Home Topic Model M(ichae)L News Research




Michael C. Dougal

Machine Learning, Causal Inference, Social Science

Modeling Today's News M(ichae)L Recommended Stories Randomized Experiments Research

Bio

I am a Ph.D. candidate in American politics and political methodology. My dissertation develops methods of text-based machine learning to assess political coverage and better understand the ability of the media to monitor elected officials on behalf of voters. I am currently working on using text-based machine learning to provide real time analytics of media coverage across the political Internet. My other ongoing research examines changes in ideology over time.

One project draws on over 1 million individual poll respondents to measure public attitudes towards the New Deal at the state level from 1936-1952. Another project scales the text of present day campaign websites to measure whether candidates moderate their campaign message between the primary and the general election. My past research has examined whether citizens vote based on candidate appearance (they do) and whether people who donate to presidential campaigns have any close friends who donated to the opposing party’s candidate (they do not).

2016 Presidential Election Media Coverage

How did the media cover the 2016 U.S. Presidential campaign? Which topics did they focus and which stories dominated the news cycle? The final component of my dissertation examines how the media covered the 2016 presidential election.

Want to see how a website talks about a particular keyword? Try Modeling Today's News yourself.

Out of Step, but in the News?

Part of my dissertation uses text-based machine learning to examine whether local newspapers foster democratic accountability by sounding the alarm on out of step representatives and alerting otherwise inattentive voters that it is time for change. I show that challengers receive less coverage than incumbents in competitive House districts, horse race coverage displaces policy coverage, and newspapers do not sound the alarm on out of step House incumbents. Instead, newspapers provide representatives who vote against a majority of their constituents on landmark legislation with the same overwhelmingly neutral coverage as more faithful representatives. Even in congressional districts that closely correspond to newspaper markets, journalists provide out of step incumbents with milquetoast coverage.

The same machine learning approach I use to measure candidate criticism can also be used to identify interesting news stories. Check out M(ichae)L's News.

Experiencing Y(1)1