Mass media and its influence on - CREI

1 Els Opuscles del CREI. num. 44. December 2017. mass media and its influence on behaviour Ruben Enikolopov Maria Petrova The Centre de Recerca en Economia Internacional (CREI) is a research centre mass media sponsored by the Universitat Pompeu Fabra (UPF). and the Generalitat de Catalunya. It is located at and its influence the campus of the UPF in Barcelona. on behaviour CREI's objective is to promote research in international economics and macroeconomics subject to the highest standards of academic Ruben Enikolopov and Maria Petrova scholarship. mass media is the important source of information Els Opuscles del CREI are the instruments for the at the macro level in most of the countries. Informa- diffusion of CREI's research beyond the academic tion provided by mass media can affect a variety of world. Each Opuscle, aimed at a general public, outcomes ranging from results of the voting, to public gathers together conclusions and observations of policies, to ethnic violence, to teenage pregnancies.

2 Specialized papers published, or about to be This opuscle reviews recent literature in economics that published, in professional journals. The opinions studies how mass media affects the behaviour of people. It first discusses recent theoretical models that can help stated in Els Opuscles del CREI are the sole to understand both incentives of media outlets and, responsibility of the authors. correspondingly, people's responses. It then provides an overview of the empirical evidence that documents Pdf version of this opuscle and all the others the effects of mass media on a variety of political and Opuscles del CREI can be downloaded at: social outcomes. Editorial Board Antonio Ciccone (editor). 1. Introduction Jordi Gal . Teresa Garcia-Mil . Jaume Ventura mass media is a major source of information for the majority of population in most countries.

3 It Published by: CREI. Universitat Pompeu Fabra can shape public opinion and ensure popular sup- Ramon Trias Fargas, 25-27 - 08005 Barcelona port of particular politicians and policies. As a re- Tel. 93 542 13 88 sult, often politicians and other interested parties CREI, 2017. of this edition: Ruben Enikolopov and Maria Petrova have incentives to influence the media to make Design: Fons Gr fic sure it is friendly enough, whenever they have op- ISSN: 2604-2266 portunity to do so. The main goal of this paper is to overview evidence on the effect of mass media on people's behaviour, primarily, in the political domain. However, since the influence of mass me- 1. dia provides incentives for politicians and other 2011). Such effects take place if media omits some agents to influence media coverage, we will also relevant facts or provides deliberately incomplete discuss evidence of media capture and the impact information.

4 In these instances even fully rational of captured media . bayesian media consumers that are aware that the media is biased cannot fully undo the bias without This Opuscle reviews the recent economic getting access to more complete information from literature on the effects of mass media . We start alternative sources. Therefore, biased media can with briefly describing theoretical framework for have an effect on people's behaviour even in the analysing the effect of mass media . We then dis- situations when the audience knows about a po- cuss the challenges of estimating media effects, tential bias in media reports (see Str mberg 2016, and review papers that quantify media impact in Prat 2016, and Gentzkow, Shapiro, and Stone 2016. various circumstances. Finally, we'll talk about the for more detailed surveys of the theoretical litera- evidence for media capture and the limits of me- ture on biased media ).

5 Dia captures. The ability of media to influence behaviour by providing information and changing beliefs is 2. Theoretical framework more apparent in behavioural models that assume specific deviations from fully rational bayesian updating that comes in the form of categorical There are two main theoretical explanations thinking (Mullainathan, Schwarzstein, and Shleif- for why media can influence people's behaviour. er 2008), limited memory (Mullainathan 2002), or First of all, it can change their beliefs by providing double-counting of repeated information (DeMar- relevant information. Second, it can have a direct zo, Vayanos, and Zwiebel, 2003). In all of these effect on behaviour, independently of people's models, mass media can have an important effect information, through persuasion (see DellaVigna on behaviour even if it provides distorted and bi- and Gentzkow 2010).

