
HERE HAVE BEEN PERIODS in U.S. history when
employers have promoted the use of alcohol during the
workday, believing that it helped to stimulate productivity
and promote employees' sense of well-being (Lender and
Martin, 1987). In recent decades, however, workers' alcohol
problems in general, and drinking on the job in particular,
have been a matter of concern to business owners and man-
agers. Indeed, in special cases drinking during (e.g., transport
workers), or even before (e.g., airline pilots), the workday
has been subject to federal regulations.
The issue of alcohol use by employees has two distinct di-
mensions. The first concerns impairment while actually per-
forming work "under the influence." The second concerns the
impact on work performance of heavy and/or habituated
drinking during nonworking hours. The acute effects of alco-
hol exposure on a variety of physiological and cognitive func-
tions have been well documented (Perfine, 1973). Less is
Received: June 24, 1996. Revision: June 26, 1997.
*This research was supported by National Institute on Alcohol Abuse and
Alcoholism grant 5R01 AA0904302 and Robert Wood Johnson Foundation
grant 18525.
*Jonathan Howland and Jennifer Cote are with the Boston University
School of Public Health, Boston, MA. Benjamin Amick is with the New
England Medical Center, The Health Institute, Boston, MA, and the Harvard
School of Public Health, Boston, MA. Marianne Lee is with the Harvard
School of Public Health, Boston, MA. Nicole Bell is with SSDS, Inc., Nat-
ick, MA. Sol Levine (deceased) was with the New England Medical Center,
The Health Institute, Boston, MA, and the Harvard School of Public Health,
Boston, MA.
known, however, about the relationship of the previous
night's drinking to work performance even when blood alco-.
hol concentration is zero during the workday.
This study explores the relationship between the frequency
of reported work performance problems and a variety of
drinking indicators (drinking during the workday, hsual
drinking amounts, drinking to get high, and drinking by those
who are alcohol dependent). The question of how, and to what
extent, on-the-job and off-the-job drinking negatively affect
work performance is important because of the economic im-
plications of reduced productivity (Harwood et al., 1985) and
because deficient job performance is typically the trigger for
initiating employer intervention into workers' drinking prob-
lems. This emphasis on work performance, as opposed to
other observable physiological and psychological symptoms,
has evolved over recent decades and is predicated on the as-
sumption that work performance of problem drinkers shows
a measurable decline well before acute alcoholism occurs.
Thus, compromised work performance has come to be viewed
as a means for case finding at a point when effective inter-
vention is feasible (Blum et al., 1993).
Several laboratory studies have been conducted on the
residual, or hangover, effects of intoxication on performance.
These studies have used simulated work situations to assess
functioning at zero blood alcohol concentration (BAC), fol-
lowing a previous night's intoxication and sleep. Wolkenberg
and associates (1975) studied the acute and delayed effects of
alcoholic intoxication on work performance as measured by
a variety of laboratory devices simulating industrial tasks.
261
262 JOURNAL OF STUDIES ON ALCOHOL / MARCH 1999
Delayed effects were observed at zero BAC up to 18 hours
after alcohol ingestion. These included lengthened reaction
time, poor motor performance, decreased motor sensory skill,
limited visual scanning and deteriorated posture.
Morrow et al. (1991) and Yesavage and Leirer (1986) have
shown significant decrements in aircraft handling perfor-
mance among military pilots performing flight simulator ex-
ercises the day after intoxication. Tomros and Laurell (1991)
have demonstrated similar hangover effects for drivers using
automobile simulators. Other investigators have also demon-
strated the residual effects of drinking on various other mea-
sures of cognitive and physical performance (Earleywine,
1993; Karvinen et al., 1962; Lemon et al., 1993; Roehrs
et al., 1991; Seppala et al., 1976; Takala et al., 1958).
Other studies also have shown positive relationships be-
tween problem drinking and absenteeism (Maywell, 1960,
1972; Observer and Maxwell, 1959; Trice, 1962, 1965a,b);
self-reports of lower work performance during periods of
problem drinking (Maxwell, 1960, 1972; Trice, 1962, 1965a);
absenteeism and the use of alcohol for coping (Brooke and
Price, 1989); and problem drinking and medically certified
absences (Beaumont and Hyman, 1987).
Two recent studies (Ames et al., 1997; Blum et al., 1993)
have examined the relationship between drinking and work
performance and have come to somewhat different conclu-
sions. Blum and her colleagues examined the relationship
between work performance and general drinking patterns
among a relatively small convenience sample (comparison
subjects in a study of the social effects of minor tranquilizer
use) of 136 employed men. These investigators found no
significant relationships between work performance and
average daily volume when performance was assessed by
self-report. Surprisingly, they found that heavier drinkers re-
ported less absenteeism and tardiness than lighter drinkers.
Nevertheless, when work performance was assessed by col-
laterals, significant decrements were found in a number of
performance measures, particularly those relating to social
and interpersonal relationships (Blum et al., 1993).
Ames and colleagues (1997) conducted their study in one
manufacturing plant among 832 randomly selected hourly
workers. Drinking was measured by self-reports of drinking
before or during work over the past year, coming to work
hungover, heavy drinking and usual drinking amounts. These
investigators found modest but significant relationships be-
tween drinking behaviors and self-reports of workplace
problems in both bivariate and multivariate analyses in
which several job characteristics and background variables
were controlled. In particular, the multivariate analyses
showed significant associations between work and coming to
work hungover with the composite indicator as well as with
the subscales of sleeping on the job and problems with su-
pervisors (drinking at work only) or problems with job
tasks/co-workers (hangovers only).
The present study uses data from a survey of employees at
16 worksites located throughout the United States to exam-
ine the relationship between drinking behaviors and self-
reported work performance. This study can make a contribu-
tion to the accumulating research on this topic because it is
based on several companies, covers a diverse group of in-
dustries and occupations, and includes employees from all
hierarchical levels.
Survey design
Method
Data reported herein are based on a second phase of data
collection in the Worksite Alcohol Project. The first phase of
the project, conducted during the fall of 1992 and winter of
1993, involved surveys of the managerial workforce in 114
worksites of seven corporations. The seven corporations in-
cluded two conglomerates, a paper manufacturer, a chemical
company, an insurance company, a building materials com-
pany and a regional utility company. The sites were ran-
domly selected from among those sites at each corporation
that had 30 or more supervisors/managers. Response rate for
the mail survey was 79%.
