and articles in national, business-focused, newspapers. By focusing on news from these
sources, we are more likely to capture new information since companies are required to
officially report price-sensitive information in a timely manner. We ignore lesser known
newspapers to focus on those with the greatest readership. We also exclude finance- and
economic-focused magazines as these are likely to refer to stale news that is more than a
week old.
The disadvantage of including newspaper articles in searches is that the majority of
information is likely to be reported events from the previous day. Most data vendors
have the ability to identify duplicate news articles to limit this issue. The advantage
therefore of including newspaper sources in our analysis is that this information may be
softer in content and contain more speculative analysis or potentially leaked informa-
tion such as asset sales or M&A speculation. Our database therefore contains a variety of
news including official releases, softer strategic information, and potentially speculative
articles.
8.2.2 Relevance of news
To ensure that news articles directly relate to a company rather than mention a company
in a non-stock-specific context, we rely on data vendors’ classifications to search for
company mentions that are either in news headlines or in the first paragraph of text
within the article. Some vendors such as RavenPack provide a relevance score to enable
users to refine their signals to company-specific information. Having decided on which
news articles to include, the second issue is to avoid double-counting news on the same
subject from multiple sources. We therefore only note if there was news on a particular
day rather than focus on how many news stories there are from different or even the
same news source for each category.
8.2.3 Classification of news
Having downloaded news, the next biggest challenge is how to classify items into
different categories. Keyword searches to analyse news means that there are potentially
hundreds of different types of event categories.
At the broadest level, corporate news can be classified as either accounting-related
‘‘hard’’ information or strategic ‘‘soft’’ news. The former consists of financial statement
information (earnings statements, trading updates) and management guidance of future
performance, while the latter includes corporate presentations, joint ventures, strategic
alliances, product launches, and broker conferences.
The key difference between these two categories is that news containing accounting-
related information can be readily adapted into analysts’ valuation models, whereas
strategic news contains more qualitative information which is useful for valuation
purposes but difficult to interpret. This means that the financial implications need to
be inferred or estimated. Accounting-related news is more likely to be on predictable
dates, enabling analysts to form an expectation of the news in advance of the announce-
ment. On the other hand, strategic information is more likely to be sporadic and
consequently takes longer for analysts and the market to decipher the information
content.
214 News and abnormal returns