Links table
Abstract and 1 Introduction 2. Data
3. Measurement of the media tilted and 3.1. Preparation and excellence
3.2. Classification of texts according to the source of the TV
3.3. Text text between newspapers and television stations and 3.4. Topic form
4. The Standard Economy Framework
4.1. Specifications of automatic variables
4.2. The instrument is the first stage and the validity
5. Results
5.1. The main results
5.2. Dumility examinations
6. Mechanisms and non -homogeneity
6.1. The content of local news versus the national or international
6.2. The Kabbieh news media is tilted by local newspapers
7. Conclusion and references
Online annexes
A. Data supplement
A.1. Newspaper articles
A.2. The alternative boycott of newspapers and A.3. Fetal the article excerpts
A.4. TV programs included peak times and A.5. Glory statistics
for. Approach, B.1. Text before processing and B.2. Bigrams is the most predictable for FNC or CNN/MSNBC
for. Human verification of the NLP model
B.4. Fox News distribution in newspapers and B.5. An example of articles from Fox News is similar
B.6. Topics from the LDA model based on newspapers
C. Approach the results
C.1. The results of the first stage and C.2. Exogeneity instrument
C.3. Faconation: The similarity of content in 1995/96
C.4. Ols results
C.5. Results of the reduced form
C.6. Sub -samples: newspaper headquarters and other provinces and C.7. Durabness: matching an alternative boycott
C.8. Durable: historical circulatory weights and C.9. Durabness: Relative blood circulation weights
C.10. Durable: absolute and relative FNC and C.11 scenes. Duble: dropping notes and assembly
C.12. Mechanisms: Language Features and topics
C.13. Mechanisms: descriptive evidence on the side of the request
C.14. Mechanisms: inclined infection and polarization
A. Data supplement
A.1. Newspaper articles
First, we provide more information about the news library. For each article, the news library provides the name of the newspaper, address, date, secondary line (if any), and (almost) the first 80 words of the article. We focus on the first 80 words, at the time of our data building (June-August 2019), allowed us to settle for this article in general. [21] Figure A.1 shows an example of the newspaper’s excerpt.
Figure A.1: An example of an excerpt of a local newspaper article
Alamida Times Star
The boycott determines ways to reduce shelter runner March 8, 2005
Can an opposition to the Alamida province allow the current opposition to allow the refugee refugees in emergency situations in hundreds of residential sites in the non -listed societies? This seems likely on Monday as the boycott planners have suggested ways that shelters cannot pass – as in monopolizing the land use game – neighbors can automatically express their approval or opposition. […]
In principle, newslibary includes about 4000 unique ports of 2005-2008. However, for many ports, there are only a few articles available: about 1500 outlets contain less than 1,000 excerpts (for a four -year collection). In all our analyzes, we are considered only outlets with more than 1000 articles. Also, many ports are not local newspapers in the sense that they cannot be set in a province (for example, “Army Communicator” or “Air & Space” magazine). Moreover, the news library often lists different versions of the same port separately. For example, “Augusta Chronicle, The (GA)”, “Augusta Chronicle, The: Web Edition Attlix (GA)”, and “Augusta Chronicle, The: Blogs (GA)” are listed separately. While our initial group covers all 2,618 ports with> 1000 articles (up to 50 million articles, see section A. We also collapse different versions of the same port (as in the example of Ugusta Cronic) to one note Because the alliance for audited media circulation is not available separately for different versions of the same port.
[21] Full articles were available on the basis of payment for each piece, which was expensive given our wide coverage in time and space.
