Kayser takes on new duties while delving into media bias, the predictive power of polling, and when to trust the numbers.
Mark Kayser stepped into the role of Dean of Research and Faculty at the Hertie School in February 2021. But he’s no newbie. The Professor of Applied Methods and Comparative Politics has been at the school since 2009 and is a familiar face to students and many alumni who have attended his popular core Statistics I class.
Kayser, a political scientist, is not only a resident expert on voter behaviour, his research addresses the economic underpinnings of democracy, including political accountability, the effects of electoral competition on policy-making, and economic influences on electoral preferences.
As Dean, Kayser is currently focusing on projects related to the school’s recent expansion, which includes shaping the PhD programme to give it a clear identity and direction for the future, expanding third-party funding and hiring new faculty. He’s also juggling number of research projects on media bias and election outcomes, when he’s not out walking his family’s new “pandemic dog” or teaching a recent Executive MPA course on informed consumption of quantitative data, “When to trust the numbers”.
The numbers also factor big in his fall core Statistics I class, which nearly all Hertie School students are required to take. Kayser is known for peppering his lectures with colourful examples to illustrate what numbers can do – whether serious or spurious – like the relationship between the size of a fire and the number of fire trucks at the scene, or between cheese consumption and death by entanglement in bedsheets. Kayser says his goal is for students to gain an understanding of data-driven, or fact-based, policymaking. Instead of just memorizing formulas, he wants to make sure they can apply statistics knowledge to things that impact people’s lives every day.
That’s why he takes a “notably different” approach to introductory statistics than courses usually do, he says. Rather than exclusively focusing on the underlying mathematics and progressing only up to “bivariate regression” – one independent and one dependent variable – Kayser wants students to understand “the most important workhorse of data analysis – multiple regression.” As he puts it: “…most variables in the social world are correlated with many others.” This method helps students grapple with the complex moving parts of today’s policy problems and prepares them to read quantitative reports with a critical eye in their future jobs.
Scraping and crunching
In this era of Big Data, crunching numbers is what Kayser calls “a sexy profession”. In fact, his latest research involves “scraping” over three million newspaper articles using automated text analysis for clues about how the papers cover the economy. In their recent paper, "Does the Media Cover the Economy Accurately? An Analysis of Sixteen Developed Democracies", published in January 2021 in the Quarterly Journal of Political Science, Kayser and co-researcher Michael Peress, Associate Professor of Political Science at SUNY Stonybrook, examine whether media bias influences voters’ perception of the economy. To what extent is bias built into this economic information?
Their paper looks at how mainstream newspapers in six languages and 16 developed countries cover the economy. The study is based on more than three million articles in 32 newspapers – one left-leaning and one right-leaning for each country and uses automated text analysis to “scrape” the articles for information related to three economic indicators.
“The great thing about this project is that it is international and multilingual, unlike nearly all other media studies, which are usually just English language. And we go back farther in time than most studies,” Kayser says. “What we discovered is that mainstream newspapers basically cover the economy with considerable fidelity. We did find negativity bias – bad news is covered more frequently than good news – but left-wing mainstream newspapers, in terms of tone or frequency, were no different than right-wing ones.”
Much of Kayser’s research relates to election forecasting and outcomes. For example, he is using automated text analysis of newspapers to determine just how the state of the economy affects elections. The question for the next paper in the project is about the extent to which economic changes directly affect how people vote, or whether hearing about economic change in the media influences them.
Predicting policy outcomes
Particularly interesting in this German election year, amid the declining popularity of mainstream centrist parties and the rise of new ones, is Kayser’s project assessing how coalitions might shift in government.
This research with Jochen Rehmert at the University of Zurich on “coalition inclusion probabilities” looks at what motivates parties to switch their policy positions, and what leads to different outcomes in negotiations and in coalition governments. “This offers a systematic means to predict policy outcomes in coalition governments,” Kayser says.
In parliamentary democracies like Germany, there are always credible threats that smaller parties might leave the coalition, he says. This credible threat can help parties when negotiating with fellow members of government. “If you're a party in government, and you could easily leave that government to join an alternative government, then you've got a lot of power.”
For every party in a large set of 21 parliamentary democracies, Kayser and his fellow researchers have calculated the probability of whether a party would be included in government if an election were held that month – with data going back decades. Via the small monthly shifts in poll percentages, they can determine the points at which a coalition member could threaten to leave the government and form another ruling coalition, and this is the point at which a “credible threat” turns into bargaining power over policy choices.
“So we can ask in any month historically, what's the probability of each party being in an alternative government, what is their leverage, how much negotiating power do they have?” Kayser says. “The idea was to create a measure, and then in a next step to predict policy change.”
The researchers have already published a first paper predicting changes in environmental policy stringency. It shows that such a shift took place in Germany just as the government decided to exit nuclear power after the Fukushima nuclear accident in 2011. A shift in polling eroded the coalition relevance of a smaller governing party, the Free Democrats, as polling for the opposition Greens made them a viable potential coalition partner for Angela Merkel’s Christian Democrats. The Free Democrats’ lack of credible alternative coalition options left them too weak to stop their coalition partner from courting the Greens by accelerating the German nuclear phase-out, Kayser says. Shifts in polling only matter for policy outcomes when they change coalition options. The paper was published online in Legislative Studies Quarterly in 2020.
Between teaching and his new administrative duties as Dean of Faculty and Research, Kayser says he tries to reserve at least one day a week and weekends for the many research projects he’s involved in. In between walking the “pandemic dog”, of course.
The Hertie School is not responsible for any content linked or referred to from these pages.
Views expressed by the author/interviewee may not necessarily reflect the views and values of the Hertie School.