기술 사업 및 스타트업 테크노에 대한 모든 것

미디어 경사 연구 : 신문 카운티 관찰

미디어 경사 연구 : 신문 카운티 관찰

초록 및 1 소개 2. 데이터

3. 미디어 경사 측정 및 3.1. 텍스트 사전 처리 및 기능화

3.2. TV 소스 별 전사를 분류합니다

3.3. 신문과 TV 방송국의 텍스트 유사성과 3.4. 주제 모델

4. 경제학 프레임 워크

4.1. 도구 변수 사양

4.2. 기기 1 단계 및 유효성

5. 결과

5.1. 주요 결과

5.2. 견고성 점검

6. 메커니즘과 이질성

6.1. 지역 대 국유 또는 국제 뉴스 내용

6.2. 케이블 뉴스 미디어 경사는 지역 신문을 양극화합니다

7. 결론과 참고 문헌

온라인 부록

A. 데이터 부록

A.1. 신문 기사

A.2. 신문의 대체 카운티 매칭 및 A.3. 기사 스 니펫 필터링

A.4. 프라임 타임 TV 쇼 및 A.5가 포함되었습니다. 요약 통계

B. 방법 부록, B.1. 텍스트 사전 처리 및 B.2. FNC 또는 CNN/MSNBC에 가장 예측되는 Bigram

B.3. NLP 모델의 인간 검증

B.4. 신문 및 B.5에서 폭스 뉴스 유사성 분포. Fox News 유사성의 기사

B.6. 신문 기반 LDA 모델의 주제

C. 결과 부록

c.1. 첫 번째 단계 결과 및 C.2. 악기의 외생

c.3. 위약 : 1995/96의 내용 유사성

c.4. OLS 결과

c.5. 형태 결과 감소

c.6. 하위 샘플 : 신문 본부 및 기타 카운티 및 C.7. 견고성 : 대체 카운티 매칭

c.8. 견고성 : 역사적 순환 중량 및 C.9. 견고성 : 상대 순환 중량

c.10. 견고성 : 절대 및 상대 FNC 시청률 및 C.11. 견고성 : 관찰 및 클러스터링 삭제

c.12. 메커니즘 : 언어 기능 및 주제

c.13. 메커니즘 : 수요 측면에 대한 설명 증거

c.14. 메커니즘 : 경사 전염 및 편광

c.6. 하위 샘플 : 신문 본부 및 기타 카운티

여기서, 우리는 기준 추정치를 복제하지만 카운티가 신문의 직계 소유자가 기반을 둔 위치와 일치하는 신문 카운티 관찰 만 고려합니다. 이를 위해, 우리는 지역 신문의 소유자가 감사를위한 Alliance for Audited Media의 데이터를 사용하여 미국 카운티에 기반을 둔 도시를 지정합니다 (섹션 2 참조). 우리는 직계 소유자에 중점을 둡니다. 즉, 대기업이 소유 한 신문에 대한 모회사의 위치를 ​​고려하지 않습니다. 반대로, 표 C.7에서, 우리는 본사 카운티를 제외합니다.

Notes: 2SLS estimates. Cross-section with newspaper-county-level observations weighted by newspaper circulation in each county. This Table only includes newspaper-county observations where the county coincides with the newspaper headquarters. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls aslisted in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are multiway-clustered at the county and at the newspaper level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.Notes: 2SLS estimates. Cross-section with newspaper-county-level observations weighted by newspaper circulation in each county. This Table only includes newspaper-county observations where the county coincides with the newspaper headquarters. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls aslisted in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are multiway-clustered at the county and at the newspaper level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.

Notes: 2SLS estimates. Cross-section with newspaper-county-level observations weighted by newspaper circulation in each county. This Table only includes newspaper-county observations where the county does not coincide with the newspaper headquarters. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls as listed in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are multiway-clustered at the county and at the newspaper level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.Notes: 2SLS estimates. Cross-section with newspaper-county-level observations weighted by newspaper circulation in each county. This Table only includes newspaper-county observations where the county does not coincide with the newspaper headquarters. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls as listed in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are multiway-clustered at the county and at the newspaper level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.

c.7. 견고성 : 대체 카운티 매칭

Notes: 2SLS estimates. Cross-section with newspaper-level observations weighted their total circulation. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls as listed in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are clustered at the state level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.Notes: 2SLS estimates. Cross-section with newspaper-level observations weighted their total circulation. The dependent variable is newspaper language similarity with FNC (the average probability that a snippet from a newspaper is predicted to be from FNC). The right-hand side variable of interest is instrumented FNC viewership relative to averaged CNN and MSNBC viewership. All columns include state fixed effects and demographic controls as listed in Appendix Table A.2. Column 2 also includes channel controls (population shares with access to each of the three TV channels). Column 3 controls for generic newspaper language features (vocabulary size, avg. word length, avg. sentence length, avg. article length). Standard errors are clustered at the state level (in parenthesis): * p < 0.1, ** p < 0.05, *** p < 0.01.

이 백서는 CC 4.0 라이센스에 따라 ARXIV에서 구입할 수 있습니다.

저자 :

(1) Philine Widmer, Eth Zurich 및 [email protected];

(2) Seregia, ETH ZURIC TH ZURICS 및 [email protected];

(3) Elliott Ash, ETH Zürich 및 [email protected].

출처 참조

Related Posts

Post Comment

당신은 놓쳤을 수도 있습니다