6 Ased information, especially if people neglect the incentives of the sender of the information (Eyster The effect of mass media through the provi- and Rabin 2009). Overall, the literature suggests sion of information can be explained by most that it is very hard to fool fully rational bayesian standard models of rational bayesian updating, consumers of media , but it is much easier to do if such as informative (Stigler 1961) and signalling the consumers are not fully rational. (Nelson 1970) models of advertising, cheap talk models (Crawford and Sobel 1982) and persua- Models in which media can change behaviour sion games (Milgrom and Roberts 1986). In most without affecting beliefs assume that information of these models it is assumed that the information directly affects utility that people derive from cer- that is provided is correct, and that if there is any tain actions (Stigler and Becker 1977, Becker and bias in media reports, it will not have any effect Murphy 1993).

7 This type of theory includes the on the behaviour of people, since it will be ful- models of persuasive advertising (Bagwell 2007). ly discounted in the process of rational bayesian and psychological models in which non-inform- updating. However, in some models media bias ative peripheral factors may play a central role can influence people's behaviour even if media (see for example, Petty and Cacioppo 1996). consumers are rational (Kamenica and Gentzkow, 2 3. There are several predictions of belief-based reflects people's preexisting beliefs and attitudes. and persuasion -based models that in principle al- Thus, if we look at two newspapers with a par- low for distinguishing them empirically (DellaVi- ticular political stance and compare the behaviour gna and Gentzkow 2010). The two main specific of their readers, we do not know if the behav- predictions of the belief-based models is that; i) iour is different because reading the newspaper the effect of media will be stronger when receivers made people behave differently or because their are less certain about the truth; and ii) the effect preexisting political attitudes made them choose of media depends on its credibility.

8 The two main a particular newspaper and, at the same time, in- predictions of the preference-based models are fluenced their behaviour. Since the earlier studies that: i) the content of media reports may affect have done little to address this issue, it remains behaviour even when it conveys no information; unclear if the correlations that they uncover in- and ii) potential audience may take costly steps to deed reflect causation. avoid exposing themselves to persuasion by mass media . As we will see in the following section, In contrast, more recent empirical studies of there is clear evidence that supports predictions of media effects are based on the idea of finding the belief-based models ( Enikolopov, Petrova, some source of exogenous variation in media ex- and Zhuravskaya 2011, Chiang and Knight 2011), posure, or media content, to ensure that self-selec- but there are also studies that provide support for tion in media consumption or supply-side factors the preference-based models, especially in the are not biasing the results.

9 In other words, modern context of advertising ( Bertrand et al 2010). researchers use different, sometimes complicated empirical strategies to identify the causal impact of media exposure. Three empirical techniques, most 3. Methodology widely used by researchers, are field experiments, difference-in-differences method, and instrumen- tal variables approach. The first systematic studies of media persua- sion date back to the 1940s. They were inspired Field experiments by the seemingly effective mass persuasion cam- paigns organized by Joseph Goebbels in Nazi The most reliable method of identifying causal Germany in the 1930s. Perhaps surprisingly, these effects in scientific research, which is often consid- early studies, which were based on data, did ered as the gold standard in empirical analysis, not find any important media effects (Lazarsfeld, is conducting a field experiment.)

10 In the context of Berelson, and Gaudet 1944, Berelson, Lazarsfeld, media studies, field experiments imply randomiz- and McPhee 1944). Concluding that media has ing access to certain media outlet or specific piece only a minimal effect on people's behaviour, they of media content ( particular radio or TV pro- found that media strengthens people's predispo- gram). sitions, so that people become more confident in the views they already hold. However, these For example, to study the effect of newspapers studies faced a fundamental empirical challenge on political behaviour, Gerber, Karlan, and Bergan of endogeneity of media exposure. The choice of (2009) offered a 10-week-long subscription, free newspapers that people read, of radio stations that of charge, to either the Washington Post or the they listen to, or of TV channels that they view Washington Times to randomly selected residents 4 5.