Periodic Reporting for period 4 - BANK-LASH (Banks, Popular Backlash, and the Post-Crisis Politics of Financial Regulation)
Berichtszeitraum: 2023-03-01 bis 2024-02-29
BANK-LASH attempts to draw together the connections between the media coverage of banks and banking, public opinion about financial regulation, and policy reforms in six countries since the 2008 financial crisis. The six countries are Australia, France, Germany, Switzerland, the United Kingdom and the United States The project poses three questions:
--How do attitudes towards financial regulation vary across countries?
--How does the framing and intensity of media coverage vary across countries?
--What are the political conditions under which countries adopt significant reforms of the regulations governing big banks in the post-crisis era?
These problems are important for society because there is a widespread perception that banks and senior bankers largely escaped accountability for the damage done to national economies during Great Recession, despite the prominent role of bank behavior in causing that crisis. The anti-bank thread runs through populist political reactions in all the countries studied, on both the political left and the political right. What the BANK-LASH project attempts to understand is what role the media plays in the process of opinion formation, as well as whether public opinion has led to big changes in the way banks are governed.
--How do attitudes towards financial regulation vary across countries?
--How does the framing and intensity of media coverage vary across countries?
--What are the political conditions under which countries adopt significant reforms of the regulations governing big banks in the post-crisis era?
These problems are important for society because there is a widespread perception that banks and senior bankers largely escaped accountability for the damage done to national economies during Great Recession, despite the prominent role of bank behavior in causing that crisis. The anti-bank thread runs through populist political reactions in all the countries studied, on both the political left and the political right. What the BANK-LASH project attempts to understand is what role the media plays in the process of opinion formation, as well as whether public opinion has led to big changes in the way banks are governed.
Our first work package (BANK-LASH 1) aimed to establish what public opinion looks like in the area of financial regulation, in the countries we studied: Australia, France, Germany, Switzerland, the UK, and the US. Based on a large survey conducted in 2020, we determined that in all six of the countries there is clear evidence that preferences for banking regulation form a coherent set of beliefs. Those beliefs, moreover, are distinct from conventional views about economic redistribution. We repeated that survey three years later (2023) in four of the six countries (France, Germany, UK and US), and we found exactly the same thing. In sum, views about financial regulation are driven by core values and by banking-related variables. They are NOT driven by ideology, political partisanship, or most of the measures associated with political populist attitudes.
Beyond measuring these preferences, we wanted to understand how media coverage of banking influences attitudes towards financial regulation. We had hypothesized in the Description of the Action that frames emphasizing scandal would work primarily through the emotion of anger and that they would increase support for financial regulation. We tested both these hypotheses in the second and third waves of our 2020 survey. We confirmed that reading about media coverage of scandals reliably increases preferences for financial regulation. We also found that anger mediates that effect. We published these results in an article in the American Journal of Political Science , entitled “Banklash: How Media Coverage of Bank Scandals Moves Mass Preferences on Financial Regulation.”
Using our 2020 data, we were further able to show, contrary to much of the existing literature on redistributive attitudes, that narrative frames not only influence views of banking regulation – they also influence attitudes on redistribution. Here our contribution was to demonstrate that it is not simple facts about the level of inequality that change preferences, but narratives about the sources of inequality. In other words, it is stories that change minds, not just facts. We published this piece in Comparative Political Studies.
Work package 2 aimed at answering our second research question, about the character of media coverage. We used machine-learning techniques to produce the first-ever comparative portrait of media coverage of banking issues, with coverage from 2007-2018. We discuss the development of these measurement techniques in section 1.2 because the techniques used are extremely cutting-edge and took us to the political science frontier of using large language models (similar to the underlying models used in generative AI such as ChatGPT).
What we found through the analysis of media coverage is consistent with our investigations of the impact of bank scandals on public opinion. Immediately after the financial crisis, media coverage of the crisis and financial bailouts was closely entwined with media coverage of financial regulation. However, from 2012 onward, media coverage of regulation marched in lock step with coverage of scandals. This is consistent with much of the work carried out on policymaking under the third work package, which found that scandals played an important role in raising salience at moments of passing laws.
