Content uploaded by Susan Bruce
Author content
All content in this area was uploaded by Susan Bruce on Oct 14, 2016
Content may be subject to copyright.
Abstract—Social norms marketing is an intervention
designed to reduce unhealthy behaviors by correcting
misperceptions. At celebratory drinking events, a greater
percentage of students consume alcohol at a higher than normal
level. A campaign focused on minimizing hazardous drinking
behaviors, negative consequences and increasing protective
behaviors of students on a specific celebratory drinking event,
Halloween. Three hundred and fifty undergraduates at a
university responded to a survey designed to evaluate the
marketing campaign as well as to characterize student drinking
behavior and associated negative consequences. The survey also
collected perception data to inform the development of social
norms marketing messages. On average students who drank on
Halloween reported consuming 6.3 drinks, approximately 1.4
drinks more than they reported drinking on a typical weekend
night. On Halloween, males, Greeks, and underclassmen drank
more than females, non-Greeks, and upperclassmen,
respectively. While students did not overestimate the percentage
of students who drank, they did overestimate the percentage of
students who got drunk. Only 35.7% of students self-reported
getting drunk and students estimated that 44.8% got drunk. If
other gaps are identified regarding healthy behaviors, they
could be promoted through a social norms marketing campaign.
While the campaign’s analysis was limited due to small sample
sizes and unrepresentative sub-populations, it is still possible to
obtain relevant results in future campaigns by continuing the
study for multiple years, increasing campaign exposure and
increasing survey responses.
I. INTRODUCTION
LCOHOL use is a major problem at American
universities. A nationwide study found that 44% of
college students are binge drinkers and that despite an
increased prevalence of interventions, this number has not
changed significantly since 1993 [1]. Binge drinking is
defined as consuming four drinks in a row for women and
five drinks in a row for men at least once during a two-week
Manuscript received April 4, 2011. This work was supported in part by
the Virginia Alcohol and Beverage Control.
L. Montealegre is an undergraduate at the University of Virginia (email:
lm3mb@virginia.edu).
E. J. Bass is an Associate Professor in the Department of Systems and
Information Engineering at U.Va., P.O. Box 400747, 151 Engineer’s Way,
Charlottesville, VA 22904 USA (email: ejb4n@virginia.edu).
S. E. Bruce is the Director for the Gordie Center for Alcohol and
Substance Education at U.Va., 170 Rugby Road, Lower Level,
Charlottesville, VA 22903 USA (email: sbruce@virginia.edu).
H. A. Foster is ADAPT Graduate Assistant for the Gordie Center for
Alcohol and Substance Education at U.Va., 170 Rugby Road, Lower Level,
Charlottesville, VA 22903 USA (email: haf6n@virginia.edu).
period [2], [3]. College students tend to engage in binge or
hazardous drinking during specific times of the year. One
type of hazardous drinking is associated with celebratory
drinking events, which can vary across colleges and
universities [3], [4]. During these celebratory drinking
events, a greater percentage of students consume alcohol at
higher levels than normal.
Halloween is a celebratory event on many college
campuses. Most students overestimate actual prevalence of
substance use, thus leading them to consider hazardous
drinking to be the norm [6]-[11], which leads to higher rates
of alcohol abuse [7], [8], [12], [13]. Furthermore, social
norms marketing addresses hazardous drinking by correcting
misperceptions about the prevalence and acceptability of
high-risk alcohol use. [14]. Using this approach, university
peer health educators can analyze survey data related to the
celebratory drinking event. The results of such an analysis
provide the basis for creating and evaluating social norms
marketing campaigns and improving future campaigns.
In 2009 the Gordie Center for Alcohol and Substance
Education (Gordie Center) in concert with the Alcohol and
Drug Abuse Prevention Team (ADAPT) peer educators
launched their first social norms marketing campaign
targeting Halloween behaviors. The campaign consisted of
marketing items such as posters, cups, and Halloween
candies to students, as well as promoting other interventions.
All of the marketing interventions displayed messages
related to healthier drinking behaviors such as “87% of
U.Va. students usually stay with and monitor their friend if
they’ve passed out” and “3 out of 4 U.Va. students usually
use the buddy system to get home from a party.” While the
peer educators created the campaign and survey, they did not
have the tools to effectively evaluate the campaign results.
