An Experiment on the Press Coverage of Child Sexual Abuse: Can Readers Differentiate Between Good and Bad Reporting?

ABSTRACT

News reporting on child sexual abuse (CSA) plays an important role in educating the public and fighting sexual violence, according to the public interest model of normative media theory. Bad reporting, however, is widespread and hinders a solution-oriented approach. Against this backdrop, the current study investigated which normative and subjective criteria are used by readers when they assess the quality of CSA newspaper reporting (RQ1). Furthermore, it was tested if readers can differentiate between good and bad CSA reporting quality (RQ2) and if their personal involvement in the topic—concerning victimization, exposure to CSA reporting, CSA knowledge—influences assessments of journalistic quality (RQ3). An experimental online study with a national quota sample of N = 2724 adults (18–65 years; M_age  = 44.1; 52.5% women) from Germany was conducted in 2020. The study is preregistered and further materials are shared on osf.io. It turned out that readers mostly used the normative criteria to assess CSA reporting quality that are suggested by the academic literature (RQ1). Readers were able to differentiate between CSA reporting with high versus low journalistic quality (RQ2)—irrespective of their own CSA victimization (RQ3). Readers rated bad reporting as mediocre, though, indicating potential unawareness of certain quality issues.

According to the public interest model of normative media theory, press articles on any social issue should demonstrate high journalistic quality and, hence, foster a rational solution-focused public debate about the respective issue for the benefit of society (McQuail, 1992, 2010). As child sexual abuse (CSA) is such a widespread and severe social issue (Barth et al., 2013), related press coverage has been the subject of critical assessment by communication scholars for quite some time (e.g., Cromer & Goldsmith, 2010; Ducat et al., 2009; Görgen & Fangerau, 2018; Görgen et al., 2013; Jones et al., 2010; Kitzinger, 2004; Weatherred, 2015). It has been shown by multiple media content analyses that a significant amount of reporting on child sexual abuse lacks journalistic quality, sensationalizes the topic, does not give a voice to survivors and experts, identifies victims, and omits relevant information on efficient prevention and intervention (Cheit et al., 2010; Jones et al., 2010; Mejia et al., 2012; Weatherred, 2015; Wolak et al., 2008).

While there is a growing consensus among experts on what makes good versus bad CSA reporting (Döring & Walter, 2020), the perception of the media audience is widely unknown. The question of whether news audiences are able to discern quality journalism is important for several reasons: From the perspective of journalism scholars, audiences’ ability to identify quality journalism contributes to the health of public discourse and informed civic engagement (McQuail, 1992, 2010). For professional journalists, journalistic integrity, credibility, and trust are deeply intertwined with the perceived quality of the information disseminated. When audiences have difficulty distinguishing between high-quality and low-quality journalism, it can undermine trust in the media (Kitzinger, 2004). Survivors of child sexual abuse and those deeply concerned about the issue expect and rely on effective communication about the issue. If audiences’ ability to discern journalistic quality in this context is compromised, it could hinder the dissemination of critical information and advocacy efforts and negatively impact societal awareness (Cheit et al., 2010; Wolak et al., 2008). Ideally, news consumers have developed news media literacy in the sense of knowing how to access, select, and critically evaluate news content (Ashley et al., 2017). But how exactly do readers evaluate the quality of newspaper articles about CSA? Do they possess the ability to differentiate between good and bad reporting according to journalistic standards? Could it even be that the audience appreciates bad reporting because sensationalized stories are more entertaining and cater to voyeuristic interests? Do quality assessments of CSA reporting differ between readers personally affected versus not affected by CSA?

The current study aims at answering these questions with the help of an experimental online study among N = 2724 adults ages 18 to 65 in Germany, of whom about 31.9% report having been personally affected by CSA to at least some degree. We first summarize the state of research and outline the research questions. Then we explain the methodology of the study before we answer the research questions. The discussion provides an interpretation of the main results, points out limitations of the current study and gives an outlook on future research and practice.

