The Journal of Consumer Research (JCR) reached forty years old in 2015. To help the celebrations we conducted an analysis of the topics featured in the journal over the years. For many journals, you can use the keywords supplied by the authors. The downside of this is that authors may use fashionable words wanting to associate themselves with the latest trends. Your analysis may then fail to pick up commonalities between older and newer research. We faced a more insurmountable problem using author-supplied keywords — JCR didn’t collect these keywords until very recently. So how could we go about understanding the topics in consumer research?
Latent Dirichlet Allocation (LDA)
We used a Latent Dirichlet Allocation (LDA) procedure. This is a topic modeling process that determines what percentage of a text relates to which topics. After downloading all JCR abstracts in the first 40 years we looked for recurring themes. We determined that 16 topics described JCR well. More topics will always describe data better. Still, you want to limit the number to make the analysis less complex. We had to decide what was enough topics to describe the data well but not so many that it got too confusing to explain.
Applying the LDA procedure allowed us to describe each abstract in terms of the topics. E.g., this is 70% about topic 1 and 30% about topic 2. The topics aren’t named by the algorithm but the researcher gets a list of words most associated with each topic. Thus, we could name the topics from our knowledge of the field.
Our topic most associated with the words social, group and identity we named social identity and influence given the words most heavily associated with it.
Wang, Bendle, Mai, and Cotte 2015
Understanding The Topics In Consumer Research
This procedure allowed us to model the rise and fall of JCR topics over its first forty years. It doesn’t allow researchers to know what the hot topics will be in the future. Still, you can probably make some pretty good guesses based on past trends. We now know a lot more about what makes for a great JCR article.
A key point is that big data techniques can allow you to analyze massive amounts of data and draw out insights that no individual could spot on their own. It is interesting stuff with massive implications for firms as well as academics.
For more on topic modeling see here and here.
Read: Xin Wang, Neil Bendle, Feng Mai, and June Cotte (2015), The Journal of Consumer Research at Forty: A Historical Analysis, June, 42 (1), pages 5-18
