Content analysis

Content analysis is a research method in which answers are categorised into different types, and the number of each type is counted up. Content analysis can be used in many different areas. In interviews general themes can be identified and the number of times they appear can be counted. If you analyse for example diaries or other documents, you could begin by counting how many times the specific topics you are interested in are referred to.

Some times content analysis is referred to as a form of qualitative analysis. This is, however, a controversial classification, since content analysis is really a type of numerical coding. In the days before the richness of qualitative analysis was widely recognised in psychology, though, content analysis was the main technique researchers used for dealing with complex meanings.

The content analysis method consists of establishing a number of different content categories, and counting up the number of times items relevant to each of them occurs in a particular set of data.

Content analysis is really a way of using summary tables to describe qualitative data, i.e. data which don’t appear in the form of numbers but as words or other meaningful information- but in a quantitative form. However, content analysis isn’t really qualitative analysis, event though it is used with qualitative data. Instead, it is a way of converting that qualitative data into quantitative information- of describing it using numbers.

There are many types of data, which are suitable for content analysis. It could be used in a study of the use of children in television advertisements; or in an interview study investigating people’s experiences of parenthood. It could be used to identify recurrent themes in a series of paintings at an exhibition, or to analyse the content of children’s reading books, or to look at reports of football hooliganism in the media, or to make sense of observational studies of children’s playground behaviour. One particular influential content analysis conducted in the 1970s analysed a popular set of children’s books, and showed how very different and stereotypical, the roles played by boys and girls were. The analysis was useful because it highlighted issues, which had been overlooked; but the process of reducing information down to numbers meant that it lacked much of the subtlety and richness of modern forms of qualitative analysis. In spite of this, content analysis may be a useful method and it can be used for all sorts of different types of information, which is why psychologists have often used it a way of making some sense out of complex topics. One recent study on Gender role stereotyping in advertisements on two British radio stations can be found in this article.

The essence of content analysis is categorisation. A content analysis describes a set of data in terms of a set of categories, and how many examples have been counted in each category. That information is usually presented as a summary table, with the categories forming the columns, and the set of data forming the rows. The numbers, which appear in the cells of the table, are the frequencies - the result of counting up how often that category occurs in the data set.

Content analysis turns qualitative information into quantitative data.

What content analysis does, then, is to turn qualitative information into quantitative data, by converting it into numbers. In doing so, it describes the information but it also opens the way for a researcher to perform additional statistical tests on the material, if that seems appropriate. The most commonly used one is chi-square, because a content analysis gives us nominal data.

One problem in content analysis is that we most be sure that the categories that we have chosen are appropriate ones for our data. That generally means that we need to spend a lot of time examining the data and our research interests, so that we can be sure that the categories reflect what we are interested in. But if the data are very complex and meaningful- like pictures in an art exhibition, for example, using content analysis always means that some of the information is lost. This means that we must do a proper qualitative analysis if we want to retain any of that richness in the data. Content analysis is not a substitute for qualitative analysis, but it does give us a general, if rather simplistic, way to look at qualitative information.

An example of content analysis: Hacker and Swan (1992) focused on different aspect of campaign strategy in forms of television advertisements paid for by political parties as a means of ‘selling’ their candidate. Some of these ads aim to promote a candidate’s strengths while at the same time highlighting the opponent’s weaknesses. Hacker and Swan suggested that such advertisements have a stronger influence than other TV spots because they are watched by a wider cross-section of the population than political debates and are presented in simpler terms. The two researchers videotaped 17 campaign advertisements in autumn 1988 and randomly selected five from each campaign for analysis. The researchers devised a coding system by watching other advertising spots, where the focus was on ‘mutually exclusive and mutually exhaustive’. Coding units were single messages (for example, a specific isolated scene, a statement about a candidate or a scene). Each was classified in terms of the media dimension: oral, visual, written, candidate nonverbal (NV) and special effects. And each was classified in terms of 14 different message appeal categories such as positive or negative trait, nationalism, family, humanitarian interest, mission statement, or fear. All coding was assessed using inter-coder reliability (0.89), and for messages for which where there was not agreement were discussed and if no agreement could be reached, they were treated as uncodeable. The results showed a difference in the sense that the Bush campaign used significantly more positive messages than the Dukakis campaign. One other difference was that the Dukakis’ campaign emphasised the visionary appeal of the candidate. These may have been perceived as irrelevant because of the insufficient number of positive images, or it may be that many members of the electorate simply find such appeals irrelevant.

The researchers presented some of the results in a summary table where the F ratio (or variance ratio) showed the differences between mean scores of the two groups and the variation of scores within each data set in order to see if the kind of differences should be expected just from random variation or individual differences. The F ratio (or variance ratio) is a statistic that expresses the ratio between the two different types of variation in the data, and it does this by dividing the between-groups variation by the within-groups variation.

Table 1. Comparison of appeals vs. campaigns

*=p<.05

Summary tables are used in almost all kinds of quantitative research. They are ways of summarising more than one set of data, so that the similarities and differences produced by variables or factors can be seen as easily as possible. Summary tables often use measures of central tendency and measures of dispersion. The convention used for summary tables in research papers is to list the variables of the study along the left-hand side, so that they form the rows of the table, and the statistical measures along the top, so that they form the columns. If we had to draw a table that reported the study of two different teaching methods (A and B) to teach children to read, the two methods would be listed as the rows of the table, and the mean scores and standard deviations obtained from our test results at the top. Reading along a row would then say how a single method scored; reading down the ‘mean’ column would tell us whether the means of the two groups were very different; and reading down the ‘standard deviation’ column would enable us to compare the standard deviations of the two sets of scores.

Table 2. Methods of teaching reading: results from a reading accuracy test

Back to Qualitative methods

Back to homepage

Hacker, K.L. and Swan, W.O. (1992) Content analysis of the Bush and Dukakis 1988 presidential election campaign commercials. Journal of Social Behaviour and Personality, 7(3), 367-74 in Flanagan, C.(1998) Practicals for Psychology. London: Routledge.