Researching Media Slant: Explaining the Instrumental Variables Specification

Table of Links

Abstract and 1 Introduction 2. Data

3. Measuring Media Slant and 3.1. Text pre-processing and featurization

3.2. Classifying transcripts by TV source

3.3. Text similarity between newspapers and TV stations and 3.4. Topic model

4. Econometric Framework

4.1. Instrumental variables specification

4.2. Instrument first stage and validity

5. Results

5.1. Main results

5.2. Robustness checks

6. Mechanisms and Heterogeneity

6.1. Local vs. national or international news content

6.2. Cable news media slant polarizes local newspapers

7. Conclusion and References

Online Appendices

A. Data Appendix

A.1. Newspaper articles

A.2. Alternative county matching of newspapers and A.3. Filtering of the article snippets

A.4. Included prime-time TV shows and A.5. Summary statistics

B. Methods Appendix, B.1. Text pre-processing and B.2. Bigrams most predictive for FNC or CNN/MSNBC

B.3. Human validation of NLP model

B.4. Distribution of Fox News similarity in newspapers and B.5. Example articles by Fox News similarity

B.6. Topics from the newspaper-based LDA model

C. Results Appendix

C.1. First stage results and C.2. Instrument exogeneity

C.3. Placebo: Content similarity in 1995/96

C.4. OLS results

C.5. Reduced form results

C.6. Sub-samples: Newspaper headquarters and other counties and C.7. Robustness: Alternative county matching

C.8. Robustness: Historical circulation weights and C.9. Robustness: Relative circulation weights

C.10. Robustness: Absolute and relative FNC viewership and C.11. Robustness: Dropping observations and clustering

C.12. Mechanisms: Language features and topics

C.13. Mechanisms: Descriptive Evidence on Demand Side

C.14. Mechanisms: Slant contagion and polarization

4. Econometric Framework

4.1. Instrumental variables specification

Estimating Equation (2) using OLS likely produces biased θ estimates. Many political and economic factors may correlate with both Fox News viewership and newspapers featuring Fox-like content – any pre-existing ideological preferences of the county. Therefore, we use an instrumental variable.

where the other terms are as above.

Combining the first stage (3) and Equation (2), we can procure causal estimates for the local average treatment effect θ using two-stage least squares (2SLS). To facilitate the coefficient interpretation, we standardize the instrument, endogenous regressor, and outcome by dividing the original values by the standard deviations. Standard errors are two-way-clustered by newspaper and county or, for robustness checks, by state.

We use weighted regressions since most newspapers serve more than one county, and the circulation across counties is unevenly distributed. To account for how much each newspaper is influenced by the channel position in its associated counties, newspapercounty observations are weighted by a newspaper’s circulation in that county. [10]

[8] This specification for the treatment is different from Martin and Yurukoglu (2017) and reflects our outcome that is also measured in relative terms. Robustness checks report the specification of non-relative FNC viewership.

[9] As mentioned above, we will also report the more standard specification of just (non-relative) FNC channel position in robustness checks.

[10] We demonstrate robustness to other weighting schemes, including a variant in line with Martin and Yurukoglu (2017) which weights by the number of households in a locality surveyed by Nielsen.