The purpose of this paper is to evaluate the impact of a
social norms marketing campaign for the Halloween
celebratory event. A data analysis has been conducted to
evaluate the following objectives:
I. To characterize and compare student drinking
behaviors by analyzing drinking patterns across
student demographics of interest
II. To evaluate the impact of the social norms
campaign
a. To characterize students’ exposure to the
campaign
b. To evaluate the effectiveness of the
campaign in reducing unhealthy drinking
behaviors
Cavman, Wonder Woman, or Too Drunk to Tell: An Evaluation of
the Effectiveness of a Halloween Social Norms Marketing Campaign
Laura-Elisa Montealegre, Ellen J. Bass, Senior Member, IEEE, Susan E. Bruce, and Holly A. Foster
A
III. To characterize and evaluate students’ protective
behaviors on Halloween
IV. To characterize and evaluate students’ drinking-
induced negative consequences
V. To illustrate the differences in students’
perceptions and the actual student population’s
drinking behaviors
VI. To give recommendations for messages to include
in the 2010 Halloween Campaign
This paper presents the methods used to evaluate and the
results of a social norms campaign targeting the Halloween
celebratory drinking event. It discusses the results and
presents recommendations.
I. METHODS
A. Sample
1,000 undergraduates (125 males and 125 females from
each of the four class years) were randomly sampled. Each
student received an email the day after Halloween 2009, with
an invitation to participate in an anonymous web-based
survey. A follow up email was sent five days later. The
survey remained open for a total of ten days.
B. Survey Instrument
The 33-question survey instrument included three
categories of questions: demographics, typical weekend
behaviors, and Halloween behaviors/experiences. There
were four demographic questions: weight in pounds for
helping to determine estimated blood alcohol content (BAC),
gender, year in school (1st – 4th), and fraternity/sorority
(Greek) affiliation.
The non-Halloween related section collected information
about drinking on a “typical weekend” (either Friday or
Saturday). Questions asked for the number of drinks and
hours over which the drinks were consumed. A Likert scale
question with options (never, rarely, sometimes usually,
always) asked students if they ever drank more than expected
because others were drinking.
The majority of the questions were Halloween-related.
Students were asked about the number of drinks and hours
over which the drinks were consumed, estimated BAC that
night, and whether the student noticed others abstaining from
drinking. There were non-drinking questions addressing
costumes (such as time spent making costumes and their
cost).
Two questions focused on protective behaviors and
another on associated negative consequences of drinking.
The types of protective behaviors analyzed were:
made sure friends were not left alone with strangers
asked friends to slow down drinking rate if drinking
excessively
intervened to stop friends from harming themselves or
others
stayed and monitored friends if they passed out
kept track of own drinks
ate beforehand
proactively set a limit on the number of drinks
drank at a rate of one drink or fewer per hour
planned to walk home in a group or arrange for alternative
transportation
alternated alcoholic and non-alcoholic beverages
carried an ID.
The types of negative consequences analyzed were:
physically injured oneself
physically injured another
behaved in ways later regretted
had memory loss
was taken advantage of sexually
engaged in sexual activity without safer-sex practices
got into trouble with the police
lost belongings.
Another question asked students whether they saw any
element of the Halloween campaign. These elements
included:
attended ADAPT presentation
received e-mail from a University administrator
received cup displaying safe Halloween behaviors
received candy displaying safe Halloween behaviors
saw flyers for Halloween campaign
Saw the Hoo View (electronic bulletin board) messages.
Three questions asked students if they had seen campaign
flyers by showing students the images shown below
(Figure1).
C. Derived Variables
Derived variables were created in order to complete the
desired data analyses. One variable was drinks per hour
(based on the self-reported drinks consumed divided by the
hours over which they were consumed).
With respect to drinking level, respondents were
categorized by drinking level. A second derived variable was
created for this purpose (low if three or fewer drinks on
Halloween, medium if four to five, and high if six or
greater).
Figure 1. Halloween 2009 campaign flyers advertising students’
healthy behaviors
The notion of risk was also measured by counting the extra
drinks one consumed on Halloween over the threshold
amount that one would drink on a typical weekend night
(number of self-reported Halloween drinks minus the number
of self-reported “Typical Weekend” drinks).
