Abstract
The focus of this chapter is the research design and considerations in design that may become apparent part-way through the project. The chapter is positioned at this point in the book as most projects are not perfectly devised at the commencement and then executed with no required changes throughout the project. Instead, there is some messiness early in a project, as research designs are refined and shifted a little as the reality becomes apparent during the project duration. Therefore, this chapter is positioned near the middle of the volume to consider the earlier chapters and discussions and consequences of design decisions. It supports the reader in making decisions that may need to be adjusted or flexed during the project. These considerations include creating the sample, managing outliers, working with a short event window, and managing confounding events.
TopGenerating The Sample
We must work carefully to generate the sample, as the sample must capture the announcements of interest. Ensuring we have the announcements of interest is particularly important when searching for news sources and press releases. There are a couple of essential elements here.
First, the set of keywords used must be carefully enhanced and expanded so that it is going to capture all the different permutations and examples of the abstract phenomenon of interest. That is, we should take care to work with colleagues, friends, and managers to ensure that there is a range of different descriptions of the phenomenon so that we can capture all the examples and different forms in the announcements we find. Getting multiple perspectives is essential, and relying on managers as part of the team is vital because of the different terminology used concerning different parts of the business or different management areas. For example, an initiative to reduce capacity could relate to plant closures or the retrenchment of staffing levels.
Key Terms in this Chapter
Non-Normal Distribution: A distribution where the observations are not normally distributed. This is commonly observed as the abnormal return distributions are frequently non-normally distributed.
Meta-Event: A wider event that affects the events we are interested in within our event study. Examples include business cycles or industry-specific issues such as the Y2K issue for information systems research studies.
Confounding Event: Another event that occurs in the event window alongside the event we are interested in. As they coincide, we cannot determine which one, or whether they both, influence stock market reactions.
Influential Observation: An observation within the event study dataset that would substantially change the estimated coefficients in the model if it was removed.
Short Window: The window used to estimate the abnormal returns, usually a period of one- to three-days in length.
Nonparametric Tests: Those tests that do not have assumptions about the form of the distribution of the parameters. Therefore, they can be used with non-normal samples.
Outlier: An observation that is an unusual distance or separate from the other values in the sample. Just because it is an outlier, however, does not make it an influential observation. Conversely, an influential observation is not necessarily an outlier.