Once a document-feature matrix
exists, a lexicon is a human-written map from word counts to a tone number. The map is the measurement. A list built
for psychology and sociology is a different measurement from a list built by reading how words are used in 10-Ks. This
note specifies the Loughran–McDonald lists, the two weighting schemes they actually test, and the file-date evidence
that separates a tonal association from a trading rule.
Learning objectives. By the end of this note you should be able to:
- state what fraction of Harvard negative counts in the 1994–2008 10-K sample are words that are not negative in a financial context;
- name the six Loughran–McDonald categories and the negation rule that applies only to the positive list;
- compute a proportional tone score and the tf.idf weight in their equation (1);
- separate a file-date association from the one-year long–short result.
In this lesson
- Why a general negative list fails on 10-Ks
- The six lists
- Two implementations of the same list
- What the file-date tests say
- When the sign flips
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Text as Data and NLP for Finance syllabus.