In the second phase of the project, data were collected in
1994 from 16 of the original 114 worksites using surveys dis-
tributed to workers as well as managers and supervisors. The
16 sites were randomly selected from a stratified list of the
114 sites. Sites were stratified on the type of workforce, in-
dusthal setting and managerial tolerance toward drinking in
the following ways. Sites were divided into manufacturing
and nonmanufacturing sites. The nonmanufacturing sites
were further divided by whether their workforce was predom-
inately professional level employees or not. This resulted in
three strata: professional, manufacturing and nonprofessional-
nonmanufacturing. The manufacturing sites employed pre-
dominately male workers and the nonprofessional-nonmanu-
facturing sites had sizable proportions of female employees.
The professional sites had the most highly educated work-
force. Within each of these three strata, sites were ordered by
tolerance of managers toward drinking (as assessed by the
Phase 1 survey) and divided into the highest one-third,
medium third, and lowest one-third. From the professional
and nonprofessional-nonmanufacturing strata, two sites were
randomly selected from the highest tolerance strata and two
sites from the lowest tolerance strata. Among the manufactur-
ing sites, since there were many more of these types of sites to
begin with, in our Phase 1 study, four sites were randomly se-
lected from the highest tolerance strata and four from the low-
est tolerance strata. In addition, back-up sites were randomly
selected in case a site was unable to participate. Four sites re-
quired substitutes: one site was consolidated with another
business group and thus ceased to exist as an independent
company in Phase 2; one declined participation due to union
resistance to the survey; one was in the midst of another large
survey and didn't want to overburden employees; and one was
sold and was no longer a part of the participating conglomer-
MANGIONE ET AL. 263
ate. The 16 participating sites included chemical processing
plants, paper mills, both heavy and light manufacturing sites,
finance, communications, insurance and public utility sites.
At 12 of the sites there were fewer than 800 employees so
all employees were selected. At four sites random sampling
by workgroup was done-to create a sample size of approxi-
mately 800 employees. All employees were identified by
workgroup•mployees who reported to the same immediate
supervisor--through a code printed on their questionnaire.
All employees in each selected workgroup were included. An
overall sample of 9,163 employees was selected.
Questionnaires were mailed directly to employees' home
addresses with the exception of two sites where they were
sent to the business address. Letters were included from the
project staff as well as from their plant/site manager (co-
signed by the union leader if it was a unionized site). Individ-
ual anonymity was guaranteed in both letters. A postcard that
did have individual identifiers on it was also included to sig-
nify return of the questionnaire and was to be mailed back
separately. Surveys were sent directly back to the research
team in business reply envelopes. Up to three reminders were
sent out at 2-week intervals for nonrespondents. Monetary in-
centives included a $5 advanced cash payment and participa-
tion in a lottery for one of three (per site) $250 prizes. This
financial incentive package varied somewhat from site to site.
Operationalization of variables
Dependent variable: Work performance. Our composite
work performance measure was the sum of answers to six
questions about problems on the job. These questions asked
how many times in the previous 12 months respondents had
(proportion saying some amount larger than never is shown
in parentheses): missed work (57%); done poor quality work
(38%); arrived late or left early (59%); done less amount of
work (50%); had an argument with a co-worker (42%); or
been hurt on the job (12%). For each of these questions there
were four response categories: never (coded as 0); 1-2 times
(coded as 1.5); 3-5 times (coded as 4); or 6 or more times
(coded as 6). Accordingly, a higher score indicates an in-
creasing number of work performance problems. The mean
(-SD) score for all employees on the overall scale was
6.15 _ 4.71.
In some of our analyses information is presented for indi-
vidual problems, but in our multivariate analyses we used the
composite score. We do so even though the inter-item corre-
lations among the problems are not large and hence the co-
efficient alpha is relatively low (.57). However, our interest
in this overall indicator is not based on the expectation that
the components should have high intercorrelations, but
rather that from the corporation's view the totality of these
events is an important and valid concept.
Because the composite scale is highly skewed, the multi-
variate analyses used a transformed version of it. We added
"1" to the score values and then took the log to the base "e."
Independent variables: Drinking measures. We used four
measures of drinking behavior. The first measure, drinking
to get high/drunk, was obtained from a single question: "Al-
most everyone has times when he/she drinks and gets high or
drunk or has a 'buzz on.' How often has this happened to you
in the past 30 days?" There were eight response categories
ranging from never to every day. Employees who drank but
not in the last 30 days were automatically assigned a score of
"never," as were those employees who did not currently
drink. Because so few employees answered in the more fre-
quent categories we combined them into one category of
twice a week or more. Therefore, the measure entered into
analyses was an ordinal variable with five categories. Rates
of this behavior were calculated for three different employee
populations--all employees, currently drinking employees
and employees who drank within the past 30 days. The pro-
ponions for these three employee populations reporting
some drinking to get high in the past month were: all em-
ployees (28%); currently drinking employees (37%); and
those drinking in the past 30 days (45%). The proportions re-
porting this happening once a week or more were: all em-
ployees (7%); currently drinking employees (10%); and
employees drinking within the past 30 days (13%).
The second measure, drinking on the job, was measured
by asking employees who had consumed alcohol in the past
30 days if during this period they had any alcohol to drink in
the following six work-related situations: within 2 hours be-
fore going to work; during a lunch break; during a work
break; while actually working; at a company sponsored
event; or before driving a company vehicle. Employees were
then classified as drinking on the job if they responded "yes"
(1) to any of the above questions. Employees who had not
drunk at all in the past 30 days were automatically coded as
"no" (0). Drinking on the job in the last 30 days was reported
by 5% of all employees, 6% of currently drinking employees
and 8% of employees who drank in the past 30 days.
