Framing and tone-of-voice of disaster media coverage: The aftermath of the Enschede fireworks disaster in the Netherlands

Abstract

In 2000 disaster struck Enschede in The Netherlands. Due to explosions at a fireworks facility 22 people were killed. This study aims to describe the developments in the media coverage of this disaster from a media amplification perspective. Media amplification was assumed to have taken place in case: (1) events in the aftermath of the disaster resulted in increased media coverage, (2) the media framed the disaster in terms of conflict and responsibility and (3) the framing was in a negative tone-of-voice regarding governmental authorities. Content analysis was performed on 3942 articles, derived from four newspapers. Three peaks in the magnitude of the coverage were distinguished. The content of the coverage during these peaks differed from the period of gradual coverage decline. During the peaks, coverage more often was framed in terms of conflict and responsibility, government was referred to more often and the tone-of-voice on government was less positive. It was concluded that the media coverage of the Enschede fireworks met the conditions of media amplification. It is discussed whether the media coverage was such that the public's risk perception might have been affected.

Keywords:

• framing
• risk
• tone-of-voice
• disaster
• media coverage
• media amplification

Notes

1. AD is a nationally available newspaper with a very high circulation in Rotterdam and surroundings.

2. An example of such an article is the life-story of a retiring architect who was not involved with the rebuilding of the destroyed neighborhood in Enschede, but made a remark about it in the light of urban development.

3. Visual inspection suggested that the decline was not monotonous and that there were fluctuations. It is tempting to attribute such fluctuations to events in the aftermath of the disaster. However, some fluctuations might be random. Inspection of the confidence intervals for the estimated number of articles gives an indication about this. If the observed number of articles falls within the confidence limits, one concludes the peak is a random fluctuation. If the observed number falls outside the confidence interval, one concludes there is a genuine and significant peak in media coverage. The analysis revealed that the decline in media coverage was steep and rather stable for the local newspaper, and steep and more volatile for the national newspapers, suggesting that for the latter the external events in the aftermath of the disaster were important.

4. Coders were four students of the Faculty of Behavioural Studies who were in their third or fourth year and who were paid for their work, and a University staff member. None of them was involved in the fireworks disaster as a victim, or suffered damages.

5. Cohen's Kappa is an index that measures interrater agreement for categorical variables. In many studies (e.g. Semetko and Valkenburg 2000), the interrater reliability is determined by calculating the percentage of agreement between the raters. However, this statistic does not take into account the agreement which occurs by chance. Cohen's Kappa overcomes this. It is the proportion of agreement that is actually observed between raters, after adjusting for the proportion of agreements that take place by chance. Cohen's Kappa ranges between -1.00 and 1.00. Landis and Koch (1977) suggest the following interpretation of the coefficient: a Kappa between .00 and .20 indicates a slight agreement between the raters, a Kappa between 0.21–0.40 a fair agreement, between 0.41–0.60 a moderate agreement, between 0.61–0.80 a substantial agreement and between 0.81 and 1.00 an almost perfect agreement between the raters.

6. Strictly speaking, the courts are not part of the national government but are independent authorities as part of the trias politica.

7. The timeline lists both the observed number of articles (per newspaper) and disaster-related events. The three-year period has been subdivided in 39 four-week periods, which allows comparisons over similar length periods. This enabled statistical analysis and guarded against fluctuations, e.g. as a result of competing events (like the Volendam Café fire) or, maybe, media hypes (Vasterman et al. 2005).

8. Oosting is the name of the chair man of the independent fact-finding Commissie Onderzoek Vuurwerkramp.

9. A two-way analysis variance was carried out. The dependent variables are all dichotomous variables. Lunney (1970) showed that ‘if the proportion of responses in the smaller response category is (…) equal to or greater than .2 and the number of degrees of freedom for the within cell variance is equal to 20 or more, it is permissible to use ANOVA when the dependent variable is dichotomous’ (p. 267). These conditions are met. Analysing our data by means of ANOVA is therefore permitted. One of the assumptions of analysis of variance is homogeneity of variance. ANOVA is robust, however, for departures from homogeneity of variance. A rule of thumb is that the ratio of largest to smallest group standard deviation should be 2:1 or less. This is the case: the highest hi/lo ratio is 1.66. As a double check, we analysed the differences between the periods by means of non-parametric statistics. In 66 of the 72 comparisons, the conclusion was the same. In case of differences, the applied Bonferroni test was more conservative in that it showed the level of significance to be between p < .01 and p < .05, whereas the Mann-Whitney U-test showed the difference to be significant at the <.001 level.