Should we believe the American Academy of Pediatrics (2001) when it tells us that television exposure impairs educational achievement? Does violence in the media trigger arousal and violent crime, as Anderson et al. (2003) suggest? Can media content affect deep-seated decisions like fertility choices? Is there consistent evidence of imitation of media behaviour?
Questions such as these – of relevance to both researchers and policymakers – have motivated the literature on the impacts of media exposure. In the survey chapter for the first Handbook of Media Economics (DellaVigna and La Ferrara 2015), we review this literature, covering a wide range of economic and social outcomes and summarizing key studies within each area. To maximise readability, we structure the content by field of study – including the topics of education, health, crime, consumption, and family choices.
While one cannot succinctly summarise the results for every different field, we emphasise five recurring themes:
- Demand for entertainment;
- Direct versus substitution effects;
- Identification and time horizon;
- The role for entertainment media; and
- Policy impacts.
Demand for entertainment
The first one is the key role of the demand for entertainment. In nearly all the settings we consider, the consumer demand for media content is largely due to demand for entertainment, with the economic impacts emerging as a by-product. Children watch television because it’s fun, and are (very) unlikely to think of the possible impacts on their education. Adults choose to watch a violent movie or to follow a soap opera for entertainment value, disregarding possible effects on their own aggression or on family values. This implies that selection into a particular media is likely unrelated to the preference for particular economic outcomes – say, education, violent crime, or fertility. The overarching role of the demand for entertainment is reflected in the fact that people spend a large share of their time on media entertainment – in the US, the average amount of time spent watching television (which is just one form of media entertainment) is 2.7 hours per day, half of people’s leisure time (Aguiar et al. 2013).
The demand for entertainment differentiates the applications surveyed from the analysis of political or financial impacts of the media. The exposure to political information often reflects a direct demand for political content, as Stromberg (2015) and Gentzkow et al. (2015) stress. The exposure to financial-themed media like CNBC also reflects interest in investment advice.
We incorporate this insight into a simple model of media choice, building on Dahl and DellaVigna (2009). We assume that consumers choose the optimal use of time between several activities, some of which are media activities and some of which are not. For example, consumers decide whether to follow a soap opera or to go out with friends. The activities chosen impact relevant economic outcomes – like education, violent crime, and fertility – but these effects are not considered as part of the utility-maximising choice, simplifying the analysis. We derive comparative statics of parameters capturing some of the identifying variation in the media studies.
Direct and substitution effects
This takes us to the second key theme – direct versus substitution effects. We stress that there are two main sources of variation of media effects in this entertainment setting. The first is a shock to the entertainment value of a channel, or to its cost, which affects the audience for that particular medium. For example, violent movies are of higher quality on a particular weekend (a positive demand shock), or soap operas become more widely available in a given year (a positive cost shock). In both cases, as the comparative statics indicates, the resulting media effect estimates incorporate both a direct effect and a substitution effect. The cinema release of a movie like Hannibal implies that more people will be watching a violent movie and thus will be doing less of the second-best alternative activity. The net effect of this shock on crime, for instance, depends on the comparative effect of violent movies on crime relative to the effect of the alternative activity on crime. Similarly, to understand the impact of the introduction of a soap opera, we need to consider the activity and content that it is substituted for. In fact, thinking through the substitute activities and evaluating the estimated impact as a net impact, relative to substitutes, is a key take-away of this approach.
Our model also highlights a second source of variation – a direct shock to the content of the media. Suppose that an episode of an ongoing soap opera features a gay couple, or an occurrence of suicide. This change in content is likely to leave the utility-maximising choice of media entertainment mostly unaffected, especially if the content of the episode is unanticipated. Still, the content may affect economic behaviour, say through imitation. In this case, the estimated media effect captures the direct impact of the media, since the consumption of substitute activities is held constant.
Identification and time horizon
The third key theme for the survey is the role played by identification and time horizon. Take the analysis of imitation of media behaviour. A first question of interest is whether there is a short-run imitation effect after a media episode features a particular behaviour. A second question is if there are long-term imitative effects of prolonged exposure. To identify the first question, high-frequency variation in the content of a widely seen outlet is sufficient. But for the identification of long-run effects, one needs plausibly exogenous variation across places and over time in the introduction of a media outlet that carries unique content. The variation in identification determines the type of media effects one can credibly estimate.
Pregnancy and telenovelas
For family choices, we have evidence of imitation in both the short run and long run. Kearney and Levine (2014) estimate, among other outcomes, the short-run impact of the US show 16 and Pregnant on Google searches for keywords related to fertility choices. Conversely, La Ferrara et al. (2012) estimate the long-term effects on fertility rates in Brazil of exposure to telenovelas, taking advantage of the staggered introduction of Globo, which largely introduced telenovelas into Brazil.