The third measure, the CAGE, is a four-item instrument
designed to identify problem drinkers. It comprises the fol-
lowing questions: "Have you ever felt the need to cut down
on your drinking? .... Have people ever annoyed you by criti-
cizing your drinking? .... Have you ever felt badly or guilty
about your drinking?" and, "Have you ever had a drink first
thing in the morning (an eye-opener)?" Any employee who
had ever consumed alcohol was asked these questions. Pur-
suant to standard usage (Mayfield et al., 1974), persons an-
swering "yes" to two or more of these questions were
classified as problem drinkers (1) and those with fewer than
two or who were not asked the questions were coded as "no"
(0). The proportion exceeding the cutpoint for all employees
was 19%; currently drinking employees, 20%; and those
drinking within the past 30 days, 23%.
Finally, average daily volume was calculated by multiply-
ing frequency by drinking during the past 30 days by usual
volume of drinking on those days and then dividing by 30
days. Employees who did not drink at all or did not drink
264 JOURNAL OF STUDIES ON ALCOHOL / MARCH 1999
within the past 30 days were assigned a score of zero. For all
employees the mean value of this variable was 0.73 _+ 1.03
drinks per day. Because the distribution on this variable was
highly skewed, in the multivariate analyses reported herein
this variable was transformed using log to the base "e" by
first adding "1" to the average amount consumed.
In some tables we describe the relationship between aver-
age daily volume and work performance by creating a six-
level categorical version of average daily volume. We
distinguish also in this categorical version among those who
are abstainers versus very light drinkers, even though both
are given a score of "0" in our construction of average daily
volume. The categories are: current abstainers, very light,
light, moderate, moderate-heavy, and heavy drinkers. Drink-
ing level categories were constructed as follows: abstainers
were respondents who had never used alcohol or who had not
had a drink in the past 30 days because they no longer used
alcohol; very light drinkers were those who had not had a
drink in the past 30 days but who were not abstainers; light
drinkers were those who drank in the past 30 days but drank
less than 13 drinks during that period; moderate drinkers
drank 13 or more drinks during the past 30 days but never had
more than five drinks (males) or four drinks (females) during
a single day (Cahalan et al., 1969; Wechsler et al., 1995);
moderate-heavy drinkers drank between 13 and 60 drinks
during the past 30 days and five or more drinks (males) or
four or more drinks (females) on at least 1 day during that pe-
riod; heavy drinkers drank 60 or more drinks during the past
30 days and had at least 1 day during that period when they
drank five or more (males) or four or more (females) drinks.
Job characteristics. Work shift was categorized as a series
of dummy variables as follows (with the proportions report-
ing shown in parentheses): rotating shift (16%); night shift
(3%); evening shift (6%); or day shift (75%). In the multi-
variate analyses day shift was the reference group. Work
hours reflected the number of hours per week (including
overtime) a respondent worked on average at the company
participating in the survey. Response categories were: less
than 30 hours (2%); 30-40 hours (32%); 41-50 hours (53%);
51-60 hours (11%); or more than 60 hours (2%). Hours
worked at other jobs were not measured.
Job satisfaction was measured with the question, "All
things considered, how satisfied are you with your job?" Re-
sponse categories were: not at all (6%); a little (16%); some-
what (49%); or very (30%). We also determined the number
of years a respondent had worked at their current job using
seven categories from "less than six months" to "more than
20 years." The median length of job service was 3 years.
Health. Health status was measured using the question,
"Overall, how would you rate your health?" Response cate-
gories were: excellent (coded as 5) ( 19 %); very good (43 %);
good (32%); fair (5%); and poor (coded as 1) (0.5%).
Other drug use. We also asked about the most recent use
of marijuana and prescription drugs for anxiety or depres-
sion. Response categories for marijuana use were: never
(coded as 1) (74%); more than a year ago (21%); within the
last year (3%); and within the last month (coded as 4) (2%).
We were constrained from asking about other illicit drug use
because of company and union reluctance to have this infor-
mation collected. Use of prescription medications during the
prior year for anxiety or depression (like Valium, Prozac or
sleeping pills) was coded as: never (coded as 1) (88%);
rarely (5%); sometimes (3%); often (2%); and very often
(coded as 5) (2%).
Demographics. Respondent age was measured using five
age categories: less than 25 years old (4%); 25-34 (22%);
35-44 (33%); 45-54 (29%); 55 or more years (12%). Position
in management hierarchy was divided into five categories: no
supervisory responsibility (coded as 1 ) (71%); first-line super-
visor (15%); middle management (11%); upper-middle man-
agement (2%); and top-level management (coded as 5) (0.5%).
Respondent gender: males (coded as 0) (61%), females (coded
as 1) (39%). We also collected data on level of education (seven
categories from less than a high school diploma to graduate de-
gree) and whether or not respondents lived in a household with
children (coded as I -- living with children) (44%).
Data analysis
Descriptive analysis. To depict the relationship between
drinking and work performance, we calculated, for respon-
dents in each of the six drinking-level categories, the mean
number of times they reported performance problems for the
previous year. These were displayed as a bar graph (mean
number of work performance problems by drinking level
category). Using ANOVA, we also analyzed the relationship
between drinking level category and each of the six types of
work performance measures as well as the composite mea-
sure of work performance problems.
Regression analysis. We performed regression analysis
with the individual's number of work performance problems
as the dependent variable and the four drinking measures as
independent variables. To control for other factors poten-
tially confounding the relationship between drinking and
performance, we also included as independent variables
other factors potentially contributing to work performance
problems. These other independent variables were: respon-
dents' age; gender; status in managerial hierarchy; shift
(evening; night; or rotating, relative to day shift); years on the
job; work hours per week at the company; self-reported
health status; job satisfaction; use of marijuana; use of pre-
scription drugs for anxiety or depression; and whether re-
spondents resided in a household with children. All variables
were entered into the regression simultaneously.
Results
Survey response. The overall response rate for Phase 2 was
71%, with individual site response rates ranging from 59%
to 91%. The variation in site response was not correlated with
variation in incentive strategy. The final response was 6,540
completed surveys returned.
MANGIONE ET AL. 265
lO
E
l
6
4
2
4.88
5.41
5.85
Abstainers Light
Very Light
5.74
6.88
7.43
Moderate-Heavy
Moderate Heavy
FmuRœ 1. Average total work performance problems for different levels of drinking
Characteristics of respondents. The demographic charac-
teristics of the responding sample were as follows: 60% of
the respondents were men, nearly 75% were married, 30%
had at least a 4-year college education, 85% were white, and
about 25% were less than 35 years old. The median job in-
come was $36,000. A third were Catholic and 27% were
Conservative Protestants. Three-quarters were current
drinkers and 18% of those who had ever drank reported a
drinking problem at some point in their lives.