Finally, the case studies of reform helped us to answer our third research question – under what conditions do we observe significant reforms of financial systems? Here, the project has shown the states with the reputation for where banks are the most powerful – the US, UK, and Switzerland – are also those where we have observed the most significant reform.
Beyond measuring these preferences, we wanted to understand how media coverage of banking influences attitudes towards financial regulation. We had hypothesized in the Description of the Action that frames emphasizing scandal would work primarily through the emotion of anger and that they would increase support for financial regulation. We tested both these hypotheses in the second and third waves of our 2020 survey. We confirmed that reading about media coverage of scandals reliably increases preferences for financial regulation. We also found that anger mediates that effect. We published these results in an article in the American Journal of Political Science , entitled “Banklash: How Media Coverage of Bank Scandals Moves Mass Preferences on Financial Regulation.”
Using our 2020 data, we were further able to show, contrary to much of the existing literature on redistributive attitudes, that narrative frames not only influence views of banking regulation – they also influence attitudes on redistribution. Here our contribution was to demonstrate that it is not simple facts about the level of inequality that change preferences, but narratives about the sources of inequality. In other words, it is stories that change minds, not just facts. We published this piece in Comparative Political Studies.
Work package 2 aimed at answering our second research question, about the character of media coverage. We used machine-learning techniques to produce the first-ever comparative portrait of media coverage of banking issues, with coverage from 2007-2018. We discuss the development of these measurement techniques in section 1.2 because the techniques used are extremely cutting-edge and took us to the political science frontier of using large language models (similar to the underlying models used in generative AI such as ChatGPT).
What we found through the analysis of media coverage is consistent with our investigations of the impact of bank scandals on public opinion. Immediately after the financial crisis, media coverage of the crisis and financial bailouts was closely entwined with media coverage of financial regulation. However, from 2012 onward, media coverage of regulation marched in lock step with coverage of scandals. This is consistent with much of the work carried out on policymaking under the third work package, which found that scandals played an important role in raising salience at moments of passing laws.
Finally, the case studies of reform helped us to answer our third research question – under what conditions do we observe significant reforms of financial systems? Here, the project has shown the states with the reputation for where banks are the most powerful – the US, UK, and Switzerland – are also those where we have observed the most significant reform.
The BANK-LASH 1 work package has demonstrated that, as predicted in our proposal, the politics of financial regulation is not derivative of either views on inequality or partisan politics. Our first work package has also deployed two experimental waves to examine the ways in which different types of media frames of stories about banks influence regulatory attitudes. We focus in particular – drawing in our work in the second work package, BANK-LASH 2 – on scandal frames. We find strong cross-national evidence that scandal frames increase preferences for regulation, though the effects of different types of scandal frames vary across countries. We find cross-country evidence that some scandal frames affect attitudes through the channel of emotion (in particular, through anger). These data were collected in 2020, and there is still much that we have to learn from them, as we work on several papers simultaneously that deal with various aspects of the study.
The BANK-LASH 2 work package has already shed light on how supervised vs. unsupervised techniques can inform questions of framing in the media, and there is much that we still have to learn from the data. In an article published in Political Communication in 2020, Tom Nicholls and Pepper Culpepper examine the performance of three unsupervised methods for identifying frames in text: k-means combined with natural language processing-feature identification, evolutionary factor analysis, and structural topic modelling. Building on that work, we applied the rapid advances in computer science large language models (LLMs) to improve the way we classify text . We developed and validated an improved method for supervised classification based on a corpus of English, French, and German news articles on banking, without requiring prior translation. We demonstrate that the new model is better across the board than a classic support vector machine (SVM) for our dataset.
The BANK-LASH 2 work package has already shed light on how supervised vs. unsupervised techniques can inform questions of framing in the media, and there is much that we still have to learn from the data. In an article published in Political Communication in 2020, Tom Nicholls and Pepper Culpepper examine the performance of three unsupervised methods for identifying frames in text: k-means combined with natural language processing-feature identification, evolutionary factor analysis, and structural topic modelling. Building on that work, we applied the rapid advances in computer science large language models (LLMs) to improve the way we classify text . We developed and validated an improved method for supervised classification based on a corpus of English, French, and German news articles on banking, without requiring prior translation. We demonstrate that the new model is better across the board than a classic support vector machine (SVM) for our dataset.