Good News!

Social Identity and Media Use

SIT postulates that people have a sense of belonging to certain groups. The evaluation of their in-group is termed social identity (Abrams & Hogg, 1990; Tajfel & Turner, 1979). SIT has widely been used to explain mass media exposure (Harwood & Roy, 2005; Reid, Giles, & Abrams, 2004; Tarrant, North, & Hargreaves, 2001). Trepte (2006) suggested the social identity model of media effects (SIM). The SIM posits that when a person reads news media or other forms of media content, this content will make one or more in-groups salient to this person. Further, people compare themselves with the out-group featured in the media. By engaging in social comparison, the individual aims to gain positive distinctiveness. In other words, the individual wants to have a positive evaluation of his or her in-group. In this process, when the media positively evaluate the in-group, this nurtures two motives: First, it positively influences a person’s self-evaluation and social identity. Second, it increases an individual’s self-esteem.

Previous research has demonstrated that these motives influence selective exposure to news and entertainment media (Harwood & Roy, 2005; Mares & Cantor, 1992; Mastro, 2003; Reid et al., 2004; Weaver, 2011). Also, it has been demonstrated that the relevance of news to a person’s identity substantially influences the amount of attention a person pays to and the person’s degree of involvement with the news content (Appiah, Knobloch-Westerwick, & Alter, 2013). Whereas prior research has studied identity effects on selective exposure to news media and other forms of media content, it did not focus on the question of how reading valenced news media might affect the readers’ social identity.

Effects of News Reading on Positive Distinctiveness

The news media as well as the entertainment media play on readers’ and viewers’ feelings of belonging to certain social groups and appear to be constantly evaluating these social groups. Rivenburgh (2000) investigated the national newspapers La Prensa (Argentina), The New York Times (United States), and Berlingske Tidende (Denmark). She demonstrated that these newspapers’ stories on how national citizens and foreigners interact serve either to enhance or to protect national identity. She revealed that all of the aforementioned newspapers tended to provide favorable reports of the respective country’s national identity. Bloom (1990) even stated that, “[...] media can appropriate any international issue and use it to defend or enhance national identity. The murder of a fellow citizen overseas […] can become an issue that is projected as threatening national identity” (p. 87).

News media have been shown to invite idiosyncratic readings that are based on group memberships. In a survey of college students, Tan, Fujioka, and Tan (2000) demonstrated that positive–negative evaluations of TV portrayals of minorities predicted stereotypical real-world responses. They showed that the valence (i.e., positive or negative) of portrayals of racial and ethnic minorities on television was closely related to the stereotypical responses associated with these groups. Also, Coover (2001) posited that racial identity determines how recipients are affected by race in the media. Both SIT and the SIM suggest that readers want to arrive at a positive evaluation of their own nationality. Therefore, they use and remember positive information from a news article according to their need to positively evaluate their in-group.

As stated here, media such as valenced news presumably trigger positive distinctiveness. Over time, the individual accumulates experiences with a certain in-group and develops a general evaluation of this group termed social identity (Abrams & Hogg, 1990; Tajfel & Turner, 1979). Social identity can thus be understood as a baseline in-group evaluation. Of course, the media add to social identity by constantly offering valenced content such as one-sided news.

Coover (2001) demonstrated that the overall acceptance of White social identity influences how media content is evaluated. Mastro and Kopacz (2006) share this view and maintain that real-world stereotyping and policy positions are crucially influenced by the extent to which representations of ethnic minorities accommodate recipients’ social identity. Hence, not only the media stimuli’s valence but also the reader’s national identity should influence positive distinctiveness. Therefore, in evaluating people’s reactions to positive or negative coverage in the form of positive distinctiveness as proposed in our first hypothesis, we controlled for the strength of national identity.

Social Identity and News Knowledge

Hearing or reading positive news about one’s in-group should be considered particularly useful by readers because it feeds into their social identity motives. A reservoir of positive knowledge about one’s in-group should nurture self-esteem and should be given more consideration than self-defeating information. In the present study, the type of in-group knowledge we focused on was factual political knowledge about a current political issue that is portrayed in an online news text. The acquisition of this kind of knowledge was defined by Eveland and Hively (2008) as the “ability to remember or recognize bits of information that can be determined by observers to be true or false” (p. 3716).

