There are basically two kinds of content analysis: a more interpretative, qualitative version that seldom uses inter-rater reliability, and a descriptive, quantitative version that often uses inter-rater reliability. For the more qualitative version, which emphasizes counting codes, the most common number of raters (coders) is two. You can find good descriptions of that kind of coding in the books by Krippendorff, Nuendorff, and Weber.

A good answer from David - as usual. With qualitative research - the notion of Trustworthiness lends itself to 'cross-matching' to assist the audit and rigour trail. Between 2-3 coders is recommended. Any more than that is perhaps seen as 'over-kill'. The attached chapter may assist.

Dear Prof Dean Whitehead and Prof David L Morgan . Thank you so much for your helpful answer. However, if I may ask, I would like to know whether the coders' background such as gender, political/religious affiliation, age, first language, nationality etc will have a great impact on the analysis? If so, is it better to have coders from similar background?

Thank you so much for your help.

If you are doing a quantitatively oriented content analysis, the coding system is supposed to be the data collection instrument, not the researcher or the coder. This means that any coder (or pair of coders) should produce the same responses, regardless of their background.

Also, in reading articles that use this kind of analysis, I have never seen any reference to the similarity or differences in the backgrounds of the coders.

Nurusshifa - if a team of coders is involved - then there should be some 'ground rules' i.e. what would constitute 'naturally' occuring content and themes. If it is felt that 'background' characteristics of coders might bring outside subjectivity into play - then this needs to be clarified as innappropriate or the coders not used in the first place. While there will always be a degree of subjectivity in data analysis (especially qualitative) it should be 'contained' and managed as much as possible.

Dear Prof. David L Morgan and Prof. Dean Whitehead, thank you again for your insight. Coming from a linguistics field, those factors are very crucial. Thus, I am quite concerned whether those factors are important in content analysis as well. Thank you again for your clarification.

Nurusshifa Zulkifli - In qualitative research, the content analysis aims at interpreting meaning from text data. Three approaches are described in the literature for content analysis; However, the main issues are approaches to coding schemes and origins of codes. Usually, keywords of content are used to interpret the underlying context. I recommend reading textbooks on content analysis and qualitative research 

I recommend for you reading:

M. Schreier. Qualitative Content Analysis in Practice 1st Edition, SAGE, London, 2012.

I think that the background of the coder is of importance but not concerning the aspects that you ask about. The researcher performing the coding ought to be familiar with the subject in some way. I can imagine that it would be extremely hard for me as a registered nurse to code interview data from technical areas concerning for example production processes in airplane engine manufacturing or casting processes since they are way outside of my area of expertise. I can also imagine that researchers unfamiliar with the area of care and caring would have trouble if they were to code my interview material with stroke unit staff telling about ethical difficulties in acute stroke care when the patient are severely afflicted and probably dying.

My opinion is that understanding of the subject, familiarity with the concepts and so on are of importance. If I were to code interviews about a totally unfamiliar subject I think that I would miss a lot of things of importance and code things that are not meaningful according to a researcher in the area, due to my lack of understanding of the subject.

I'm wondering if the coder that you mentioned is for intercoder reliability test in content analysis? If it's for intercoder reliability test, there're usually two or more independet trained coders (Lavrakas, 2008; Riffe et al, 2005). I don't think it's possible to have only one coder for the reliability test since the test is to ensure a minimum agreement percentage among coders about categorising content. The minimum percentage of realiability itself is various depending on what reliablity formulas that you'll be using, whether it's Cohen’s Kappa, Krippendorff’s alpha, etc.

Before carrying out the intercoder reliability test, it is essential that the coders are properly trained in using the coding and categorising protocol in order to be familiar with the definitions of the protocol. Hopefully it helps.

Prof Samy A. Azer , Dr. Åsa Rejnö and Lupita Wijaya , thank you for your input. It was really helpful for my study.

Good Answer Dean Whitehead

Question: What are the qualifications for a content analysis coder. The subject of the study is beer television advertisement

Nhiecole Abejo The typical approach is for a senior investigator to create codebook, and then teach either a graduate or undergraduate assistant how to apply that codebook.