Testing the group polarization hypothesis by using logit models

Rodrigo, María F.; Ato, Manuel

doi:10.1002/ejsp.86

Cited by 7 publications

(2 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We conducted an analysis of polarization scores because Rodrigo and Ato ( 50 ) showed that changes in the means on an aggregate level, as in an (M)ANOVA, do not provide unequivocal evidence of the presence or absence of polarization. To assess the extent of polarization, we first determined whether an individual posttest item response showed a shift toward a higher value on the same pole as its pretest counterpart.…”

Section: Methodsmentioning

confidence: 99%

“…Another type of attitudinal change results from attitude polarization, which is the tendency to shift toward a more extreme position on the same pole of the attitude continuum. This can occur after group discussion ( 39,50 ) or by personal reflection. ( 51 ) Specifically, when group members begin their discussions by having generally similar nonneutral attitudes, any relevant, valid, and novel arguments that are shared during group discussions tend to be consistent with members’ existing attitudes, thus increasing their confidence in the correctness of these attitudes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Does Communicating (Flood) Risk Affect (Flood) Risk Perceptions? Results of a Quasi‐Experimental Study

Terpstra¹,

Lindell²,

Gutteling³

2009

Risk Analysis

170

134

View full text Add to dashboard Cite

People's risk perceptions are generally regarded as an important determinant of their decisions to adjust to natural hazards. However, few studies have evaluated how risk communication programs affect these risk perceptions. This study evaluates the effects of a small-scale flood risk communication program in the Netherlands, consisting of workshops and focus group discussions. The effects on the workshop participants' (n = 24) and focus group participants' (n = 16) flood risk perceptions were evaluated in a pretest-posttest control group (n = 40) design that focused on two mechanisms of attitude change-direct personal experience and attitude polarization. We expected that (H1) workshop participants would show greater shifts in their flood risk perceptions compared with control group participants and that (H2) focus groups would rather produce the conditions for attitude polarization (shifts toward more extreme attitudinal positions after group discussion). However, the results provide only modest support for these hypotheses, perhaps because of a mismatch between the sessions' contents and the risk perception measures. An important contribution of this study is that it examined risk perception data by both conventional tests of the mean differences and tests for attitude polarization. Moreover, the possibility that attitude polarization could cause people to confirm their preexisting (hazard) beliefs could have important implications for risk communication.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Does Communicating (Flood) Risk Affect (Flood) Risk Perceptions? Results of a Quasi‐Experimental Study

Terpstra¹,

Lindell²,

Gutteling³

2009

Risk Analysis

170

134

View full text Add to dashboard Cite

show abstract

Polarization is first and foremost a categorical phenomenon: a comment on Kiers

Rodrigo

Ato

2001

Euro J Social Psych

Self Cite

View full text Add to dashboard Cite

The main problem that we focus on in our work`Testing the group polarization hypothesis by using logit models ' (Rodrigo & Ato, 2001) is that group polarization is de®ned at the level of individual groups but empirically tested at the sample level, using some form of aggregation across groups. We refer to this approach as`parametric', based on the fact that the main procedure requires some form of averaging' pre-test and post-test measurement across groups and the application of standard statistical methods like linear regression or ANOVA. Against this traditional approach, we propose à categorical' approach (the term`nonparametric' is not used in our work), based ®rst on de®ning a categorical variable that assesses change between post-test and pre-test in terms of the individual polarization for each group.In his kind commentary to our work, Henk A. L. Kiers is in agreement with our argument that group polarization' hypothesis cannot be adequately assessed by the traditional analytical procedure based on the aggregation of positions across groups (Kiers, 2001). We very much appreciate his recognition of this problem, given that this procedure has been applied as a matter of course during the last thirty years in group polarization research.In essence, Kiers agrees with our ®rst main proposal to study changes in polarization at the individual group level. However, he disagrees with our second main proposal to do this by using a categorical variable and consequently with our proposal to analyse data with logit models. He argues that there is no reason to test the polarization hypothesis by means of a categorical variable. In addition, he points out that logit models are not easy to use and drawing conclusions from their results is dif®cult, involving some complexities for interpretation of results in standard social research. Instead, he proposes a simpler alternative based on some quantitative operationalization of`degree of polarization' that would allow us to reapply the standard parametric techniquesÐsuch as ANOVA or linear regressionÐto relate this new quantitative variable to the experimental variables. We consider that there are some critical points in Kiers' proposal that merit our consideration in this commentary.To summarize, Kiers mainly proposes the next quantitative operationalizations of`degree of polarization' (P), where X 1 denotes initial opinion values (pre-test) and X 2 ®nal opinion values

show abstract

Categorization is not the only way: a comment on Rodrigo and Ato

Kiers

2001

Euro J Social Psych

View full text Add to dashboard Cite

Rodrigo and Ato deal with the interesting problem of operationalization of the concept of group polarization. Subjects are ®rst asked to indicate their opinion (or attitude or judgment) on a particular topic on a bipolar scale. The subjects are assigned to a number of groups, and for each group the initial group opinion' on the bipolar scale is determined by simply taking the average of the opinions of the individuals in that group. Next, some form of group discussion takes place, and subjects are again asked to indicate their opinion. These opinions are again averaged within groups to determine the group opinion. Then`group polarization' is de®ned as a shift of the group opinion towards a more extreme position on the bipolar opinion scale.In research on polarization one often studies whether certain types of discussion groups will lead to more polarization than other, hence, groups are often assigned to different conditions in an experimental set-up, and the aim of the research is to compare the (degree of ) polarization between the different kinds of groups (i.e. different experimental conditions).As Rodrigo and Ato rightly emphasize and nicely demonstrate, the concept of group polarization cannot adequately be assessed by the`usual analytical procedure' based on aggregation of positions across groups. This is because group polarization may refer both to a shift from a slightly positive to a strongly positive scale value (thus pertaining to a positive shift) and to a shift from a slightly negative to a strongly negative scale value (thus pertaining to a negative shift). In principle, even if all groups show group polarization, but in half of them this pertains to a positive shift, and in the other half to a negative shift, the average over all groups may hardly change. Therefore, methods testing the group polarization hypothesis by means of procedures that average across groups are not suitable. Rodrigo and Ato interestingly refer to this as`a lack of ®t between the de®nition and the statistical analysis of the group polarization phenomenon'. Indeed, this is a nice example of a situation where standard statistical approaches may lead researchers to answering the wrong research question. In this case polarization is inadvertently studied at the sample level rather than the individual group level, and rather than group polarization, the usually uninteresting concept of`sample polarization' (polarization of the opinion averaged over all groups in each condition) is studied (see McGarty, Turner, Hogg, David & Wetherell, 1992).Rodrigo and Ato refer to the above-described approach as the`traditional parametric approach'. This is somewhat unfortunate, because the`parametric' aspect of the approach (i.e. the fact that

show abstract

Testing the group polarization hypothesis by using logit models

Cited by 7 publications

References 41 publications

Does Communicating (Flood) Risk Affect (Flood) Risk Perceptions? Results of a Quasi‐Experimental Study

Does Communicating (Flood) Risk Affect (Flood) Risk Perceptions? Results of a Quasi‐Experimental Study

Polarization is first and foremost a categorical phenomenon: a comment on Kiers

Categorization is not the only way: a comment on Rodrigo and Ato

Contact Info

Product

Resources

About