eprintid: 34992
rev_number: 17
eprint_status: archive
userid: 3114
dir: disk0/00/03/49/92
datestamp: 2024-06-25 12:00:47
lastmod: 2024-07-17 08:25:10
status_changed: 2024-06-25 12:00:47
type: workingPaper
metadata_visibility: show
creators_name: Arndt, Sarah
title: Different Newspapers – Different Inflation Perceptions
subjects: ddc-330
divisions: i-181000
keywords: Inflation expectations, text mining, forecasting, monetary policy, LLM, ChatGPT
abstract: In this paper, I investigate how inflation signals from different types of newspapers influence household inflation expectations in Germany. Using text data and the large language model GPT-3.5-turbo-1106, I construct newspaper-specific indicators and find significant heterogeneity in their informativeness based on the genre—tabloid versus reputable sources. The tabloid’s indicator is more effective for predicting perceived inflation among low-income and lower-education households, while reputable newspapers better predict higher-income and more educated households’ expectations. Local projections reveal that tabloid sentiment shows an immediate decrease following a monetary policy shock, whereas responses from reputable newspapers are smaller and delayed. Household expectations also vary depending on the type of newspaper affected by the sentiment shock and the socioeconomic background of the household. These findings underscore the differentiated impact of media on inflation expectations across various segments of society, providing valuable insights for policymakers to tailor communication strategies effectively.
date: 2024
publisher: Universität
id_scheme: DOI
id_number: 10.11588/heidok.00034992
schriftenreihe_cluster_id: sr-3
schriftenreihe_order: 0748
ppn_swb: 1895451612
own_urn: urn:nbn:de:bsz:16-heidok-349926
language: eng
bibsort: ARNDTSARAHDIFFERENTN20240604
full_text_status: public
series: Discussion Paper Series
volume: 0748
place_of_pub: Heidelberg
pages: 45
citation:   Arndt, Sarah  (2024) Different Newspapers – Different Inflation Perceptions.  [Working paper]     
document_url: https://archiv.ub.uni-heidelberg.de/volltextserver/34992/7/Arndt_dp748_2024.pdf