Discovering higher-level Patterns (E1) - Grimmer 2013

[HyunkuKwon]

Post questions about the following exemplary reading here:

Grimmer, Justin. 2013. “Appropriators not Position Takers: The Distorting Effects of Electoral Incentives on Congressional Representation.” American Journal of Political Science 57(3): 624-642.

[linghui-wu]

There are two highlights of the article I want to emphasize:

• It complemented grounded theories from previous studies with comprehensive empirical analysis and a quantitative case study: press releases on the Iraq war.
• I think the use of press releases as the data is well-suited for supporting the claim because compared to other textual resources, the formality of press releases systematically captures the arguments of the senators.

However, my critiques for this paper are as follows.

• The target research group is senators in congress. If we cannot generalize the conclusions to all legislators including representatives, the interchangeability of "representatives", "senators", "legislators" in the paper not only makes it confusing at first but also derives the conclusions questionable.
• The stem words used for categorizing press releases are minimal and overlapped, so the accuracy of topic classification is not very convincing.

[WMhYang]

This paper gives fruitful insights on the home styles of the senators and its corresponding consequences. After reading the paper, I am quite confused about how the author categorizes the topics into claim credit related and articulate position realted. It was mentioned in the paper that the categories are identified by the Bayesian statistical model. However, among the four outputs of this model, I cannot find any clue on how the categorization is done. Is it done manually by the author?

[Lesopil]

This is interesting research, and I wonder if something similar has been done for presidential elections and the president's style of messaging in general. This also seems to imply that running against an incumbent marginal legislator is particularly difficult, because a candidate without any appropriations to their name (if they are running for the first time) will have to revert to policy, thereby undermining support from part of the base that they need to win. I would again be interested to see this research extended to the campaign promises of defeated legislators.

In terms of my question, the author writes: ". The first column in Table 1 is a summary label that I constructed after reading a sample of 10–15 press releases from the topic. The second column presents a set of identifying words: words that distinguish the documents in each topic from the other topics, identified using a statistical method. The final column is the percentage of press releases assigned to each topic" (5). I have many questions about this. Are the press releases the same size? Does their size affect the percentage in any way? What if a press release is about more than 1 topic? What is the statistical method? Is there a better way to classify press releases? Surely the senators do not just dump press releases into the wild. Maybe they provide a category themselves?