Research

Job Market Paper

Quantifying Echo Chambers and Their Impact on News Engagement: Evidence from a Facebook Algorithm Update

I show that social media platforms face incentives to make algorithmic design choices which amplify misinformation and polarization in news consumption – specifically, by increasing the prevalence of echo chambers. This study leverages a 2018 Facebook algorithm update to investigate how increased network homophily affects news spread in a social network. My findings support a model of news consumption on social media where rational consumers re-share news articles based on reputational concerns and a desire to spread factual news; I extend this model to analyze how network structure influences tribalism in news engagement choices. Using an empirical reworking of the model, I measure the magnitude of the increase in social network homophily caused by the algorithm update. I then use this credibly exogenous shift to test the model’s main predictions. As predicted, greater homophily increases engagement with less reliable, more divisive news, and intensifies tribalism of engagement behaviour with media via an ‘agitation bubble’ effect. The results demonstrate that echo chambers created by social media platforms can drive tribalism and misinformation, rather than merely reflecting the existing prevalence of these phenomena in society. As the algorithm update benefited the platform, the results further suggest that platform incentives in shaping network structure misalign with social welfare and clarify how the global shift toward news consumption via social media can damage news diets. This work underscores the role of the communication network structure in explaining polarization, shifting focus away from explanations based on changing beliefs or cognitive biases.

Works in Progress

If it Bleeds, it Leads: Multivariate Firm Objectives with Concave Arguments

News outlet editors face a tradeoff between profit maximisation and journalistic integrity. While there is a social benefit to producing informative, truthful news which editorial staff often internalise, this can conflict with incentives to sensationalise news to gather more clicks and/or subscriptions which generate revenue. This mechanism has been used to explain a decrease in the informativeness and reliability of news as advertising revenue for news outlets has fallen over the last two decades. I propose a model of the market for news where news outlets maximise an objective which integrates company profits and journalistic integrity, demonstrating that the trend of declining informativeness and increasing sensationalism observed in the news industry is consistent with a firm objective which is concave in its profit argument. I develop this into a structural model which I use to disentangle the roles of supply and demand in driving a ‘sensationalist’ or ‘negativity’ bias in the news output amongst the top newspapers in the UK. I estimate the model using granular data from Facebook on news demand, data on the prominence that news outlets give articles on their websites (supply choices), and quantitative sentiment data on article headlines generated using NLP. I find that there is a strong sensationalist bias in demand for news, which news outlets respond to with sensationalist supply choices. I then use a series of (credibly exogenous) shocks to news outlet profits to test the hypothesis that the company’s objective functions are concave in their profit argument. I further develop the model to derive the policy implications of the dual firm objective. My findings have broad implications for the regulation of the news industry in light of an increasingly fragmented market.

The Impact of Social Media Bargaining on Media Market Competition

I exploit a change in social media regulation in Australia to examine the impact that media organisations’ bargaining power with social media companies can have on concentration in newspaper markets. I use a difference in difference approach with a synthetic control to compare media market concentration in Australia to other countries before and after the regulatory change. I find that the shift, which forced social media companies to compensate newspapers for news content appearing on their platforms, increased concentration in newspaper markets by cementing the advantage larger media conglomerates have as a result of their disproportionate bargaining power with social media companies. I develop a bargaining model to rationalise the results, drawing conclusions about the welfare effects of compensatory policies like the one implemented in Australia.

Boundaries of the State

With Thiemo Fetzer and Jacob Edenhofer Formal conceptions of state capacity have mostly focused on indirect measures of state capacity – by, for instance, using tax revenues as a proportion of GDP as a proxy for state capacity. Yet, this input view of state capacity falls short, especially since cross-country empirical evidence suggests that similar tax revenues, as a percentage of GDP, can produce starkly different outputs - both in classic economic terms and in broader terms that citizens would recognize as good outcomes, such as life expectancy, infant mortality, deaths of despair, and equality of opportunity. This paper argues that conventional views of state capacity ignore one crucial boundary of the state or dimension of state capacity, namely its capacity to gather and process information, and how the presence or lack of such informational capacity constrains governments in responding to crises, such as the recent energy price shock. Our framework provides the analytical toolkit to examine how this informational boundary of the state shapes the incentives for policymakers to resort to untargeted and/or distortionary policy instruments, as opposed to targeted and non-distortionary ones, in responding to crises. In doing so, our framework draws attention to the need to redefine the boundaries of the state, the firm and notions of property rights around data and information more generally. The policy response to the energy crisis following the invasion of Ukraine provides the empirical context upon which we bring this theoretical apparatus to bear, though the latter can be straightforwardly extended to other recent crises.

Vertical Integration and Demand Steering with Information Frictions: Evidence From the Online Advertising Industry

I develop and estimate a structural model of the online display advertising intermediary market. I find substantial evidence that vertically integrated entities benefit from information frictions which restrict information transfer between non-integrated firms, giving integrated downstream  firms a competitive advantage. A counterfactual simulation reveals that the market share of the downstream firm which is vertically integrated with the most dominant upstream firm (Google) would be 1.66 percentage points lower if information frictions were eliminated. My findings imply that privacy protection regulation like GDPR which inhibits the ease of data transfer between firmscan have anti-competitive effects.