Bullying

ORIGINAL ARTICLE
A useful evaluation design, and effects of the Olweus
Bullying Prevention Program
DAN OLWEUS
Research Centre for Health Promotion (HEMIL), University of Bergen, Norway
Abstract
The article presents the logic and other characteristics of an ‘‘extended selection cohorts’’ quasiexperimental
design. Possible threats to the validity of conclusions based on this kind of designare
discussed. It is concluded that chances are good that conclusions about the effects or non-effects of
school-based intervention programs will be roughly correct in most cases. The design may be
particularly useful in studies where it is not possible or desirable to use a random selection of ‘‘control
schools’’ and it should be of value to both practitioners and researchers. The design isillustrated with
a study in which three consecutive cohorts of students (n approximately 21 000) were administered
the Bully/Victim Questionnaire before and after some 8 months of intervention with the Olweus
Bullying Prevention Program (OBPP). Results indicated quite substantial reductions (by 32
/49%) in
bully/victim problems. The ‘‘time-series’’ nature of the data showed convincingly that a‘‘history
interpretation’’ of the findings (Cook & Campbell, Quasi-experimentation . Chicago: Rand McNally,
1979) is very unlikely. The data in this project were obtained in the context of a government-funded
new national initiative against bullying in Norway. The characteristics of this initiative and the model
used in implementing the program in more than 450 schools were briefly described.Keywords: Selection cohorts design, evaluation, prevention, bully, victim
Introduction
A researcher or practitioner who is interested in evaluating the effects of an intervention
program, for example against bully/victim problems in school, is very often faced with a
situation where it is not possible or desirable to use a traditional experimental design. This
means that the observational unitssuch as students or classes/schools are not randomly
assigned to the various treatment conditions (e.g. intervention versus no intervention/
control, or various degrees of intervention versus no intervention/control). In such
situations, the investigator usually must turn to what is called a quasi-experimental design.
How can the investigator then evaluate the effects of an intervention in areasonably
rigorous way?
Correspondence: Dan Olweus, Research Centre for Health Promotion (HEMIL), University of Bergen, Christies
Gate 13, N-5015 Bergen, Norway. E-mail: Olweus@psyhp.uib.no and bvpolweus@c2i.net.
ISSN 1068-316X print/ISSN 1477-2744 online # 2005 Taylor & Francis
DOI: 10.1080/10683160500255471
Psychology, Crime & Law, December 2005; 11(4): 389
/402
There is a large literatureon various quasi-experimental designs (see, e.g. Cook &
Campbell, 1979; Shadish, Cook, & Campbell, 2002, and a number of standard textbooks in
design and statistics) the strength and weaknesses of which will not be discussed in the
present context. However, here I will focus brief attention on one particular design which I
have found particularly useful and which is relatively easy to use,also for investigators who
are not primarily researchers. The general structure of this design, sometimes called a
selection cohorts design is described in Cook and Campbell’s classical book (1979) and some
textbook treatments (e.g. Judd & Kenny, 1981, under the name of ‘‘age cohort design’’).
However, one does not see many examples of it in the literature, in particular not the
‘‘extended’’version of the design (below) which I recommend and have used in several
intervention studies. Important aspects of this variant of the design are that it consists of
several adjacent or contiguous cohorts and that there is a 1-year (or possibly 2-year) interval
between measurement occasions.
A concrete illustration
I will start by giving a brief description of this extended version as it...