A. Data Analysis
The data were analyzed using scripts written in R version
2.11.1. Scripts addressed data cleansing, derived variable
creation, demographic analysis, and the analysis of the
campaign.
Since data were not normally distributed, Wilcoxon
signed-rank tests were used for mean comparisons. In
addition, Kruskal-Wallis analyses of variance by ranks were
used for single factor analyses with more than two factor
levels.
To characterize and compare behaviors of different
demographics, Wilcoxon signed-rank tests were used to
analyze the number of drinks and “extra drinks” for the
following groups:
Greek Affiliation
Gender
In addition, a Kruskal-Wallis analysis of variance by ranks
was used to compare the number of drinks and “extra drinks”
for following groups:
Year in school
Drinking Level (low, medium, high)
To characterize the frequency and significance of students’
protective behaviors, descriptive statistics and Chi-squared
tests of comparison were used to characterize and compare
those students who did and did not assume particular
protective behaviors.
Descriptive statistics and Chi-squared analyses were used
to compare student groups who did and did not experience
particular negative consequences.
Descriptive statistics and Chi-squared analyses were used
to compare student groups who did and did not see certain
campaign elements.
II. RESULTS
Significant results are reported using α=0.05 and trends at
α=0.10.
A. Respondents and Sub-Populations
363 students of the 1000 students originally contacted
responded to the survey. Thirteen responses were discarded
for having incomplete surveys and unreliable information
(e.g., consuming 40 drinks in less than one hour). Table I
presents basic demographic information for the analyzed
sample.
This sample includes 350 responses, of which 37.2% were
males and 62.8% females. These respondents do not match
the U.Va. undergraduate population, since U.Va. is
approximately 46% male and 54% female. The analyzed
sample includes 19.8% Greeks and 80.2% non-Greeks,
which under-represents the Greek-affiliated student
population; approximately 23% of the undergraduate
students are members of sororities or fraternities in the fall.
First-year students are also over-represented in the sample,
given that they accounted for 32.2% of responses but only
made up 21.6% of the 2009 student population.
In addition to demographic populations, the study
characterized the Halloween drinking behaviors of other sub-
populations. Table II presents the number of responses for
these different drinking sub-categories.
B. Halloween Drinkers
350 respondents provided their self-reported number of
drinks at Halloween. Table III shows overall self-reported
drinking behaviors on typical Saturday nights and during this
celebratory event.
TABLE I
SURVEY RESPONSE DEMOGRAPHICS
Demographic
Responses (percent)
Male/Female
124/209 (37.2%/62.8%)
Greek/Non-Greek
66/268 (19.8% 80.1%)
Year in School
First-Year
107 (32.2%)
Second-Year
76 (22.9%)
Third-Year
79 (23.8%)
Fourth-Year
70 (21.1%)
TABLE II
SURVEY RESPONDENTS BY DRINKING TYPE SUB-CATEGORIES
Drinker Type
Total
Percent / 350
Typical Saturday night drinkers
258
73.7
Halloween drinkers
223
63.7
Drinkers on both nights
220
62.9
Halloween celebrants
271
77.4
Non-drinker that drank on Halloween
3
8.6
Non-Drinkers
89
25.4
TABLE III
SELF-REPORTED NUMBER OF DRINKS
Drinkers Only
N
Drinks
Std. Dev.
Typical Saturday
258
4.90
2.67
Halloween
223
6.27
3.40
TABLE IV
SELF-REPORTED DRINKING ON HALLOWEEN
Demographic
N
Mean Drinks
Std. Dev.
Gender
Male
83
7.80
3.57
Female
136
5.40
2.98
Greek Affiliation
Greek
59
7.00
3.64
Non-Greek
161
6.00
3.31
Year in School
First-Year
68
6.60
3.22
Second-Year
51
6.37
3.24
Third-Year
49
6.02
3.34
Fourth-Year
51
6.08
3.97
Student drinkers self-reported consuming an average of
6.37 drinks on Halloween and an average of 4.90 drinks on a
typical Saturday night. A paired Wilcoxon signed-rank test
was used to compare drink differences on Halloween and
typical Saturday nights. It was found that the average
difference in drinks was significantly higher at Halloween by
approximately 1.3 drinks (V= 13080, p =<0.001).