Work performance problems by drinking level categories.
The overall number of work performance problems in-
creased as a function of increasing drinking level category
(see Figure 1). When we repeated this analysis for each of the
six types of work problems individually, there were gener-
ally positive relationships between drinking category and
each type of performance problem with the exception of i.n-
jury on the job. The relationship between drinking and injury
on the job was U-shaped, with the highest rates of reported
injuries occurring among abstainers, and among heavy
drinkers, and lower rates for respondents in the intermediate
drinking level categories. We ran t-test contrasts for each
category of drinker and found that abstainers were signifi-
cantly different from each level except heavy drinkers. The
relationship between drinking category and work problems
is strongest for late to work/early to leave and for doing less
work (see Table 1).
TABLE 1. Mean number of work performance problems by drinking category
Low
Drinking Missed quality
category days work
TYPE OF WORK PERFORMANCE PROBLEM
Late Argue
in/early Less with
out work co-worker
Injured All
on job problems
Abstainer 1.24 0.57
Very light 1.42 0.68
Light 1.38 0.73
Moderate 1.17 0.67
Moderate-heavy 1.53 0.86
Heavy 1.52 0.91
F 6.23 10.29
p .0001 .0001
R 2 .005 .008
1.08 0.93 0.77 0.30 4.88
1.29 0.98 0.83 0.21 5.41
1.65 1.11 0.82 0.16 5.85
1.68 1.15 0.91 0.16 5.74
1.95 1.42 0.99 0.14 6.88
2.01 1.55 1.18 0.30 7.43
41.51 18.98 7.19 12.48 29.19
.0001 .0001 .0001 .0001 .0001
.032 .015 .006 .010 .022
266 JOURNAL OF STUDIES ON ALCOHOL / MARCH 1999
Regression analysis predicting work performance prob-
lems. We regressed work performance problems on the four
measures of drinking, job characteristics, demographic vari-
ables, and the other potential causes of work performance
problems. Overall, our model was significant at p = .0001,
with an adjusted R 2 of. 15 (see Table 2). Three drinking vari-
ables were independently, positively and significantly asso-
ciated with the dependent variable: drinking at work, scoring
high on the CAGE, and frequency of drinking to get high or
drunk. Average daily volume was reduced to a nonsignifi-
cant association primarily because of its correlation with fre-
quency of getting high or drunk.
Among those variables that we included because of their
potential confounding relationships, we found several sig-
nificant relationships. Job satisfaction and health status were
negatively related to problems. Both use of marijuana and
use of prescription drugs for anxiety or depression were in-
dependently and positively associated with work perfor-
mance problems.
Among the demographic characteristics, age was signifi-
cantly and negatively associated with work performance
problems, and having children at home was positively asso-
ciated with work performance problems. Gender and years at
current job were not significantly associated with the num-
ber of work performance problems.
Job characteristics also showed mixed results. The more
hours worked per week at the company the more work per-
formance problems there were. Also, working the evening or
rotating shift (relative to the day shift) were negatively asso-
ciated with work performance problems. Working the night
shift, however, was not statistically significant, nor was po-
sition in the management hierarchy.
We also ran the regression analyses for each individual
work performance indicator, but have not shown these re-
sults because of space considerations. However, the general
pattern of results in relationship to the drinking indicators
was replicated in each of the analyses.
Discussion
Overall, there was a positive linear relationship between
the frequency of reported work performance problems and
drinking behavior, although the amount of variance ex-
plained was modest. Our findings in general are noteworthy
in several respects. First, we observed a relationship between
drinking during the workday and work performance prob-
lems, independent of general drinking practices. Second, we
found a relationship between heavy drinking during non-
working hours and performance problems during working
hours, controlling for drinking on the job. Third, these find-
ings were demonstrated with a sample that included a range
of industrial settings and company cultures.
Since our study was similar to those of Ames et al. (1997)
and Blum et al. (1993), comparisons to their findings are
noteworthy. First, Blum and colleagues found no significant
relationships between average daily volume and self-reported
work performance, but did find significant decrements when
performance was assessed by collaterals. These results led
the authors to speculate about the validity of self-reported
work performance data. Our study, however, used self-reported
work performance measures exclusively, and we did find sig-
nificant and positive relationships between alcohol use and
work performance problems. Second, these investigators
found that heavier drinkers were less apt than light drinkers
TABLE 2. Predicting work performance problems
Unstandardized
parameter
Variable estimates SE t p
CAGE .117 .0258 4.54 .0001
Drinking on job .189 .0461 4.09 .0001
Freq. drinking to get high/drunk .027 .0138 1.95 .05
Avg. daily volume .016 .0340 0.47 .63
Age -. 136 .0105 - 12.92 .0001
Health status -. 111 .0116 -9.61 .0001
Job satisfaction -. 134 .0115 - 11.68 .0001
Marijuana use .131 .0162 8.08 .0001
Anxiety/depression drug use .092 .0128 7.20 .0001
Children at home .067 .0190 3.54 .0004
Hours worked/week .027 .0142 1.93 .05
Rotating shift -.222 .0274 - 8.10 .0001
Night shift -. 145 .0575 -2.52 .0116
Evening shift -.031 .0413 -0.76 .4472
Gender (female) .032 .0215 1.469 .1420
Level in hierarchy .016 .0123 1.292 .1963
Years at current job -.0001 .0066 - 0.024 .9807
MODEL
N F p Adj. R 2
5,943 61.894 .0001 .15
MANGIONE ET AL. 267
to report work absences and tardiness, whereas we found a
strong positive association between drinking and tardiness.
Indeed, of all our performance measures, tardiness and/or
leaving work early was the variable most strongly associated
with increased alcohol consumption. Third, they found that
collateral reports of work performance indicated a stronger
relationship between drinking and workplace interpersonal
problems than between drinking and technical aspects of
work performance. Using our self-reported data, we found
stronger relationships between drinking and technical aspects
of work performance (low quality work; less work) than we
did for interpersonal aspects (argue with co-workers).
Our findings corroborate those reported by Ames et al.