State of research

At first sight, journalistic quality is an abstract concept that might seem hard to grasp. So-called quality criteria were, hence, established to structure what defines good versus bad press reporting. Two types of quality criteria can be distinguished: normative quality criteria (NQC), which are derived from scholarship and rooted in academic theory, and subjective quality criteria (SQC) that are based on the preferences and taste of the audience.

Normative quality criteria are rooted in the public interest model of normative media theory and cover all reporting aspects that are necessary to establish and support the aforementioned solution-focused debate on a specific topic relevant to society (Christians et al., 2009; McQuail, 2010). Due to the significance of the free press and high-quality reporting on social issues for democratic societies, NQC have received considerable attention from researchers for decades, reaching from initial definitions of media accountability and responsibility toward the society (McQuail, 1992, 1997) to updated normative quality dimensions for news media in the twenty-first century (Karlsson et al., 2023). At its core, normative media quality encompasses a range of mechanisms and concepts aimed at ensuring accuracy, fairness, ethical standards, and further relevant aspects in news reporting. Accordingly, quality criteria can be either assigned, contracted, self-imposed, or denied depending on the issue being addressed (McQuail, 1997).

Despite this high interest in defining normative media quality in general, topic-specific quality models are still rare in this field of research. For the issue of child sexual abuse, 10 different normative criteria for CSA reporting with good journalistic quality have been established (Döring & Walter, 2020): NQC 1: Thematic framing, NQC 2: Non-sensational reporting, NQC 3: Use of appropriate terms, NQC 4: Inclusion of stakeholders, NQC 5: Non-stereotypical reporting, NQC 6: Inclusion of prevention and intervention, NQC 7: Ethical treatment of survivors in interviews, NQC 8: Lawful reporting, NQC 9: Balance of survivors’ and alleged perpetrators’ interests, NQC 10: Disclosure and reflection of official sources.

In contrast to normative criteria, the subjective quality criteria are not rooted in journalism theory, but in the media audience’s needs, preferences, and tastes. Research on SQC is rather limited, and some of the findings are contradictory. While some authors show that normative and subjective quality criteria for good journalism coincide (e.g., Costera Meijer, 2013), others show that they diverge (e.g., Gil de Zúñiga & Hinsley, 2013). To date, the academic literature has identified mainly five subjective quality criteria for news coverage (Jungnickel, 2011) that are applicable to CSA coverage: SQC 1: Personal relevance, SQC 2: Interest in the subject, SQC 3: Entertainment, SQC 4: Voyeurism, SQC 5: Follow-up communication.

Previous research has not yet explored how readers evaluate the overall quality of CSA reporting and which normative and/or subjective criteria they use in their evaluations.

It is also unknown if readers can clearly differentiate between CSA reporting with high versus low journalistic quality, which is largely dependent on the individual news media literacy of the readers. For this investigation, it is crucial to differentiate between two key concepts in media quality assessment: motivation and ability. While motivation reflects the intrinsic willingness to select and engage with media content, ability represents the capacity to critically assess and discern the nuances of journalistic quality (Ashley et al., 2017; Urban & Schweiger, 2014). The present study specifically focuses on the assessment of the audience’s ability. In journalism research, experimental studies are commonly employed to test the audience’s ability to assess journalistic quality, a major aspect of news media literacy. Previous studies have demonstrated that readers are able to differentiate good versus bad journalistic quality of newspaper articles, for example when it comes to cross-topic comparisons (e.g., sports-related versus political topics; Jungnickel, 2011), cross-media comparisons (e.g., newspaper press articles versus blog posts; Trepte et al., 2008), and cross-brand comparisons (e.g., reputable quality newspaper brand versus yellow press/tabloid brand; Urban & Schweiger, 2014; Voigt, 2016). Experimental studies examining audience evaluations of the press coverage on sensitive topics like domestic violence, suicide, and child sexual abuse are scarce, representing a research gap. Thus, it remains unclear to what extent readers can discern good from bad CSA reporting in an experimental context. The audience’s ability to detect bad CSA reporting based on recipients’ news media literacy (Ashley et al., 2017) is relevant because research reviews have shown that bad reporting is quite prevalent (Popović, 2018; Weatherred, 2017).