Table IV provides the self-reported drinks by each
demographic category, as well as the corresponding standard
deviations. The data showed similar trends for several
groups across various demographics. Specifically, males,
Greeks, and U.Va. underclassmen, drank more than females,
non-Greeks, and upperclassmen, respectively.
Males reported consuming 7.8 drinks. A two-sided
Wilcoxon signed rank test showed that their self-reported
number of drinks was significantly higher than females’ (V=
8464, p<0.001). Men drank approximately 2.4 more drinks
on Halloween than women. In addition, descriptive statistics
show that Greeks drink on average one more drink than non-
Greeks; yet, the results of a two-sided Wilcoxon test showed
that this difference is not significant at the 0.05 level
(V=5644, p= 0.1378). First through Fourth-years self-
reported an average of 6.60, 6.37, 6.02, and 6.08 drinks,
respectively.
One way to analyze higher-risk student drinking behaviors
is by comparing the number of “extra drinks” each student
consumed by drinker type (low, medium, high). Figure 2,
shows the 95% confidence intervals of the differences in
students self-reported drinks on Halloween and a typical
Saturday night (i.e. “extra drinks”), conditioned by drinker
type.
Medium drinkers consume more extra drinks on average
than low or high drinkers (Figure 2). While medium
drinkers’ higher average number of extra drinks is not
significant at the 0.05 level, a Kruskal-Wallis test
(K=5.0235, 2 d.f., p=0.081) and Tukey’s comparison of
means (p=0.098) showed a trend at the 0.1 level. On
average, Fourth-Year students reported approximately 3
extra drinks on Halloween and First-Years reported 2.4 extra
drinks; although, drinking differences between classes was
not significant at the 0.05 level.
A. Perception vs. Reality
350 students reported whether they abstained, drank or got
drunk on Halloween; additionally, these students reported
the percent of students they perceived either drank or got
drunk on the day they celebrated Halloween. Table VI
reports the actual percentages for students that self-reported
drinking or getting drunk on Halloween, as well as the
perceptions for the entire population as well as the
perceptions of only drinkers.
TABLE VI
SELF-REPORTED DRINKING PERCEPTIONS FOR HALLOWEEN
Measure
N
Drank
%
Got Drunk
%
All Respondents
Percent that Drank on Halloween
350
66.55
35.71
Percent Perception of Others
350
63.71
43.59
Drinkers Only
Percent Perception of Others
223
64.51
44.85
Students were slightly less likely to believe that other
students drank on Halloween compared to students’ actual
behaviors. (66.6% of students self reported drinking while
students believed that only 63.7% drank and drinkers
believed only 64.5% drank). Students did overestimate the
percentage of students that got drunk; while only 35.71% of
students self-reported getting drunk, students estimated that
43.59% of students got drunk and drinkers estimated that
44.85% got drunk. This perception gap indicates that a social
norms marketing campaign may be helpful.
B. Campaign Exposure
Given limited data, it was not possible to fully evaluate the
effects of the interventions on drinking behaviors, negative
consequences and protective behaviors on Halloween. Figure
3 illustrates the differences in percent exposure that students
received for each of the campaign elements. The campaign
element with the most exposure was the University
administrator email with 84.44% exposure, followed by the
Figure 3. 95% CI of Extra Drinks on Halloween by Year in school for
Halloween Drinkers
Figure 2. 95% CI of Extra Drinks on Halloween by Saturday
Drinking Level for Halloween Drinkers
Halloween flyers with 24.17% exposure. A t-test comparing
the drinking levels of those who saw the campaign element
versus those who did not showed the effect was not
statistically significant.
A. Protective Behaviors
Another goal in this study is to characterize and evaluate
student protective behaviors and alcohol induced negative
consequences on Halloween. Figure 5 reports the percentage
of students participating in protective behaviors out of 350
survey respondents. The most popular reported behavior was
making sure that friends were not left alone (88.6%). The
other most popular protective behaviors were staying and
intervening to stop students from harming themselves
(84.87%), and monitoring students if they passed out
(83.53%). The protective behaviors with less frequency were
alternating alcoholic and non-alcoholic drinks (15.43%),
drank at a rate of one drink or fewer per hour (14.57%), and
set a limit on the number of drinks (12.86%).