News knowledge crucially influences voting decisions, political engagement, and deliberation processes (Bakker & de Vreese, 2011; Gehrau et al., 2014). Hence, it seems important to further clarify potential antecedents of political news knowledge. Previous research has shown that in-group favoritism and out-group derogation influence selective reading and selective attention (Appiah et al., 2013; van Bavel & Cunningham, 2012). Both processes demand higher cognitive investment and cognitive organizing. That said, knowledge building as another process of cognitive investment and organization might also be influenced by valenced information. Not only selective exposure as shown in previous research (Knobloch-Westerwick & Hastall, 2010) but also selective remembering should be determined by positively valenced news texts. In the present study, we expected that identity processes will foster the acquisition of knowledge about a specific current political issue that is portrayed in an online news article.

Further, we suggest that readers differ in the degree to which they have a tendency to compare themselves with others when reading news articles (Gibbons & Buunk, 1999; Hinkle & Brown, 1990). Donsbach (2009) has shown in several studies that comparison behavior motivates people to read news that is pleasing to them. In other words, people who have a greater tendency to compare themselves with others should be more eager to find information that is agreeable to them. Thus, when evaluating the influence of the valence of a news article on political knowledge, social comparison behavior should be controlled for.

National Differences in Social Identity Processes

Previous research has assumed that there are no differences between Germany and the US; further, research has relied primarily on experimental methods and was mostly conducted in a single country (Smith & Long, 2006). Social identity was usually assumed to be an overarching social-psychological phenomenon (Abrams & Hogg, 1990; Tajfel & Turner, 1979).

However, cross-cultural differences seem possible as nationality might be a proxy for many influences such as the media system or cultural norms. For example, the news media and the ways in which readers elaborate on news differ tremendously across cultures (Iyengar et al., 2010). It has been demonstrated that people living in countries with a public service media system (e.g., Germany) are more likely to be confronted with different kinds of news than citizens of the US, which has a market-driven media system (Curran, Iyengar, Lund, & Salovaara-Moring, 2009). Moreover, historically derived norms might influence how international news is interpreted; for example, displaying signs of national pride is not perceived as equally appropriate across different countries (Smith & Kim, 2006).

The goal of our work was to predict the influence of news reading on positive distinctiveness and news knowledge while using nationality as the social group in question. A specific analysis of cultural influences was not within the scope of this study. Nonetheless, we argue that, when investigating social identity, research should account for the possibility that these influences may be at work. Consequently, we asked:

Overview

We conducted an experiment with two groups. The first group read two news articles that both had a positive valence (VAL_pos) with regard to the manner in which relevant societal issues were handled by the respondent’s home country and a negative valence with regard to the manner in which relevant societal issues were handled by a foreign country. The second group (VAL_neg) read two news articles that both had a negative valence with regard to the manner in which relevant societal issues were handled by the respondent’s home country and a positive valence with regard to the manner in which relevant societal issues were handled by the foreign country. Between the two valenced articles, participants read a distractor article. The study was conducted in the US and in Germany. We ran pretests in a laboratory beforehand with the aim of finding news issues that participants from both countries would consider equally relevant (see the Stimuli section). In addition, we used the pretest to investigate the effectiveness of the experimental manipulation.

Participants

A total of 160 participants were recruited for the pretest; 102 students from a large Midwestern university took part in the US (58.82% women, 29.41% men, 11.77% did not report gender; age: M = 30 years, SD = 3.32, range = 19–48), 58 from a university in south Germany (75.86% women, 22.41% men, 1.73% did not report gender; age: M = 23 years, SD = 2.49, range = 20–33). In both countries, participants were recruited from different university courses on communication.