According to previous studies, personal involvement can help to better understand complex topics and hence to evaluate journalistic quality more competently (Jungnickel, 2011). This observation also aligns with the elaboration likelihood model (Petty & Cacioppo, 1979), which suggests that readers highly involved in a topic are more likely to closely investigate and examine related news articles, leading to better quality assessments. On the issue of CSA, involvement factors include personal victimization, a high exposure to CSA reporting in the press, and great topic knowledge. CSA survivors are assumed to be better in judging CSA reporting quality in comparison to non-affected people because of their personal experience with the issue and often having to deal with the consequences for a long time. Also, people who read a lot of news articles about CSA and know more about CSA in general might be better prepared to adequately evaluate CSA coverage quality.

Current study

As a dedicated study on the audience’s quality assessment of CSA press reporting has not been conducted so far, the current study was designed to answer, for the first time to our knowledge, the following three research questions.

The first research question is based on the framework of normative and subjective quality criteria of news reporting and draws on the ten NQC introduced by Döring and Walter (2020) and the five SQC based on Jungnickel (2011) presented above.

RQ1:

How do readers assess the overall quality of newspaper articles on CSA and which normative and subjective quality criteria play a role in this assessment?

The second research question addresses the audience’s ability to differentiate between good and bad CSA reporting when confronted with respective example articles that fulfill or neglect all normative criteria of journalistic quality mentioned above (Döring & Walter, 2020).

RQ2:

To what extent can readers distinguish between high and low quality of newspaper articles on CSA?

The third research question deals with the influence of personal involvement in this context: People who are personally affected by CSA, are more often exposed to CSA press reporting, or know more about the topic in general are more involved and might judge the reporting quality differently from less involved readers.

RQ3:

How does personal involvement (in terms of CSA victimization, exposure to CSA reporting, and CSA knowledge) influence the readers’ assessment of good or bad CSA reporting in newspapers?

Method

To answer the three research questions, we conducted an experimental online study. In the following sections we present the (1) stimulus material; (2) instrument and procedure, (3) sampling, data collection, and data cleaning, (4) sample description, and (5) statistical analysis. The study is preregistered and follows an open science approach: a corrigendum, all materials, instruments, data, and the analysis script are shared via the Open Science Foundation (OSF) server (https://osf.io/pwth5/). Prior to data collection, the study received ethical approval by the institutional review board of the Friedrich-Schiller-Universität Jena on July 17, 2020, file number FSV 20/026.

Stimulus material

For stimulus material we created two comparable newspaper articles about a typical, but fictional, CSA case: Sexual abuse of several young girls in a youth center in Germany. One newspaper article showed high journalistic quality and the other one low journalistic quality according to the framework of CSA-related quality criteria agreed upon by experts (Döring & Walter, 2020). The low-quality article violated all normative quality criteria while the high-quality article fulfilled them all. The stimulus articles were presented in German language in the design of a typical print newspaper (Scheufele, 2005). Figure 1 shows a side-by-side preview of the articles in English translation. The complete stimulus articles are available in Figures A1 and A2 in the Appendix.

Figure 1. Side-by-side preview of the stimulus press articles Q+ (high normative journalistic quality, left-hand side) and Q- (low normative journalistic quality, right-hand side).

The differences in journalistic quality between the two stimulus articles are pointed out in Table 1. The stimulus material was constructed based on an established quality model and checked by three independent experts from the field of journalism and child sexual abuse to ensure its validity.

Table 1. Experimental variation of the ten Normative Quality Criteria (NQC) in stimulus press articles (based on Döring & Walter, 2020).