B. Negative Consequences
Out of 223 drinkers 16.59% reported having memory loss,
6.73% reported losing their belongings, and 7.17%
reported behaving in ways they later regretted (Figure.6).
III. DISCUSSION
Hazardous drinking is a major concern on college
campuses across the country, especially during
celebratory events like Halloween. Because students
drink more on Halloween than they do on a typical
weekend night, it is important to ensure that
interventions are effective and target the sub-populations
at most risk.
A. Student Halloween Drinking Behaviors
In order to create targeted social norms marketing
campaigns, it is necessary to characterize and identify
student-drinking behaviors for diverse sub-populations.
Results from the 2009 Halloween survey showed that on
Halloween, males drink more on average than females and
Greeks drink more than non-Greeks. The results also showed
a trend for First-Year students drinking more on Halloween
than other undergraduates; yet in order to ensure that this
trend is statistically significant, it is necessary to collect a
higher sample size. The idea that younger students drink
greater quantities is found in other celebratory drinking
studies [1].
In addition to reporting student drinking behaviors, the
study investigated gaps in students’ perceptions of drinking
behaviors. 35.71% of students get drunk on Halloween;
however, students believe that this percentage is actually
43.59%. If other gaps are identified regarding healthy
behaviors, they could be promoted through a social norms
marketing campaign.
B. Campaign Effectiveness
It cannot be evaluated whether the Halloween campaign
was successful. Evaluation of the campaign and all of its
additional elements were limited to small sample sizes.
Analyzing the effects of the campaign on drinking,
protective behaviors, or negative consequences was
impossible due to low sample sizes. For example, it was not
possible to conduct a Chi-Square analysis for the ADAPT
presentation, as this had less than five data points. Moreover,
a test of proportions showed that exposure to individual
campaign elements was not significant. The only element
with significant exposure was the email; yet its larger
exposure did not appear to have an impact on behavior.
C. Limitations of the Sample
To properly evaluate the campaign it is necessary to
ensure a valid and representative sample. ADAPT and the
Gordie Center sampled 1,000 students (125 males and 125
females from each class). Sending more surveys could
increase the response rate, the number of usable responses
and the validity of statistical tests. Men have consistently
shown lower response rates than women, as identified by
previous Gordie Center studies [15], [16]. Thus, a targeted
survey could achieve more representative demographics and
allow for better recommendations for future Halloween
campaigns.
Figure 6. Self-reported student alcohol induced negative
consequences on Halloween
Figure 5. Self-reported student protective behaviors on Halloween
FIgure 4. Campaign Elements and Percent Exposure
D. Recommendations
While interventions should target the entire student body,
more Halloween interventions should target the male Greek
population. In addition, more statistically significant results,
and hence a more effective analysis can be achieved with a
larger sample size. If the survey is sent to more than 1,000
students the odds of having more representative results is
increased.
Future surveys should include a question to determine
students’ perceptions of how many standard drinks are
consumed on Halloween as well as perceptions of how
frequently students engage in specific protective behaviors.
These data will allow an evaluator to know each subject’s
perceptions in order to calculate a paired t-test comparison
for normative misconceptions and knowledge gaps instead of
just percentage differences.
Even if results from the 2009 survey do not indicate any
impact of individual campaign elements, previous studies
have shown that campaigns work best with other
interventions and take years of iterations [17]. It is important
to add relevant questions to fully assess the impact of
prevention campaigns in future Halloween surveys.
ACKNOWLEDGEMENT
The authors thank the students who responded to the 2009
Halloween survey. They also thank the Gordie Center for
Alcohol and Substance Education for financial support and
for access to the data collected regarding Halloween.
REFERENCES
[1] J. Turner, A. Keller, and J. Bauerle, “The longitudinal pattern of
alcohol-related injury in a college population: emergency department
data compared to self-reported data,” The American Journal of Drug
and Alcohol Abuse, vol. 36, pp. 194-198, 2010.
[2] H. Wechsler and T. F. Nelson, “Binge drinking and the American
college student: what’s five drinks?,” Psychology of Addictive
Behaviors, vol. 14 no. 4, pp. 287-291, 2001.