A total of 364 participants were recruited for the main experiment (60.71% women, 39.01% men, 0.28% did not report gender; age: M = 20 years, SD = 2.42, range = 17–47); 119 participants took part in the US (78.99% women, 21.01% men, age: M = 20 years, SD = 3.13, range = 18–47), 245 in Germany (51.84% women, 47.55% men, 0.61% did not report gender; age: M = 20 years, SD = 2.00, range = 17–28). In both countries, we applied the same recruitment procedure that we used for the pretest. In the US, participants were granted one course credit for their participation. In Germany, participation was voluntary, and there were no penalties for students who decided not to take part in the experiment.

Stimuli

On the basis of authentic news reports (“U.S. Students Still Lag Behind Foreign Peers”; Zhao, 2012), stimulus texts were first written in English and afterwards translated into German. In order to guarantee the accuracy of the translation, five people who were native speakers of both English and German worked on each article. Previous research (Rasmussen Reports, 2013; Westle, 2006) showed that the following six current political topics were the most important issues for young adults in both Germany and the US: the educational system, economics, the health-care system, the job market, energy policies, and national security. For all topics, two different versions were prepared and pretested: One featured the US in a positive way and Germany in a negative way, the other presented Germany in a positive way and the US in a negative way. The structure, length, and language were similar between versions.¹ In order to disguise the real research topic, one neutral distractor article was chosen. The distractor article did not refer to national aspects and featured a story on self-checkout lanes.² Distractor articles were not pretested. We informed the participants that the study was about information from news articles on the consent forms that were administered with both the pretest and the main study.³

Pretest

For the pretest, six articles each dealing with one of the aforementioned political issues were presented consecutively in a random order in an online survey. Respondents rated the articles according to the dimensions of personal relevance, salience of nationality, detection of text valence (positive vs. negative with respect to one’s own social identity), national identity relevance, social comparison, and credibility. All articles were ranked according to these dimensions. The two highest-ranking articles were selected: the article about national security (NAS) and the article about national education (EDU).

A three-way ANOVA was run to (a) check the manipulation (version: VAL_POS vs. VAL_NEG), (b) ensure that both articles (topic: NAS vs. EDU) were perceived similarly, and (c) ensure that the perceptions of the articles did not differ between countries (US vs. GER) with respect to text valence, relevance, and credibility.

The manipulation checks confirmed that VAL_POS was perceived as more positive than VAL_NEG. Next, results showed that the two versions (VAL_POS vs. VAL_NEG) and the two articles (NAS vs. EDU) did not differ from one another in relevance or credibility. Finally, people in both nations perceived the articles to be equally relevant. There was, however, a marginal influence of nationality on the overall credibility of the article, F(1, 145) = 2.85, p = .09, η_p² = .019, since German students rated the articles as slightly less credible (M = 3.54, SD = 0.16) than the U.S. students did (M = 3.90, SD = 0.13).

Main Study Procedure

The experiment was run as an online questionnaire on personal computers under controlled conditions. First, experimenters asked for the participants’ formal consent to take part in the study. Afterwards, they read the experiment’s instructions aloud. Participants were randomly assigned to one of the two experimental conditions. Participants first read a manipulated article, then a distractor article, and afterwards the other manipulated article. Participants were given either two positively valenced articles and the distractor article or two negatively valenced articles and the distractor article. The sequence of the topics (national education and national security) of the manipulated articles was randomized. After reading the articles, participants answered the main questionnaire.

Measures

For the psychometrics of the variables, please refer to Table 1. Table A1 in the Electronic Supplementary Material 1 (ESM1) provides a list of all of the items. To assess reliability, McDonald’s composite reliability omega was used (McDonald, 1999). Values above Ω = .60 indicate good fit (Bagozzi & Yi, 1988). In addition, as suggested by Eisinga, te Grotenhuis, and Pelzer (2013), reliability for two-item scales was assessed with the Spearman–Brown prophecy formula and for longer scales with Cronbach’s α. Results showed that the reliabilities were close to or above the recommended thresholds.