(1997). The strength of our associations for total problems
are of similar, if modest, magnitude. We both find that self-
reports of drinking on the job as well as residual effects of
heavy drinking are related to work performance p•blems.
Ames and colleagues used a direct question about coming to
work hungover, while our study used a question about fre-
quency of drinking to get high or drunk to represent the resid-
ual effects of heavy drinking. Both studies also showed that
these more immediate behaviors explained the bivariate re-
lationship between general drinking levels and work perfor-
mance problems. The only difference in our results were the
direction of effects for some of the confounders. They found
women less likely to report problems, whereas we found no
gender effect after controlling for the other factors in the
model. They found day shift workers less likely to have work
performance problems than those working other shifts. In
contrast, we found that those working evening and rotating
shifts were less likely to report work performance problems.
Our findings also differ from those of previous investiga-
tors with respect to workplace injuries. A number of re-
searchers have found little evidence to support an association
between heavy drinking and job injuries. Ames et al. (1997)
did not find an association between drinking and their injury/
medical problems subscale in their multivariate models al-
though they did report a significant bivariate relationship.
Some have suggested that drinking problems may even have
a protective effect in that workers with drinking problems,
and their supervisors, may modify work demands or situa-
tions to reduce the likelihood of on-the-job injury (Archer,
1977; Blum et al., 1993; Trice and Roman, 1972). It is pos-
sible that workers with drinking problems are indeed pro-
tected from dangerous work assignments but may experience
higher rates of minor injuries. Thus, depending on the source
of injury data--workplace reports or self-reports•ifferent
relationships between drinking and work-related injury
might be found. In addition, our results show a U-shaped re-
lationship between drinking levels and job injuries, suggest-
ing another reason why previous studies have failed to find
linear associations. Moreover, the change in the slope of the
relationships we found occurred between moderate and
heavy drinkers. Therefore, whether or not heavy drinkers are
shown to have higher rates of job injuries may vary depend-
ing upon the cutpoint used to differentiate moderate from
heavy drinkers. Our results also showed relatively high rates
of injury among abstainers. A large proportion of the ab-
stainers in our study worked in blue-collar manufacturing
environments, and this might account for their higher rates.
Leigh (1996) has suggested that heavy drinkers tend to work
at hazardous jobs and thus the alcohol-injury association is
confounded by workplace exposure.
Although it was not the primary focus of this study, we ac-
knowledge two points about the other significant variables in
our regression analysis. These variables were included be-
cause they might be confounders to the observed relationship
between drinking and work performance. Many of these con-
founders were indeed themselves significantly associated
with work performance problems. The demographic charac-
teristics associated with performance problems were younger
age and the presence of children at home. Interestingly, after
controlling for all other factors in the model, gender was not
significant. These findings may reflect changes in society
such that the family lives of both women and men impact on
work. Not surprisingly, job satisfaction was significantly and
negatively associated with work performance problems. The
strongest associations, however, were related to self-rated
health status, use of marijuana, and the use of other drugs for
the treatments of anxiety or depression. These relationships
deserve more analysis than our dataset allows. Nevertheless,
what is important about these findings is that the three alco-
hol use measures remain significant predictors of work per-
formance problems even when controlling for a variety of
other contributing factors.
The level of cooperation we received from both labor and
management at our research sites was extensive, although
not without compromise. A major concern of theirs was that
individual respondents could be identified by combining de-
mographic data with information on job characteristics.
Thus, several of our variables were measured with limited
precision. Education was collected as an ordinal variable al-
though it would have been better collected as a continuous
(years of education) variable. We do not believe, however,
that in this case having had years of education would have re-
suited in a substantial change in our results.
Both labor and management were reluctant to have us col-
lect data on the use of drugs other than alcohol. We were,
however, able to ask questions about marijuana use and the
use of prescription drugs for anxiety or depression. Thus, our
results could have been confounded by the unmeasured use
of cocaine and heroin. Again, we do not believe this to be the
case because the use of these drugs tends to be infrequent
among contemporary workers, compared to a decade earlier,
and the relationship between alcohol use and work perfor-
mance in our study was robust.
We also acknowledge that our measure of on-the-job
drinking included drinking at company sponsored events,
some of which could have involved nonwork situations (e.g.,
a weekend company picnic). We could not distinguish these
268 JOURNAL OF STUDIES ON ALCOHOL / MARCH 1999
events. However, to the extent that they were included in our
data, they would only tend to attenuate the relationship be-
tween drinking and most of our performance measures.
We should also point out that our measure of frequency of
drinking to get high or drunk included drinking on Friday and
Saturday nights, which would be less likely to affect work
performance through hangovers. Therefore, the relat!onship
that we observed is weakened by this measurement issue.
Finally, we should note that our dependent variable has a
relatively low coefficient alpha of reliability, because the va-
riety of job performance indicators that we used do not •nec-
essarily correlate together. We did not have other measures
of reliability such as test-retest reliability as part of this study.
We justified using an overall indicator because the totality of
these problems were a legitimate concern from a corpora-
tion's perspective. However, the strength of the association
between this overall indicator and drinking behaviors may be
weakened because of this.
Because our study was cross-sectional, we cannot disen-
tangle the causal order of the relationship between drinking
and work performance problems. Several hypotheses are
evident: (1) drinking can affect work performance, whether
it be drinking on the job or drinking during off-work hours;
(2) workers who, for some reason other than alcohol use,
have problems at work, drink in response to having perfor-
mance problems; or (3) some third factor confounds the re-
lationship and is causal for both drinking and performance
problems.
Although it is generally accepted that drinking on the job
can impair workplace functioning, drinking during lunch or
at company-sponsored events is often tolerated. Moreover,
there is little research on the residual effects of off-work
drinking on following-day work performanc e . There is some
evidence of such effects, however, and this relationship war-
rants further investigation. Studies of industrial work perfor-
mance (Wolkenberg et al., 1975), driving performance
(Tomros and Laurel, 1991), cognitive and physical function-
ing (Earleywine, 1993; Karvinen et al., 1962; Lemon et al.,
1993; Roehrs et al., 1991; Seppala et al., 1976; Takala et al.,
1958), aircraft handling performance (Morrow et al., 1991;
Yesavage and Leirer, 1986) and sleeping on the job (Ames
et al., 1997) suggest that heavy drinking can contribute to im-
pairment the day after alcohol consumption, even when
blood alcohol concentration is zero and even when subjects
report no ill effects of previous drinking. In contrast,
Streufert et al. (1995) found that prior night's intoxication
did not compromise managerial decision-making ability but
did produce physical discomfort.