Normative quality criteria	Stimulus Article Q+ (high normative journalistic quality)	Stimulus Article Q- (low normative journalistic quality)
NQC 1: Thematic framing	In the article, CSA is framed thematically as a topic that affects the entire society: “Crime statistics show that child sexual abuse occurs in all social classes and can affect the entire population.”	In the article, CSA is framed episodically as a tragic unfortunate event with no relation to society as a whole: “Such an incident has never happened here before. It is a tragic individual case.”
NQC 2: Non-sensational reporting	A non-sensationalist, objective, and fact-based reporting style dominates the article: “ … severe child sexual abuse in multiple cases over a five-year period … .”	A sensationalist, excessive, and attention-grabbing reporting style dominates the article: “It is an incident that takes the breath away of the citizens.”
NQC 3: Use of appropriate terms	In the article, appropriate terms are used for the survivors (“abused children“), the perpetrator (“47-year old man”), and the crime (“sexualized violence”).	In the article, inappropriate terms are used for the survivors (“violated girls“), the perpetrator (“monster”), and the crime (“indecent behavior”).
NQC 4: Inclusion of stakeholders	The article includes stakeholders relevant to society such as the police and the Independent Commissioner for CSA Issues in Germany.	The article does not include stakeholders relevant to society, but only purports a statement of the affected youth center.
NQC 5: Non-stereotypical reporting	The article dispels common CSA stereotypes, e.g., that survivors can never recover from CSA: “The psychological consequences for the victims vary from person to person and are usually long-lasting, but they can be treated and eventually overcome … .”	The article disseminates common CSA stereotypes, e.g. that survivors can never recover from CSA: “[The girl] cannot forget what happened, her childlike innocence has been lost forever since then.”
NQC 6: Inclusion of prevention and intervention	The article provides references to help and support services on CSA (website, telephone hotline).	The article does not provide references to help and support services on CSA.
NQC 7: Ethical treatment of survivors in interviews^a	not applicable	not applicable
NQC 8: Lawful reporting	In line with current media laws, the article does not disclose any names or other identifying attributes of the CSA survivors or the perpetrator.	Violating current media laws in Germany, the names of both the perpetrator (“Bernd S.”) and one of the CSA survivors (“Anna B.”) are disclosed in the article.
NQC 9: Balance of survivors’ and alleged perpetrators’ interests	In the article, survivors and the perpetrator are referred to in a non-judgmental manner (“suspected abused children” and “alleged perpetrator”), no pre-judgement is present. Instead, it is mentioned that the case will be decided at court.	In the article, the case is already judged and the perpetrator is one-sidedly pre-sentenced: “[The perpetrator] belongs behind bars for a long time.”
NQC 10: Disclosure and reflection of official sources	The articles refers to trustworthy official sources such as scientific studies and includes statements from the police and the Independent Commissioner for CSA Issues in Germany.	The article does not refer to further official sources apart from a single statement of the youth center’s manager, which is also not reflected critically.

^aNQC 7: Ethical treatment of survivors in interviews is a quality criterion of good CSA reporting that is not directly represented in the press article itself.

Instrument and procedure

The experiment was conducted as an online study. Participants first gave informed consent and then filled out the online questionnaire. As part of the experimental variation participants were randomly assigned one of the two stimulus articles that were either high or low normative journalistic quality as shown in Figure 1 above and Figures A1 and A2 in the Appendix.

Overall quality assessment

After reading the assigned stimulus newspaper article on the fictional case of child sexual abuse, participants gave their overall quality assessment (“How do you rate the quality of this article on child sexual abuse in general?”) on a 5-point Likert scale ranging from 1 = very low quality to 5 = very high quality.

Normative quality criteria

Then, participants rated the article on 10 different normative quality criteria (Döring & Walter, 2020) on 5-point Likert scales ranging from 1 = I don’t agree at all to 5 = I completely agree using the following statements:

• NQC 1: Thematic framing: “The information in the article is relevant to society.”

• NQC 2: Non-sensational reporting: “The article does not convey a sensational style.”

• NQC 3: Use of appropriate terms: “The information in the article is presented with adequate terms and in a way that is easy to understand.”

• NQC 4: Inclusion of stakeholders: “The article takes different opinions and sources into account.”