Table 1 Psychometric properties of the study variables

			Range					SEM
Variable	M	SD	Potential	Actual	Skew	Kurt	Rel (α, ρ)	Rel (Ω)	AVE
Note. AVE = average variance extracted. GER = Germany. aCronbach’s α. bSpearman–Brown ρ.
Positive distinctiveness
All	0.67	1.93	−6 – 6	−5 – 5	−0.37	−0.39	.66b	–	–
US	0.22	2.24	−6 – 6	−5 – 5	−0.15	−0.81	.79b	.87	.70
GER	0.89	1.73	−6 – 6	−4 – 4.5	−0.34	−0.34	.56b	.57	.41
News knowledge
All	0.78	0.20	0 – 1	0 – 1	−1.24	1.41	.71a	–	–
US	0.82	0.19	0 – 1	0 – 1	−1.25	1.82	.69a	.55	.42
GER	0.77	0.21	0 – 1	0.1 – 1	−1.23	1.17	.71a	.64	.50
Social comparison
All	5.02	1.30	1 – 7	1 – 7	−0.67	0.51	.71b	–	–
US	5.33	1.43	1 – 7	1 – 7	−0.93	0.73	.83b	–	–
GER	4.88	1.20	1 – 7	1 – 7	−0.64	0.54	.62b	–	–
National identity
All	5.63	0.83	1 – 7	1.25 – 7	−1.03	2.30	.64a	–	–
US	5.48	0.85	1 – 7	2 – 7	−0.87	1.20	.58a	.58	.26
GER	5.70	0.82	1 – 7	1.25 – 7	−1.12	3.02	.68a	.66	.36

Table 1 Psychometric properties of the study variables

View as image HTML

Data Analysis

All hypotheses were tested with multilevel structural equation modeling (SEM). One group consisted of the sample from the US and the other of the sample from Germany. In order to compare the two groups, we checked for measurement invariance according to the approaches recommended by Hirschfeld and von Brachel (2014).

The experimental manipulation was operationalized by means of an observed dichotomous variable (labeled “Article Valence” in Figure 1). Missing data were treated as missing at random and were replicated with the full information maximum likelihood approach. The research question was answered by comparing the confidence intervals of the effects (Masson & Loftus, 2003); 95% confidence intervals were estimated by means of bootstrapping standard errors with 1,000 draws.

Model Characteristics

The data fit the baseline model well: All indicators were above common thresholds as presented in the literature (see Table 2).

Table 2 Fit indices for the standard error of the mean values

		Models
Fit indices	Criteria	Baseline	Configural	Weak invariance	Strong invariance
Note. CFI = comparative fit index. GFI = goodness-of-fit index. TLI = Tucker–Lewis index. RNI = relative noncentrality index. RMSEA = root mean square error of approximation. SRMR = standardized root mean square residual.aByrne, 2010. bHu & Bentler, 1999.
χ2		37.82	62.57	68.47	97.15
df		31	62	67	72
p	> .05a	.19	.46	.43	.03
CFI	> .95a	.99	1	1	.96
GFI	> .90a	1	1	1	1
TLI	> .95a	.98	1	1	.95
RNI	> .95b	.99	1	1	.96
RMSEA	< .10a	.02	.01	.01	.04
SRMR	< .08b	.04	.04	.05	.06

Table 2 Fit indices for the standard error of the mean values

View as image HTML

To check for measurement invariance, we imposed several constraints on the model (Hirschfeld & von Brachel, 2014). Results showed that models with (a) configural, (b) weak, and (c) strong invariance constraints also fit the data well (Table 2). The model with weak invariance constraints was used to test the hypotheses because it provided the best way to compare potential effects between the two countries (Hirschfeld & von Brachel, 2014).

Next, we checked convergent validity. First, all items loaded significantly on their respective latent factor (see Figure 1). Second, four of six average variance extracted (AVE) values were below the recommended value of AVE = .50 (Fornell & Larcker, 1981; see Table 2). Thus, the AVE values showed that the measurement quality of the model was only moderate.

To analyze discriminant validity, we checked for substantial individual cross-loadings and error covariances. Analyses of the modification indices did not show significant cross-loadings, thus implying satisfactory discriminant validity.