With respect to the second hypothesis, our evidence is in-
direct. We reported herein on work performance but not on
working •on•tions. There is, however, a large body of liter-
ature suggesting that unfavorable work conditions lead to a
variety of untoward effects for workers, including compro-
mised health status (Kasl and Amick, 1995; Janes and Ames,
1992) and alcohol use (Cooper et al., 1990; Crum et al., 1995;
Greenberg and Grunberg, 1995) and misuse (Parker and
Farmer, 1990). Work conditions likely to provoke drinking
include social conflict or isolation (Richman, 1992) and the
absence of occupational self-direction in work (Parker and
Brody, 1982). In addition, work conditions can provide a
context in which drinking is facilitated, or not discouraged,
by the absence of supervisory control (Ames, 1990; Trice
and Roman, 1972). Thus, if poor working conditions are hy-
pothesized to cause substance abuse, by extrapolation it is
possible that having the kinds of workplace problems we
measured could also contribute to heavy drinking.
Another possible interpretation of the observed associa-
tion between drinking and work performance problems is
that both are results of a deviant lifestyle syndrome (Nor-
mand et al., 1994). From this perspective, persons who have
deviant lifestyles would be likely to break rules, abuse alco-
hol and other drugs, hold deviant attitudes, and engage in
other behaviors that are contrary to social norms. Several
studies have shown evidence for this point of view (Donovan
and Jessor, 1985; Gillmore et al., 1991; Newcomb and
Bentier, 1988). For instance, Newcomb and Bentier found
that 6% of the variance in on-the-job drinking was accounted
for by deviance variables, compared to 20% by cocaine use
at work and 36% by the use of other hard drugs at work. We
do not have clear measures of deviant lifestyles in our study,
although we did control for use of marijuana and still found
independent and significant effects for both drinking at work
and away from work. However, in an analysis not shown, use
of marijuana was a significant correlate of alcohol use at
work and this result would be consistent with the notion that
lifestyle deviance is a contributor to both drinking and work-
day performance problems. In other words, we have found
evidence in our study that supports both the notion that de-
viant lifestyle contributes to drinking at work and low work
performance, and that drinking at work (as well as heavy
drinking away from work) independently contributes to per-
formance problems.
It is likely that the association we found between the
drinking-level categories and work performance represents
the joint effect of all three of the hypotheses arrayed above.
Further research with different (prospective) study designs
will be necessary to untangle the relative contribution of
each. While further research may help elucidate these rela-
tionships, our findings suggest that current corporate alcohol
policies may be deficient in several regards. First, these poli-
cies tend to focus on drinking at the workplace, as if only
immediate alcohol exposure were of significance to work
performance. Second, workplace interventions tend to focus
on case-finding and treatment for problem drinking, as if
problem drinkers were the sole source of alcohol-related
work performance problems. By investing exclusively in on-
site drinking among workers with acute alcohol problems,
these policies: (1) utilize the most resource intensive (and ex-
pensive) types of interventions (e.g., inpatient care for alco-
hol abuse); (2) focus resources on workers who are least apt
MANGIONE ET AL. 269
to be able to alter their drinking patterns; and (3) ignore the
potential contribution of moderate and moderate-heavy
drinking to work performance problems (Morawski et al.,
1991).
Some corporations do not have employee assistance pro-
grams, or their equivalent, to aid workers who are affected by
substance abuse. Our findings provide compelling evidence
that employees who are heavy drinkers are most likely to ex-
perience unproductive workdays. Thus, it is in the com-
pany's interest to provide, and facilitate access to, substance
abuse services.
Some companies do have employee assistance programs
but the focus of these services is limited to case-finding and
treatment for problem drinkers. Our findings suggest that
drinking affects work performance even among workers
without drinking problems. This includes workers who drink
at times when work activities will soon follow (e.g., lunch)
and workers who drink moderate-heavy amounts during
nonworking hours, presumably at home during evenings
prior to the workday. Accordingly, there may be benefits for
corporations to also focus on nonproblem drinkers whose
drinking may still affect their work performance. These
workers account for a large proportion of the unproductive
workdays reported by our subjects. These moderate and
moderate-heavy drinkers may, in fact, be more responsive to
company interventions than problem drinkers by virtue of
the fact that they are less dependent on alcohol. A similar ob-
servation was recently reported by Cherpitel et al. (1995)
who studied alcohol exposure among injured subjects. These
investigators concluded that the risk for injury may be in-
creased at relatively low levels of consumption and that pre-
ventive efforts aimed at more moderate drinkers may have a
greater impact on reductions of alcohol-related injuries than
efforts aimed at heavier drinkers who are fewer in number
(Cherpitel et al., 1995).
In addition to employee assistance programs, there are two
primary ways by which corporations can address employee
drinking behaviors: through promotion of education on the
association between drinking and work performance prob-
lems, and through delineation of clear policies. The public
health education and prevention approach is consistent with
a corporation's interests in promoting employee wellness in
several domains, including blood pressure control, choles-
terol management, weight control, exercise, nutrition, smok-
ing, seatbelt use, and drinking and driving. All of these health
goals represent intersections of worker and corporate inter-
ests. This approach would be consistent with findings of other
investigators, which have suggested the benefit of workplace
alcohol interventions focusing on prevention of drinking
problems through changing workplace drinking norms
(Ames, 1990, 1993; Cook and Youngblood, 1990; Delaney
and Ames, 1995; Towers et al., 1994).
Policy is another way in which companies can clarify and
communicate risks and expectations for behaviors. Some-
times policies may explicitly prohibit any drinking in specific
situations. However, many times corporate policy is silent on
drinking in a wide array of circumstances. We believe that
companies should also consider including in their policies
guidance and recommendations to employees about circum-
stances where light or moderate drinking may be more pru-
dent than heavier drinking for those employees who drink.
