Media Coverage Shifts and Policy Overreactions

• In Event: Agenda Setting Dynamics in Contemporary Policymaking

Abstract

By exploiting a novel dataset and Machine Learning techniques, our paper studies the relationship between informational signals (captured by newspaper opinion columns) and policy change (proxied by adjustments to individual budget programs). First, we empirically test two seminal models of government information processing (parallel and serial). Then, we structure a unique empirical setting to provide causal estimates of bottlenecks in government attention, and policy overreactions to media coverage.

The paper sheds new light on the long-standing debate in Political Science and Public Administration between incrementalism and the punctuated equilibrium view of policy change. It does so by providing novel causal evidence at a granular level that has not been addressed by the literature so far. While existing studies focus on changes in the total budget of a government, our approach analyzes hundreds of individual government programs. To perform such analysis, we develop an innovative methodology–based on Natural Language Processing–that allows linking informational signals from thousands of opinion columns to these highly disaggregated expenditure programs.
Beyond the granular level at which we conduct our analyses, we believe that APSA members will be greatly interested in the empirical setting that we have structured in the second part of the paper. It consists of a natural experiment in which two independent earthquakes took place in different (and distant) cities of Mexico within the same month. More interestingly, one of these seisms happened on the exact day of the 1985 Mexico City earthquake anniversary, the most devastating in the country’s modern history (which contributed to a differentiated media coverage of the two events). These exogenous shocks, differentiated media coverage, and detailed data on budgetary responses to natural disasters provide a unique empirical setting that is rare to find in this research area. One that we exploit to produce nuanced evidence on the causal relationship between media signals and policy change. In addition, we complement the analyses by performing several robustness checks and testing different alternative hypotheses.

Overall, our results on disproportionate government reactions are consistent with the theoretical framework proposed by Jones and Baumgartner (2005). Their model provides a sound foundation for arguing that aggregate changes in national budgets follow non-normal distributions. However, this conjecture has not been properly tested at a level of disaggregation where changes are meaningful in terms of policy instruments and outputs. This is a major gap in the literature; one that our paper tries to fill by combining cutting-edge computational methods with a rigorous empirical strategy.

Understanding the link between media coverage and budget allocation is especially important in today’s world, where the digitization of news content, the propagation of information through social media, and the push for fiscal transparency are becoming critical, and a common component of the political culture of many countries and societies.

Author

• Daniele Guariso, The Alan Turing Institute