In most other cases, however, it is not possible to estimate both short-term and long-term effects. In the research on media violence and violent crime, Dahl and DellaVigna (2009) exploit the natural experiment induced by the idiosyncratic release of violent movies to estimate the short-run effects of exposure to media violence. Their design, however, does not lend itself to the analysis of long-term effects. Indeed, to our knowledge, there is no study providing credible estimates of the long-run impact of exposure to media violence. The difficulty is that violent content on the media has been pervasive for a long time, making identification of long-term exposure near impossible, at least in the US. In cases such as this, it is tempting to look for alternative evidence on long-term effects – indeed, even the American Academy of Pediatrics cites correlational evidence between television usage and violent behaviour to support its policy recommendations. Unfortunately, this evidence is plagued by bias – taste for violence is likely to drive both behaviours – and should in our mind be disregarded. In these settings we have credible estimates of short-term effects but are left in the dark regarding long-term effects. In other cases, like the impact of video games consumption on crime, even short-run effects are problematic to estimate because high-frequency variation in video game consumption is very limited.
The fourth theme is the role for entertainment media. A striking feature is the scarcity of evidence about print media, i.e. newspapers and periodicals. To put things into perspective, variation in newspaper circulation plays a key role in the identification of political impacts of the media, as local papers provide critical information on local politics (Stromberg 2015). And yet, when it comes to the impact on outcomes such as education, health, crime, or family choices, most of the available evidence concerns the entertainment media – television, movies, and the internet.
Last is policy impacts. A number of the topics we examine reflect policy concerns, like the impact of television and of violent media. Yet, the research papers suggest policy implications which can appear surprising. A first example is that the studies reviewed do not find consistent evidence of a negative effect of television on education (the evidence in this respect is mixed), and find that the availability of violent movies in the short run leads to reductions, not increases, in crime. The key insight here goes back to the substitution effects – television or violent movies may be substituting other activities that are not better for the relevant outcomes. A second example is that one of the most clearly documented policy objectives achieved by media exposure – the reduction of the fertility rate in developing countries (e.g. Brazil in the 1970s and 1980s) – was attained as a by-product of the success of telenovelas, and was not a goal in the design of the entertainment material. This second example brings us back to the first theme, the overarching role of demand for entertainment.
A more recent set of studies takes stock of the primacy of demand for entertainment for policy purposes, and takes it one step further – why not attempt to incorporate policy goals into the entertainment material? The parallel with advertising is clear – marketing companies have for decades used product placement to sell products, suggesting a natural path for the use of entertainment to ‘sell’ policies. Still, this strategy faces thorny issues, as one may legitimately worry about propaganda.1
Aguiar, M, E Hurst, and L Karabarbounis (2013), "Time Use During the Great Recession", American Economic Review 103(5): 1664-1696.
American Academy of Pediatrics (2001), "American Academy of Pediatrics: Children, Adolescents, and Television", Pediatrics 107(2): 423-426.
Anderson, C A, L Berkowitz, E Donnerstein, L R Huesmann, J D Johnson, D Linz, N M Malamut, and E Wartella (2003), "The Influence of Media Violence on Youth", Psychological Science in The Public Interest 4: 81-110.
Dahl, G and S DellaVigna (2009), "Does Movie Violence Increase Violent Crime?", Quarterly Journal of Economics 124(2): 677-734.
Gentzkow, M, J Shapiro and D Stone (2015), "Media Bias in the Marketplace: Theory" in S Anderson, D Strömberg, and J Waldfogel (eds.), Handbook of Media Economics 1: Amsterdam.
Kearney, M S, and P B Levine (2014), "Media Influences on Social Outcomes: The Impact of MTV's 16 and Pregnant on Teen Childbearing", NBER Working Paper No. 19795.
La Ferrara, E (2015), "Mass media and social change: Can we use television to fight poverty?", JEEA-FBBVA lecture.
La Ferrara, E, A Chong, and S Duryea (2012), "Soap Operas and Fertility: Evidence from Brazil", American Economic Journal: Applied Economics 4(4): 1-31.
Strömberg, D (2015), "Media Coverage and Political Accountability: Theory and Evidence", in S Anderson, D Strömberg, and J Waldfogel (eds.), Handbook of Media Economics 1: Amsterdam.
1 We examine the issue of ‘educational entertainment’ (or ‘edutainment’), discussed more in depth in La Ferrara (2015), in the survey.