For instance, many companies (but by no means all) have
policies prohibiting drinking during the workday on the
worksite. Nevertheless, there are specific situations that are
often excluded, or go unmentioned, in these policies. Few
companies advise that employees should not drink immedi-
ately prior to work, although companies may have policies
regarding intoxication on the job. Few companies explicitly
prohibit drinking at lunch or even set standards for modera-
tion concerning this practice. Few companies have policies
specifically addressing the dangers of heavy drinking after
the workday, even if this involves workers driving home un-
der the influence of alcohol. Few companies have policies
addressing the prudence of avoiding heavy drinking while
traveling on company business even though workers on
the road may drink more heavily than they do at home. Some
companies are not explicit about driving a company vehicle
and drinking while on company business. While not all of
these circumstances require bans on alcohol consumption
(some do), companies should provide more explicit guidance
to their employees about the risks of alcohol for injury and
workplace problems. Corporate policies can also have an im-
portant educational effect and as such may promote amoi•g
workers behavioral norms (Ames and Janes, 1992) that are
of benefit to both workers and the corporation's bot/.om line.
Our study adds to the growing body of literature that
points both to the risks to work performance of drinl•ing im-
mediately before or during the work shift and to the risks of
heavy drinking that results in hangovers. Also, while not de-
finitive, our data suggest that corporate interest in employee
drinking should extend beyond focusing exclusively on the
alcoholic or problem drinker.
Acknowledgment
The authors gratefully acknowledge the following for their assistance in
this study: Diana Chapman Walsh, Lois Beiner, Gary Gregg, Karen
Kuhlthau and Virginia Chomitz. We would also like to make special men-
tion of our co-author Sol Levine, who died in 1997. He was the epitome of
what a mentor should be. It was his life's avocation and he was superb at it.
We will sorely miss his guidance.
References
AMES, G.M. The workplace as an enabling environment for alcohol prob-
lems. Anthropol. Work Rev. 11: 12-16, 1990.
AMES, G.M. Research and strategies for the primary prevention of work-
place alcohol problems. Alcohol Hlth Res. World 17: 19-27, 1993.
AMES, G.M., GRUBE, J.W. ^Nn MOORE, R.S. The relationship of drinking
and hangovers to workplace problems: An empirical study. J. Stud. Al-
cohol 58: 37-47, 1997.
270 JOURNAL OF STUDIES ON ALCOHOL / MARCH 1999
AMES, G.M. AND JANES, C. A cultural approach to conceptualizing alcohol
and the workplace. Alcohol Hlth Res. World 16:112-119, 1992.
ARCHER, J. Occupational alcoholism: A review of issues and a guide to the
literature. In: SCHRAM, C.J. (Ed.) Alcoholism and Its Treatment in In-
dustry, Baltimore, MD: Johns Hopkins Univ. Press, 1977, pp. 2-28.
BEAUMONT, P.B. AND HYMAN, J. The work performance indicators of prob-
lem drinking: Some British evidence. J. Occupat. Behav. 8: 54-62, 1987.
BLUM, m.c., ROMAN, P.M. AND MARTIN, J.K. Alcohol consumption and
work performance. J. Stud. Alcohol 54: 61-70, 1993.
BROOKE, P.P., JR. AND PRICE J.L. The determinants of employee absen-
teeism: An empirical test of a causal model. J. Occupat. Psychol. 62:
1-19, 1989.
CAHALAN, D., elSIN, I.H. AND CROSSLEY, H.M. American Drinking
Practices: A National Study of Drinking Behavior and Attitudes, Rut-
gers Center of Alcohol Studies, Monograph No. 6, New Brunswick,
N.J., 1969.
CHERPITEL, C.J., TAM, T., MIDANIK, L., CAETANO, R. AND GREENFIELD, T.
Alcohol and non-fatal injury in the U.S. general population: A risk func-
tion analysis. Accid. Anal. Prey. 27:651-66 l, 1995.
COOK, R.F. AND YOUNGBLOOD, A. Preventing substance ab•se as an in-
tegral part of worksite health promotion. Occupat. Med. 5: 725-
738, 1990.
COOPER, M.L., RUSSELL, M. AND FRONE, M.R. Work stress and alcohol
effects: A test of stress-induced drinking. J. Hlth Social Behav. 31:
260-276, 1990.
CRUM, R.M., MUNTANER, C., EATON, W.W. AND ANTHONY, J.C. Occupa-
tional stress and the risk of alcohol abuse and dependence. Alcsm Clin.
Exp. Res. 19: 647-655, 1995.
DELANEY, W.P. AND AMES, O. Work team attitudes, drinking norms, and
workplace drinking. J. Drug Issues 25: 275-290, 1995.
DONOVAN, J.E. AND JESSOR, R. Structure of problem behavior in adoles-
cence and young adulthood. J. Cons. Clin. Psychol. 53: 890-904, 1985.
EARLEYWINE, M. Hangover moderates the association between personality
and drinking problems. Addict. Behav. 18: 291-297, 1993.
GILLMORE, M.R., HAWKINS, J.D., CATALANO, R.F., JR., DAY, L.E., MOORE,
M. AND ABBOTT, R. Structure of problem behaviors in preadolescence.
J. Cons. Clin. Psychol. 59: 499-506, 1991.
GREENBERG, E.S. AND GRUNBERG, L. Work alienation and problem alco-
hol behavior. J. Hlth Social Behav. 36: 83-102, 1995.
HARWOOD, J.J., KRISTIANSEN, P. AND RACHAL, J.V. Social and Economic
Costs of Alcohol Abuse and Alcoholism, Issue Report No. 2, Research
Triangle Park, NC: Research Triangle Institute, 1985.
JANES, C.R. AND AMES, G.M. Ethnographic explanations for the clustering
of attendance, injury, and health problems in a heavy machinery assem-
bly plant. J. Occupat. Med. 34: 993-1003, 1992.
KARVINEN, E., MIETTINEN, M. AND AHLMAN, K. Physical performance
during hangover. Q. J. Stud. Alcohol 23: 208-215, 1962.
KASL, S. AND AMICK, B. Work stress. In: MCDONALD, J.C. (Ed.) The Epi-
demiology of Work Related Disesases, London, UK: BMJ Press, 1995.
LEIGH, J.P. Alcohol abuse and job hazards. J. Safety Res. 27: 17-32, 1996.
