ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Combining content analysis and survey data: How much precision do we need?

Nicole Podschuweit, Christine Heimprecht

Building: Law Building
Room: Breakout 10 - Law Building, Room 105
Date: 2012-07-10 01:30 PM – 03:00 PM
Last modified: 2012-06-19


For studying media effects under real conditions, a combination of content analysis and survey data is needed. Content analysis enables us to draw conclusions about the amount and character of media coverage which are supposed to affect people while surveys enable us to draw conclusions about the tendency and strength of media effects. The combined data can be more or less precise. That means that the content analysis data can be examined over all observed media or for every single one separately. The survey data can be added on an aggregate level or the specific media use of every single interviewee can be taken into account. Our question is: Does more precision lead to a better explanation of media effects?
We conducted a secondary analysis of parts of the German Longitudinal Election Study (GLES). This study provides a number of open access data sets collected since the run-up to the 2009 federal elections in Germany. For our purpose we examine how well voters' perception of political parties can be explained by the coverage of German TV newscasts. We assume that the more precise people's individual media use is taken into account, the better their perception of the parties can be explained.
We examined three levels of precision: On an aggregate level the average tenor of the coverage of all newscasts about the parties is compared to their perception among all viewers of TV news - regardless of their specific media usage. On a second level, the tenor of the single newscast is taken into account. Therefore, the content analysis data of the newscasts every individual interviewee uses is linked with his or her specific perception of the parties. This method is based on the assumption that media can only affect people who were exposed to it. On a third level, the individual media input is additionally weighted by the usage frequency of the respective newscast.
Our analyses show (1) that the tenor of TV newscasts has no distinct effect on the perception of the Christian Democrats. This is true for all levels of the analysis. (2) In case of the Social Democrats we found diverse results. Whilst on the first two levels of the analysis no effect could be found, a distinct, significant effect of media input showed when the usage frequency was taken into account additionally (third level).