«Matthias Heinz 1 University of Cologne Johan Swinnen 2 University of Leuven and Stanford University July 2014 Abstract We reviewed all articles ...»
Media Slant in Economic News: A Factor 20
Matthias Heinz 1
University of Cologne
Johan Swinnen 2
University of Leuven and Stanford University
We reviewed all articles reporting on job creation and job destruction by companies in Germany
between December 2000 and September 2008 in Die Welt, one of the leading German
newspapers, using experiments to test our selection criteria. There is a large difference in
coverage of job creation and job shedding. Despite the fact that the economic situation in Germany improved over the period (unemployment rate fell by 2.0%), more than ten times as many articles report on negative employment news compared to positive news. When we control for the number of jobs involved, we find an even stronger slant: on a per-job basis, the slant to downsizing increases to a factor greater than 20. Additional tests indicate that these effects are similar in other leading German newspapers.
Keywords: Media economics, rational ignorance, negative news coverage JEL Codes: L82, D83 This research was supported by the University of Leuven’s Research Council (Methusalem Project).
1. Introduction Perceptions on the economy matter. They influence investor decisions, company strategies, stock market behavior, and decisions of individuals on employment choices and investment in education and skills. Perceptions of the public and politicians will influence public policies.
They matter beyond the economy. Psychological problems such as depressions, etc. may result and it is well know that perceptions on the economy are a prime factor in voting behavior in elections, probably best reflected in Bill Clinton’s 1992 election “war room” slogan: “It’s the economy, stupid!”. Not surprisingly, incumbent politicians and parties often complain that voters and the public’s view on the economy is too pessimistic, as e.g. reflected in the 2012 electoral campaign of President Obama.3 A key factor in influencing perception is mass media and the nature of information provision by the media, i.e. the existence of ‘media slant’. Media slant can take various forms, and there is evidence that slant can be significant (Groseclose and Milyo, 2005; Gentzkow and Shapiro, 2010; Strömberg, 2001).4 Studies have identified several possible explanations for the existence of slant. Media slant can be induced by supply and/or demand factors. The most obvious source is preferences from the owners, editors, or journalists who may affect the news coverage (Bovitz et al., 2002).
There are several reasons why, in an era of massive information provision, there is still room for perceptions which deviate from reality. McCluskey and Swinnen (2010) argue that even in today’s media world most consumers and voters are “rationally ignorant” – a term coined by Downs (1957) in a different media era. They provide several reasons why, despite a massive amount of available information, people choose to be less than fully informed. First, most obviously, if the price of news is high compared to the marginal benefits of information, it is rational for individuals not to be fully informed. Second, reducing the price of news will increase consumer information, but only up to a point.
Even when news and information is free, it takes time, energy and attention to process the information.
Consumers will stop acquiring more information when the opportunity costs of processing the information become larger than the benefits. Opportunity costs play an important role, especially when considering trade-offs in information accumulation on various issues. A third reason why consumers may choose to be less than fully informed has to do with the source providing the information.
Ideological bias or distrust of the source may cause consumers not to inform themselves any further.
The literature (e.g. Gentzkow and Shapiro, 2010) differentiates between slant and bias, and defines bias as a behavior that is not in line with maximization of profits. In line with this, we use the word “slant” in most parts of the paper. If we discuss, however, about individual behavior that is not necessarily in line with profit maximization of media companies, we use the work “bias”.
This is evident in mass media owned by the state. However also in commercial media, owners or advertisers may push for their preferences being reflected in the reporting. Typically there is a tradeoff between political objectives (i.e. using the media to express owners’ ideological bias) and commercial objectives (Mullainathan and Shleifer, 2005). Too much bias may reduce sales – because of consumers’ distaste for bias or the differences with their personal political preferences – and advertising revenues.5 Another supply-induced form of slant is from falsehoods or from information hidden or distorted by sources or journalists. Dyck and Zingales (2002) argue that journals spin stories to reward sources for providing information. Competition between information sources affects news reports (Baron, 2006). Journalists eager for a scoop or under pressure to attract attention may provide biased information.
On the demand side, Mullainathan and Shleifer (2005) argue that readers or viewers have a preference for news that is consistent with their initial beliefs, and that media organizations have therefore an incentive to slant their reporting towards confirming their readers or viewers initial beliefs. In Gentzkow and Shapiro’s (2006) theory consumers are uncertain about the quality of information. As consumers think that a newspaper confirming their expectations is more valuable, newspapers have an incentive to slant their reporting towards readers’ beliefs to build a reputation for accuracy, Our study focuses on one particular form of slant: the tendency of media to over-report negative stories. McCluskey and Swinnen (2004) use a demand-side argument to explain a slant towards “bad news”. As readers or viewers give more weight to negative than to positive information, media organizations respond to this in their coverage selection. The rationale for Gabszewicz et al. (2001) show that the media’s incentives to appeal to a larger audience and hence be more attractive to advertisers may induce editors to moderate the political messages they display to their readers. Baron (2006) explains how bias may be larger in competitive media markets, while Sutter (2001) and Corneo (2006) argue that collusion or a concentration in media ownership makes bias more likely, which is consistent with the findings of Gentzkow and Shapiro (2006). Mullainathan and Shleifer (2005) separate the impact of competition on slant along two axes and find that competition will neutralize ideological bias but intensify spin.
the higher valuation of information about issues concerning negative welfare is that media consumers can use that information to make decisions that avoid income losses (Kahneman and Tversky, 1979). Siegrist and Cvetkovich (2001) also found in psychological experiments that people place greater trust in research results indicating health risk and the confidence in the results increases with an increasing indication of health risk.