Effects of News Article Valence on Positive Distinctiveness

Hypothesis 1 predicted that readers’ positive distinctiveness would be higher when they read news articles containing news that was positive with respect to their own social group (nationality) than when they read articles containing negative news about their social group. We tested Hypotheses 1 and 1.1 for the two subsamples to detect both the general psychological processes as predicted by the hypotheses and the national differences stated in RQ1. The data confirmed Hypothesis 1 for both countries: The results showed significant effects for the U.S. subsample (b = 4.34, 95% CI [3.71, 4.93], β = .83, p < .001, SE = 0.32) and the German subsample (b = 2.08, 95% CI [1.65, 2.53], β = .64, p < .001, SE = 0.22). This implies that participants who read a news article containing news that was positive with respect to their social group (nationality) showed more positive distinctiveness after reading the article as compared with respondents who read an article containing news that was negative with respect to their social group (nationality). For example, U.S. participants who read a news article that featured a positive story on the U.S. educational system afterwards believed that the US handled national educational system issues better than Germany did. For participants in both countries, the effect was strong (US: β = .83, GER: β = .64). With regard to RQ1, the effect of valence on positive distinctiveness existed for both subsamples. At the same time, the effect was larger in the U.S. subsample (b = 4.34, 95% CI [3.71, 4.93]) than in the German subsample (b = 2.08, 95% CI [1.65, 2.53]).

In Hypothesis 1.1, we controlled for the influence of a potential third variable. That is, no matter what kind of article people read, we expected that people with a strong national identity would show more signs of positive distinctiveness. The data did not confirm this hypothesis for participants in the US (b = 0.25, 95% CI [−0.42, 0.87], β = .07, p = 0.455, SE = 0.33) or for participants in Germany (b = 0.22, 95% CI [−0.20, 0.77], β = .10, p = 0.357, SE = 0.24). In sum, we did not find an influence of national identity on positive distinctiveness.

Effects of News Article Valence on the Knowledge Learned From the Articles

With Hypothesis 2, we proposed that readers would obtain more knowledge from news articles presenting a positive view of their own social group (nationality) than they would from articles presenting a negative view. Again, we tested Hypotheses 2 and 2.1 for both subsamples to detect not only the general psychological processes as predicted in the hypotheses but also national differences as suggested in RQ1. The data confirmed Hypothesis 2 for the German subsample (b = 0.12, 95% CI [0.06, 0.18], β = .33, p < .001, SE = 0.03) but not for the U.S. subsample (b = 0.05, 95% CI [−0.02, 0.12], β = .16, p = 0.177, SE = 0.04). In Germany, participants who read articles that presented a positive view of their own social group showed more knowledge about the articles after reading than respondents who read articles that presented a negative view of their own social group. In other words, German participants who, for example, read a news article that featured a positive story on Germany’s national security were likely to remember more facts from the article than German participants who read a news article that featured a negative story on Germany’s national security. The effect was moderate (β = .33).

In Hypothesis 2.1, we controlled for the influence of a potential third variable. That is, no matter what kind of article people read, people who compared their own social group (nationality) intensely with another social identity while reading a news article were expected to show more knowledge about the news article in general. In other words, we predicted that the intensity with which participants compared the two countries would influence knowledge acquisition while reading a valenced news article. The data supported Hypothesis 2.1 for the German subsample (b = 0.03, 95% CI [0.01, 0.06], β = .22, p = 0.028, SE = 0.01) but not for the U.S. subsample (b = 0, 95% CI [−0.03, 0.02], β = −.04, p = 0.755, SE = 0.01).

Additional Analyses

In a supplementary model, we evaluated whether the manipulation had an effect on the two independent variables social comparison and national identity. Results did not show a significant effect for both samples. In addition, the model did not show a significantly better fit (Δχ² = −1.53, p > .999). We also ran supplementary analyses in which we controlled all variables for gender effects. Likewise, we did not find significant effects.

Differences Between the US and Germany

To evaluate RQ1, we compared the confidence intervals of the unstandardized effects. The results showed that the German and U.S. samples differed in terms of the strengths of the effects for three of the four hypotheses: First, the effect on positive distinctiveness of positive news articles about the person’s own country (H1) was larger for the U.S. participants than for the German participants. Second, the influence of the text’s valence on factual news knowledge (H2) was larger for the German sample than for the U.S. sample. Third, the influence of comparison behavior on factual news knowledge (H2.1) was larger in the German subsample than in the U.S. subsample. Concerning the potential influence of national identity on positive distinctiveness (H1.1), we did not find any differences between the two countries.

