Replication Data for: Wait For Free: A Consumption-Decelerating Promotion for Serialized Digital Media

Below are the descriptions for each of the two main datasets used for analyses: Comic-level Dataset and Episode-level Datasets. Comic-level Datasets Table 5 results pertain to the comic-level (i.e., aggregated across all episodes within each comic) response to the WFF promotion. There are two mostly similar datasets (“free_dataset” and “paid_dataset”) that are identical except for the outcome variables (“n_views” below). Each record is at the comic-week level. The variables included in these datasets are the following: n_views: the log number of total free (“free_dataset”) or paid (“paid_dataset”) views (+1) for all the episodes included in the comic series on a given week comic_treated: a binary variable indicating whether comic was treated (i.e., under WFF) at the given week, equivalent to 1{c∈WFF}⋅{t∈post WFF_c} in page 12 of the manuscript Discount: log number of days (+1) during the week on which the “Discount” promotion was offered for the comic Free: log number of days (+1) during the week on which the “More Free Episodes” promotion was offered for the comic Episode-level Dataset Figure 5 results pertain to the much more detailed episode-level response to the WFF promotion. More specifically, the left panel results are coefficient estimates and standard errors for the “treated_weeks” variable and the right panel results are those for the “episode_treated_weeks”. Note that in contrast to the comic-level models that were run separately for the free and paid viewership, they were run jointly for the episode-level models, which means that there are two copies of columns with the same name, but with identifiers in the end (“.x” for free model and “.y” for paid model). The variables included in this dataset are the following: Outcome variable: outcome: the log number of free or paid views (+1) of each episode on a given week Treatment-effect-related variables: treated_weeks_w: a binary variable indicating whether it was the w-th week since the comic was treated (i.e., first became subject to WFF), equivalent to WFF_ctw^comic on pages 22 and 23 episode_treated_weeks_w’: a binary variable indicating whether it was the w’-th week since the episode was treated (i.e., first available for free for those following the WFF schedule maximally), equivalent to WFF_(ctw^')^episode on pages 22 and 23 genres_{adult, romance, bl, drama, school, day, fantasy, mystery}: observed heterogeneity in treatment effects relative to the baseline “others” genres. Comics that belong to 2 genres are considered to belong half to each genre. treated:{episode_number3, popularity, conversion_rate, weeks_since_completed, bingibility}: observed heterogeneity in treatment effects according to comic characteristics as defined in Table 3 on pages 13 and 14 of the manuscript treated:similarity_sum_treated: The sum of similarity scores with all the WFF comics treated up to that week, equivalent to Sim.〖WFF〗_cet on page 25 of the manuscript Control variables: lagged_Free(Paid)_Views2: the lagged outcome variable for the episode from the previous week Discount and Free: Non-WFF promotional variables defined in the same way as for the comic-level dataset, also related are first_week_Free, end_week_Free, first_week_Discount, and end_week_Discount that indicate the first and last week of each type of promotion similarity_sum_raw: The sum of similarity scores with all the WFF comics treated up to that week, calculated only for non-WFF episodes to account for spillover effects, equivalent to Sim.NonWFF_cet on page 25 of the manuscript recent_episode, new_episode, episode_how_old, not_really_treated, n_episodes, price, episode_number3 over_predict, under_predict: indicator variables for outliers, as described in Figure B7 of the Online Appendix