[3] H. Wechsler, J. E. Lee, M. Kuo, H. Lee, M. Seibring, and T. F.
Nelson, “Trends in college binge drinking during a period of
increased prevention efforts,” Journal of American College Health,
vol. 50 no. 5, pp. 203-217, 2002.
[4] T. C. Harford, H. Wechsler, and B. O. Muthen, “Alcohol-related
aggression and drinking at off-campus parties and bars: a national
study of current drinkers in college,” Journal of Studies on Alcohol,
vol. 64 no. 5, pp. 704-711, 2003.
[5] S. Jones, J. Oeltmann, T. Wilson, N. Brener, and C. Hill, “Binge
drinking among undergraduate college students in the United States:
implications for other substance use,” Journal of American College
Health, vol. 50, no. 1, pp. 33-38, 2001.
[6] Berkowitz, A.D. (1997) From reactive to proactive prevention:
Promoting an ecology of health on campus. P.C. Rivers and E.R.
Shore, Eds., Substance Abuse on Campus: A Handbook for College
and University Personnel, Westport, CT: Greenwood Press, 227, 121-
139.
[7] Guha, P.N., Bass, E. J., & Bruce, S. E. (2007). I Drink, I Get Drunk, I
Fall Down, No Problem: An Analysis of College Student Binge
Drinking and Related Decision Making Behaviors. Proceedings of
the 2007 IEEE Systems and Information Engineering Design
Symposium. Charlottesville, VA, April 27, 2007. Michael D.
DeVore, ed.
[8] Perkins, H.W. and Wechsler H. (1996). Variation in perceived college
drinking norms and its impact on alcohol abuse: A nationwide study.
Journal of Drug Issues 26(4), 961-974.
[9] Perkins, H.W., Meilman, P.W., Leichliter, J.S., Cashin, J.R., and
Presley, C.A. (1999). Misperceptions of the Norms for the Frequency
of Alcohol and Other Drug Use on College Campuses. Journal of
American College Health, 47(6), 253-258.
[10] Sher, K. J., Bartholow, B.D., Nanda, Shivani. (2001). Short- and
long-term effects of fraternity and sorority membership on heavy
drinking: a social norms perspective. Psychology of Addictive
Behaviors 15(1): 42-51.
[11] Wechsler, H. (1995). Binge Drinking on College Campuses: A New
Look at an Old Problem. Harvard School of Public Health, College
Alcohol Study. Boston, MA: Department of Health and Social
Behavior: 3-16.
[12] Agnostinelli and Miller, 1994; Guha, et al., 2007; Perkins and
Wechsler, 1996; Perkins, et al, 1999; Prentice and Miller, 1993)
which leads to higher rates of alcohol abuse (Berkowitz 1997; Perkins
and Wechsler 1996; Sher et al. 2001).
[13] Prentice, D. A., and Miller, D. T. (1993). Pluralistic ignorance and
alcohol use on campus: Some consequences of misperceiving the
social norm. Journal of Personality and Social Psychology. 64 (2)
243-256.
[14] M. P. Haines, A social norms approach to preventing binge drinking
at colleges and universities, Newton, MA: The Higher Education
Center for Alcohol and Other Drug Prevention, Education
Development Center, Inc, 1996.
[15] M. C. Purvis, M. S. Odioso, M. C. Weaver, M. H. White, E. J. Bass,
and S. E. Bruce, “Did You See a Horse at Foxfield? A Social Norms
Approach for Targeting the Negative Consequences of Binge
Drinking,” Proceedings of the 2008 IEEE Systems and Information
Engineering Design Symposium.
[16] M.H. White, M.S. Odioso, M.C. Weaver, M.C. Purvis, E.J. Bass, and
S.E. Bruce, “Horses? There are Horses at Foxfield? An Analysis of
College Student Hazardous Drinking and Related Decision Making
Behaviors” Proceedings of the 2008 IEEE Systems and Information
Engineering Design Symposium.
[17] W. DeJong, S.K. Schneider, L.G. Towvim, L.G.; et al. A multisite
randomized trial of social norms marketing campaigns to reduce
college student drinking. Journal of Studies on Alcohol 67(6):868–
79, 2006.