Limitations

We chose the topics of the stimuli according to the interests of the age group under question. Further, we extensively pretested the stimulus material in terms of structure, wording, and length. However, we could not exclude the potential influence of other third factors such as the articles’ topics or rhetoric, because a certain rhetoric and topic might be associated with a certain valence. For instance, Americans might be used to a negative evaluation of their education system, whereas Germans might be used to reading that their standards of national security are inferior to the standards of the US.

Regarding methodological limitations, it can be argued that the U.S. subsample of 119 participants was too small. It is recommended that for model comparisons between groups, each group should ideally consist of at least 200 cases (Kenny, 2011). In addition, the likelihood of detecting significant effects increases with sample size (Field, Miles, & Field, 2012). Our results imply that the effects of the articles’ valence on knowledge were small to medium-sized. In order to show that small effects are statistically significant, when using a multivariate design with two predictors, sample sizes of approximately 600 cases are needed (Field et al., 2012). Future research might hence want to develop other experimental designs in which such large sample sizes can be obtained more easily.

Further, we compared samples from specific regions in both Germany and the US. Therefore, we were not able to generalize the results for an entire country or a specific culture. However, in order to ensure comparability, we tried to keep the differences between the German and U.S. samples as small as possible by using student samples in both countries.

Finally, the composite reliability and discriminant validity did not always meet the criteria that are recommended in the literature. This implies that further work would be helpful to develop scales that measure the employed variables with more reliability.

Future Perspectives

The findings stemming from both the empirical results and the theoretical considerations of how social identity processes interact with media effects might have an impact on how news reading and news making is considered in the future. For further research, it seems worthwhile to combine or compare different theoretical approaches such as stereotyping, priming, or even mood management with the social identity processes shown in this study (cf. previous section). In addition, it might be interesting to look at other formats such as entertainment media and how valenced entertainment media might influence both in-group evaluations and learning processes.

In terms of future directions for policy and news making, we see our results from a somewhat critical standpoint and have more open questions than answers to deliver. As demonstrated by this study, reading valenced news texts has manifold consequences. This study showed that a positive valence resulted in learning for the German subsample. This result is intriguing and somewhat frightening at the same time. We can think of two major questions that arise from this result. First, in a media environment – both in the US and in Germany – where political interest and the reading of political news is constantly declining (Pew Research Center, 2012; Shell Deutschland Holding, 2011), might it be the case that we need news that are more positively valenced toward the ingroup to make people pay attention? Should we suggest that the media offer news texts that are as positively valenced toward the ingroup as possible in order to foster people’s learning about international issues? A story that is positively valenced toward the ingroup might be easier to read because it is closer to the identity needs of a reader. Thus, one could harness these identity needs in order to leverage interest in and personal relevance of news texts, which could finally result in a better political understanding. However, one-sided news as a strategic rhetoric to initiate learning processes does not appear to be a reasonable solution. Even though learning processes occurred in the German subsample, it seems important to prevent the dissemination of only positively valenced and pronational views about international affairs. Moreover, as a second major question, do the results by contrast imply that we need to teach young adolescents to prefer news that is not positively valenced toward one’s own nationality? Can we teach readers to overcome the social–psychological principle of in-group favoritism? Such a noble aim seems unrealistic: Tajfel (1978) showed that simply by separating people into two groups according to their favorite painter, strong feelings of in-group favoritism and out-group derogation arose. And after all, both ways – offering positively valenced news to make people learn on the one hand and teaching them to learn without needing positively valenced information on the other – appear to be equally paternalistic approaches, and, as such, problematic.

In a final note, we can state that the German subsample remembered news text that were positively valenced towards Germany better, and that in both Germany and the US, news articles that were positively valenced towards the own country stirred in-group favoritism. This seems particularly important to understand when considering the history of Germany and our future reading of good news.