There are only a few empirical studies on this, and only two studies on coverage of economic issues. Almost all studies argue that there is a slant towards ‘negative coverage’ in the mass media for a variety of policy and public interest areas. Cohen (1983) finds that mass media pay more attention to negative stories indicating the presence of nuclear power risks than to positive ones indicating the absence of risks. Koren and Klein (1991) compared media reports on two medical studies which appeared simultaneously and which both analyzed radiation as a risk for cancer. One study found no impact while the other study’s results pointed at a risk of cancer from radiation. The authors find that the study that showed a potential risk for cancer received significantly more attention than the other report. Ditton and Duffy (1983) and O’Connell (1999) find that the number of crime stories dealing with extreme and violent offences is disproportionate compared to the actual occurrence. Kalaitzandonakes et al. (2004) find that mass media reports on food safety issues and biotechnology focus disproportionately on negative aspects.
Swinnen and Francken (2006) find that newspapers’ coverage of issues related to globalization, trade, and the WTO is predominantly negative and that the vast majority of media coverage about trade and globalization issues is about riots and demonstrations that surround summits of political leaders on these issues. Harrington (1989) shows that U.S. television networks give greater coverage to bad economic news. Reports on unemployment, inflation, and growth were 34% longer and twice as likely to lead the evening news broadcasts when these statistics were worsening than when they were improving, ceteris paribus. There are measurement problems in many of the empirical studies due to the lack of good data, low number of observations and the difficulty in identifying “positive” and “negative” impacts, and as a consequence none have been able to measure the extent of the slant.
Our study contributes to the literature by using a more elaborate and carefully constructed dataset on economic reporting, including more than 7,700 newspaper articles appearing in a time span of almost eight years. Moreover, only a few studies have focused on economic reporting. Our media coverage dataset is an extended version of the dataset used by Friebel and Heinz (2013) who analyze economic xenophobia by comparing media reporting on downsizing of domestic and foreign firms. We have extended the dataset by analyzing the media coverage on upsizing in all leading German quality newspapers. We use lab experiments to check and test the robustness of our selection of words and concepts to identify good or bad news.
As a result we have a cleaner indicator of good and bad news than previous studies and we are the first to be able to estimate a statistical measure of slant. For example, in Harrington (1989) either a positive or negative event occurs at a point in time, and media have to decide whether to report about it or not. However, in reality these events may overlap. Our data on firm employment clearly show that it is often the case that some firms shed jobs and other create jobs at the same time.
Being able to measure this slant is important since a negative slant in news coverage can have important implications as it may distort public opinion. For example, it has been shown that public awareness of crime is substantially different from official statistics due to the negative slant in newspaper reporting (Smith, 1984). Moreover, negative information has a much greater impact on individuals’ attitudes than positive information (Soroka, 2006) and negative information plays a greater role in voters’ opinion formation and voting behavior (Aragones, 1997; Easaw, 2010). The predominance of negative news in the mass media is likely to reinforce these effects. This might be in particular important in the case of the slant we identified, as the size of the effect is enormous and as we focus on a topic (upsizing vs.
downsizing), where people seem to react quite strongly and emotionally on (bad) news.6 The paper is organized as follows. Chapter 2 provides information about the data. Chapter 3 provides the main results, Chapter 4 contains robustness checks. Chapter 5 concludes.
2. Data Using the media data base LexisNexis, we identified all articles reporting on up- and downsizing from companies between December 2000 and September 2008 in Die Welt, one of the leading German newspapers. We only record articles that mention the creation (shedding) of jobs by a firm in Germany. Articles reporting about upsizing (downsizing) in general or in whole sectors are not recorded.
We utilize the following approach to identify all articles reporting about downsizing.
The identification strategy for the upsizing articles is the same, with one exception, which will be explained below. Firstly, we read thousands of newspaper articles in order to identify German synonyms for the term “downsizing”. The list of synonyms is provided in the Appendix. There are some words that are direct German synonyms for the word downsizing, for example “Stellenabbau” (staff reductions). However, there are many other terms that without the right context would not be immediately identifiable as a synonym for downsizing.
For example, the word “Restrukturierung” (restructuring) can be used for financial restructuring but also for reporting about job destruction.
Friebel and Heinz (2013) provide a number of case studies illustrating how consumers react to downsizing news. For example, in 2008, after Nokia’s decision to shut down a plant in Germany attracted massive negative media reporting, the company lost 8 percentage points market share in the German mobile phone market in the subsequent six months. Nokia’s market share in the rest of Europe remained constant, the global market share even increased.
Secondly, we tested these synonyms in two laboratory experiments at the FLEX laboratory at Goethe University in Frankfurt.7 In the first one, participants were asked to create an own list of German synonyms of downsizing. The aim of this experiment was to ensure that we had not missed synonyms on our list. In order to verify our understanding of terms related to downsizing we conducted a second experiment. Participants were confronted with 40 words.
8 were synonyms for downsizing, 17 were terms where depending on the context would suggest a downsizing event. 15 words did not concern job destruction at all, e.g. the term “Verlust” (loss). Using a scale from “by no means” (1) to “by all means” (5), participants were requested to state to what extent these terms relate to downsizing. The synonyms had a mean score of 4.43 (standard deviation 0.24). For the context-depending terms we find a mean of 3.61 (s.d.
0.45), for the words that we deemed not to be related to downsizing 2.19 (s.d. 0.56).
Thirdly, based on the list of synonyms, we identified and checked all articles in Die Welt which contained any of our downsizing terms. In total, 498 different firms are mentioned in one or several of these articles. In a next step, firm by firm, we checked in detail all articles in Die Welt in our period of observation in which the companies are mentioned. In total we read around 40,000 articles.