Working Paper: CEPR ID: DP17931
Authors: David Byrne; Robert Goodhead; Michael McMahon; Conor Parle
Abstract: Discussions of time are central to many questions in the social sciences and to official announcements of policy. Despite the growing popularity of applying Natural Language Processing (NLP) techniques to social science research questions, before now there have been few attempts to measure expressions of time. This paper provides a methodology to measure the "third T of Text": the Time dimension. We also survey the techniques used to measure the other Ts, namely Topic and Tone. We document key stylised facts relating to temporal information in a corpus of policymaker speeches.
Keywords: Textual Analysis; Communication
JEL Codes: C55; C80; E58
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| temporal dimension of communication (C41) | policymakers' economic decisions (E60) |
| temporal references in speeches (C41) | understanding of economic policymaker communications (E60) |
| temporal references (C41) | macroeconomic understanding of economic cycles (E32) |
| temporal references (C41) | risk assessments (D80) |
| temporal references (C41) | policy evaluations (H43) |
| quantifying temporal references (C41) | insights into policymakers' communications (D72) |
| temporal references (C41) | market expectations and responses (D84) |
| frequency and orientation of temporal references (C41) | market behavior (D40) |
| speech characteristics (Z00) | macroeconomic indicators (E66) |