Deductive theming is a type of qualitative data categorization or coding in which your themes or codes are pre determined through your experience or existing literature.
In cases where we already have an idea of criteria from which to categorize our data (for example, searching for application of UN Development Goals, or using existing open government best practices) we may want to engage in deductive theming of unstructured data.
To analyze data deductively, follow these steps:
- Using insight gained from your previous professional experience or from existing academic or professional literature (or both) develop a list of categories that you expect to find when you collect your data.
- If data collection includes interviews, make sure your questions are designed so that it is possible to locate your categories in your data.
- Use the categories to put together a coding guide. Make sure you explain them in enough detail so that others can use them to categorize the data once it is collected.
- Often it is a good practice to include examples of how to identify the category in context.
- Collect your data.
- If possible, multiple coders should use the coding guide to code a portion of the data (~5 to 10 percent).
- At this stage coders must code the same data.
- Use an excel spreadsheet with columns for the section of text being coded and the code assigned to it.
- Then the coders meet to compare their results; this is an opportunity to refine the coding guide to make it more clear.
- If engaging in a hybrid interpretive/deductive analysis approach, this is also an opportunity to add interpretive patterns and themes to the coding guide.
- Coders return to coding and code another 10%, then meet again to discuss.
- This pattern continues 10% at a time until coders are mainly in agreement – i.e. getting the same results, then the remainder of the data can be coded independently.
- Once agreement is set, coders no longer have to code the same data but can divide up the remainder of the data.
- This pattern continues 10% at a time until coders are mainly in agreement – i.e. getting the same results, then the remainder of the data can be coded independently.
This video from a professor at Ontario Tech University provides a solid and concrete example of what coding can look like when you have predetermined criteria. He even shows how he sets up his excel spreadsheet to code his data (6 mins):
Test Your Knowledge:
Go to the University of West of England’s example of Qualitative Interview Analysis at the following link: Example of Qualitative Interview Analysis. Conduct either an inductive or deductive theming of the interview transcript provided and then compare your results with the ones given. What is different? What is similar? What can you learn about qualitative analysis by comparing your results to others?