Refine
Has Fulltext
- no (2)
Document Type
- Article (2)
Language
- English (2)
Is part of the Bibliography
- yes (2)
Keywords
- Critics (1)
- Difference-in-difference (1)
- Propensity score matching (1)
- Reviewer (1)
- Selection effects (1)
- propensity score matching (1)
- selection bias (1)
- spam (1)
- spam filter (1)
Institute
Dealing with spam is very costly, and many organizations have tried to reduce spam-related costs by installing spam filters. Relying on modern econometric methods to reduce the selection bias of installing a spam filter, we use a unique data setting implemented at a German university to measure the costs associated with spam and the costs savings of spam filters. Our methodological framework accounts for effect heterogeneity and can be easily used to estimate the effect of other IS technologies implemented in organizations.
The majority of costs stem from the time that employees spend identifying and deleting spam, amounting to an average of approximately five minutes per employee per day. Our analysis, which accounts for selection bias, finds that the installation of a spam filter reduces these costs by roughly one third. Failing to account for the selection bias would lead to a result that suggests that installing a spam filter does not reduce working time losses.
However, cost savings only occur when the spam burden is high, indicating that spam filters do not necessarily reduce costs and are therefore no universal remedy. The analysis further shows that spam filters alone are a countermeasure against spam that exhibits only limited effectiveness because they only reduce costs by one third.
We propose a combined approach of propensity score matching with difference-in-differences methods for reducing selection biases of products being reviewed by critics. Critics' decision to review products may be driven by observable (e.g., star power) and unobservable (e.g., critics' individual preferences) factors, raising the question of reverse causality and selection biases. Our proposed approach enables to rigorously control for selection biases by observable and unobservable characteristics. We apply our methodological framework on data from the German book market and estimate the sales effect of a well-known TV critic. We identify substantial selection effects of individual critics, which result in serious underestimation of the short-term effect (up to 29 %) and the long-term effect (up to 37 %). The results emphasize the relevance of the proposed methodological framework by demonstrating that observable and unobservable factors drive selection effects.