FROM WORDS TO SHOCKS: Textual Analysis and Inquiry into the Explanation of DSGE-Inferred Shocks Using Monetary Policy Text in Pakistan

Ateeb Akhter Shah SYED*, Kevin Haeseung LEE** and Mohsin WAHEED***

This paper uses Natural Language Processing (NLP) techniques to discern how central bank communication in an emerging economy gets shaped during normal and crisis periods and across different governors’ tenures. Firstly, the Monetary Policy Statements (MPS) are modelled as topics using two distinct approaches: the Latent Dirichlet Allocation (LDA) and the discovery of topics through text nets. Secondly, governor regime-wise analysis is conducted. Thirdly, optimism scores are measured using three distinct techniques. We also shed light on the discussion of core and headline inflation in the MPS over time. Finally, we find that the topics we discover explain shocks to the economy obtained through a DSGE model covering the period 2005-2023.