How to make causal inferences using texts-Reference-Cited by-同舟云学术

How to make causal inferences using texts

Published:2022-10-21 Issue:42 Volume:8 Page:
ISSN:2375-2548
Container-title:Science Advances
language:en
Short-container-title:Sci. Adv.

Author:

Egami Naoki¹^ORCID,Fong Christian J.²,Grimmer Justin³⁴^ORCID,Roberts Margaret E.⁵⁶^ORCID,Stewart Brandon M.⁷⁸^ORCID

Affiliation:

1. Department of Political Science, Columbia University, New York, NY 10027, USA.

2. Department of Political Science, University of Michigan, Ann Arbor, MI 48109, USA.

3. Department of Political Science, Stanford University, Stanford, CA 94305, USA.

4. Hoover Institution, Stanford University, Stanford, CA 94305, USA.

5. Department of Political Science, University of California San Diego, La Jolla, CA 92093, USA.

6. Halıcıoğlu Data Science Institute, University of California San Diego, La Jolla, CA 92093, USA.

7. Department of Sociology, Princeton University, Princeton, NJ 08544, USA.

8. Office of Population Research, Princeton University, Princeton, NJ 08544, USA.

Abstract

Text as data techniques offer a great promise: the ability to inductively discover measures that are useful for testing social science theories with large collections of text. Nearly all text-based causal inferences depend on a latent representation of the text, but we show that estimating this latent representation from the data creates underacknowledged risks: we may introduce an identification problem or overfit. To address these risks, we introduce a split-sample workflow for making rigorous causal inferences with discovered measures as treatments or outcomes. We then apply it to estimate causal effects from an experiment on immigration attitudes and a study on bureaucratic responsiveness.

Publisher

American Association for the Advancement of Science (AAAS)

Subject

Multidisciplinary

Reference57 articles.

1. Computational Methods for Predicting and Understanding Food Judgment

2. Predicting leadership perception with large-scale natural language data

3. What's Advertising Content Worth? Evidence from a Consumer Credit Marketing Field Experiment*

4. Uniting the Tribes: Using Text for Marketing Insight

5. The Elasticity of Science

Cited by 16 articles. 订阅此论文施引文献订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献

1. What's a parent to do? Measuring cultural logics of parenting with computational text analysis;Social Science Research;2024-11

2. Seeded Topic Models in Digital Archives: Analyzing Interpretations of Immigration in Swedish Newspapers, 1945–2019;Sociological Methods & Research;2024-08-22

3. The Fill-Mask Association Test (FMAT): Measuring propositions in natural language.;Journal of Personality and Social Psychology;2024-07-08

4. Natural Language Processing in Psychology;Zeitschrift für Psychologie;2024-07

5. Causal inference with latent outcomes;American Journal of Political Science;2024-06-16