A Song of Reviews and Q&As

Investigating the Role of Q&A Sections on Review Sentiment on Goodreads

User-generated content is becoming increasingly important in the buying process of consumers. In this study, the role of a Q&A section on Goodreads on review sentiment of books is studied. Based on literature, the following three hypotheses were formulated:

H1 The presence of a Q&A section has a positive effect on the average sentiment of reviews.
H2 The effect of the presence of a Q&A section on the average sentiment of reviews increases (decreases) as the number of answers on the featured question goes up (down).
H3 The effect of the presence of a Q&A section on the average sentiment of reviews increases (decreases) as the number of likes on the featured question goes up (down).

Source: The 50 great books on education by Dennis Hayes, 2014

The study uses data from Goodreads and includes data on questions, reviews, and related information. Using fixed effects regression models and a difference-in-difference analysis, the effect of having a Q&A section present is estimated. In the difference-in-difference analysis, reviews on Goodreads were considered as the treatment group while Amazon reviews were part of the control group.

The fixed effect regression models showed a negative relationship between having a Q&A and the average sentiment of reviews and difference-in-difference analysis suggested that an increase in question likes, lead to more positive reviews on Goodreads.

Running instructions

Necessities

R and R Studio Installation Guide.
For required R packages, source code files where lines starting with library indicate the required packages.
Various datasets with raw data. Contact Shrabastee Banerjee for access to the following raw data files:

1. 2864_goodreads_com_book_full.csv
2. 2864_web_archive_org_one_year_level.csv
3. qa_subset_amazon_text_merged.RDS
4. qa_subset_goodreads_text_merged.RDS
5. overlap_titles_amazon_gr.txt

Running the code

Clone the repository to your device
Set working directory to the structure of your directory by using the setwd() function in R Studio
Run the source code files for data preparation (located in the src/dataprep folder) in the following order:

1. gr_questions_cleaning.R
2. gr_reviews_cleaning.R
3. am_reviews_cleaning.R
4. preparing_gr_and_am.R
5. gr_sentiment.R
6. am_sentiment.R

Then, run the source code files for the statistics and analyses (located in the src/analysis folder) in the following order:

1. summary_statistics.R
2. regression.R

Shortcut

If the cleaning and sentiment measurement processes are not of interest, unzip sentiment.zip located in the gen/dataprep folder to gain access to the sentiment files used for the analyses in regression.R.