LEMON, J., CriESHER, G., FOX, A., GREELEY, J. AND NABKE, C. Investiga-
tion of the "hangover" effects of an acute dose of alcohol on psychomo-
tot performance. Alcsm Clin. Exp. Res. 17: 665-668, 1993.
LENDER, M.E. AND MARTIN, J.K. Drinking in America: A History, New
York: Free Press, 1987.
MAXWELL, M.A. Early identification of problem drinkers in industry. Q. J.
Stud. Alcohol 21: 655-678, 1960.
MAXWELL, M.A. Alcoholic employees: Behavior changes and occupa-
tional alcoholism programs. Alcoholism 8: 174-180, 1972.
MAYFIELD, D., MCLEOD, G. AND HALL, P. The CAGE questionnaire: Val-
idation of a new alcoholism screening instrument. Amer. J. Psychiat.
131: 1121-1123, 1974.
MORAWSKI, J., MOSKALEWlCZ, J. AND WALD, I. Economic costs of alcohol
abuse, with special emphasis on productivity. In: AASLAND, O.G. (Ed.)
The Negative Social Consequences of Alcohol Use, Oslo, Norway: Nor-
wegian Ministry of Health and Social Affairs, 1991, pp. 95-128.
MORROW, D., LEIRER, V., YESAVAGE, J. AND TINKLENBERG, J. Alcohol, age,
and piloting: Judgment, mood, and actual performance. Int. J. Addict.
26: 669-683, 1991.
NEWCOMB, M.D. AND BENTLER, P.M. ConseqUences of Adolescent Drug
Use: Impact on the Lives of Young Adults, Thousand Oaks, CA: Sage
PubEs., Inc., 1988.
NORMAND, J., LEMPERT, R.O. AND O'BRIEN, C.P. (Eds.) Under the Influ-
ence? Drugs and the American Work Force, Washington, DC: National
Academy Press, 1994.
OBSERVER AND MAXWELL, M.A. A study of absenteeism, accidents and
sickness payment in problem drinkers in one industry. Q. J. Stud. Alco-
hol 20: 302-312, 1959.
PARKER, D.A. AND BRODY, J.A. Risk factors for alcoholism and alcohol prob-
lems among employed women and men. In: NATIONAL INSTITUTE ON AL-
COHOL ABUSE AND ALCOHOLISM. Occupational Alcoholism: A Review of
Research Issues, Research Monograph No. 8, DHHS Publication No.
(ADM) 82-1184, Washington, De: Government Printing Office, 1982.
PARKER, D.A. AND FARMER, G.C. Employed adults at risk for diminished
self-control over alcohol use: The alienated, the burned out, and the un-
challenged. In ROMAN, P.M. (Ed.) Alcohol Problem Intervention in the
Workplace: Employee Assistance Programs and Strategic Alternatives,
New York: Quorum Books, 1990, pp. 27-43.
PERRINE, M.W. Alcohol influences on driving-related behavior: A critical
review of laboratory studies of neurophysiological, neuromuscular, and
sensory activity. J. Safety Res. 5(3): 165-184, 1973.
RICHMAN, J.A. Occupational stress, psychological vulnerability and alco-
hol-related problems over time in future physicians. Alcsm Clin. Exp.
Res. 16: 166-171, 1992.
ROEHRS, m., YOON, J. AND ROTH, m. Nocturnal and next-day effects of
ethanol and basal level of sleepiness. Human Psychopharmacol. 6:
307-311, 1991.
SEPPALA, T., LEINO, T., LINNOILA, M., HUTTUNEN, M. AND YLIKAHRI, R.
Effects of hangover on psychomotor skills related to driving: Modifica-
tion by fructose and glucose. Acta Pharmacol. Toxicol. 38:209-218,1976.
STREUFERT, S., POGASH, R., BRAIG, D., GINGRICH, D., KANTNER, A., LAN-
DIS, R., LONARDI, L., ROACHE, J. AND SEVERS, W. Alcohol hangover and
managerial effectiveness. Alcsm Clin. Exp. Res. 19:1141-1146, 1995.
TAKALA, M., SIRO, E. AND TOIVAINEN, Y. Intellectual functions and dex-
terity during hangover: Experiments after intoxication with brandy and
with beer. Q.J. Stud. Alcohol 19: 1-29, 1958.
TORNROS, J. AND LAURELL, H. Acute and hang-over effects of alcohol on
simulated driving performance. Blutalkohol 28: 24-30, 1991.
TOWERS, A.M., K•S•CHUK, N., SYLVESTRE, M., PETERS, C. AND DOUR-
GAULT, C. A qualitative investigation of organizational issues in an al-
cohol awareness program for blue-collar workers. Amer. J. Hlth Promot.
9: 56-63, 1994.
TRICE, H.M. The job behaviors of problem drinkers. In: PITTMAN, D.J. AND
SNYDER, C.R. (Eds.) Society, Culture and Drinking Patterns, New York:
John Wiley & Sons, Inc., 1962, pp. 493-510.
TRICE, H.M. Alcoholic employees: A comparison of psychotic, neurotic,
and "normal" personnel. J. Occupat. Med. 7: 94-99, 1965a.
TRICE, H.M. Reaction of supervisors to emotionally disturbed employees:
A study of deviation in a work environment. J. Occupat. Med. 7: 177-
188, 1965b.
TRICE, H.M. AND ROMAN, R.M. Spirits and Demons at Work: Alcohol and
Other Drugs on the Job, Ithaca, NY: New York State School of Indus-
trial and Labor Relations, Cornell University, 1972.
WECHSLER, H., DOTBALL, O.W., DAVENPORT, A. AND RIMM, E.B. A
gender-specific measure of binge drinking among college students.
Amer. J. Publ. Hlth 85: 982-985, 1995.
WOLKENBERG, R.C., GOLD, C. AND TICHAUER, E.R. Delayed effects of
acute alcohol intoxication on performance with reference to work safety.
J. Safety Res. 7(3): 104-118, 1975.
YESAVAGE, J.A. AND LEIRER, V.O. Hangover effects on aircraft pilots 14
hours after alcohol ingestion: A preliminary report. Amer. J. Psychiat.
143: 1546-1550, 1986.