• NQC 5: Non-stereotypical reporting: “The article is plausible, well researched and most likely factually correct.”

• NQC 6: Inclusion of prevention and intervention: “Further support services are referenced in the article.”

• NQC 7: Not applicable

• NQC 8: Lawful reporting: “The article respects the privacy and honor of the people that are reported about.”

• NQC 9: Balance of survivors’ and alleged perpetrators’ interests: “The article is neutral and non-judgmental.”

• NQC 10: Disclosure and reflection of official sources: “It is clear where the information mentioned in the article comes from.”

Subjective quality criteria

Finally, participants evaluated the stimulus article on five subjective quality criteria (based on Jungnickel, 2011) on 5-point Likert scales ranging from 1 = I don’t agree at all to 5 = I completely agree using the following statements:

• SQC 1: Personal relevance: “The subject of the article is relevant to myself.”

• SQC 2: Interest in the subject: “I find the topic of the article interesting.”

• SQC 3: Entertainment: “The article is thrilling.”

• SQC 4: Voyeurism [self-constructed]: “I am fascinated by the detailed description of the course of the crime in the article.”

• SQC 5: Follow-up communication: “The article gives me material for conversations with family and friends.”

Participants’ sociodemographic and personal involvement

Participants provided information about five sociodemographic variables used as quota variables for the sampling process: (1) gender, (2) age, (3) federal state of Germany, (4) school education/vocational training, and (5) family status.

Furthermore, three variables of personal involvement in the topic of CSA were measured. CSA victimization was measured using the Sexual Abuse 5-item Subscale of the Childhood Trauma Questionnaire (CTQ) after Häuser et al. (2011; updated version: Klinitzke et al., 2012). The scale was calculated as a sum score and shows excellent reliability for the present sample: α = .94, GLB_a = .96 (Trizano-Hermosilla & Alvarado, 2016).

Exposure to CSA reporting was measured with the item “How often have you come into contact with mass media reporting about child sexual abuse so far?” with response options ranging from 1 = very rarely to 5 = very often.

Knowledge about CSA was measured with 11 items of the Child Sexual Abuse Myth Scale (CSAMS) after Collings (1997; German translation: Bienstein et al., 2019). The scale was calculated as a mean score and showed excellent reliability for the sample: α = .88, GLB_a = .90.

A more detailed introduction to all variables included in the statistical analysis, along with a manipulation check, is provided as online supplementary material on OSF (https://osf.io/bxqsv/). For the manipulation check, nine independent-samples t-tests were used to check for mean differences between the two experimental groups Q+ and Q- concerning the normative quality criteria. All t-tests yielded significant results with higher means for group Q+, indicating a successful experimental manipulation of the normative quality criteria.

Sampling, data collection, and data cleaning

An a-priori power analysis was conducted to determine the optimal sample size using the G*Power software in version 3.1.9.7 for both planned statistical procedures: hierarchical linear regression and 2 × 2 between-subjects ANCOVA. The calculation for the hierarchical regression model resulted in the largest required sample size of N = 2160 (test family: F tests; statistical test: linear multiple regression: fixed model, R² deviation from zero; small effect size of Cohen’s f² = 0.02; number of predictors = 18; Bonferroni-corrected α‘ = .003; β = 0.05). To accommodate for possible missing values, we aimed at a gross sample of N = 3000 participants.

The sample was collected using an online access panel provider. Based on the best4planning 2019 quota plan, an uncrossed quota sample for the Internet population of Germany age 18 to 65 years was drawn based on five sociodemographic quotation variables: (1) gender, (2) age groups, (3) federal states of Germany, (4) school education and vocational training, and (5) family status.

The data collection took place in summer 2020. Participants who completed the study were paid a compensation of 1€ by the online panel provider. After data collection, the COMR completion rate for online study according to American Association for Public Opinion Research (Ed.) [AAPOR], (2016) was calculated. A satisfactory COMR of 26.7% was achieved for the study, indicating an acceptable response rate. A gross sample of N = 3111 cases was drawn in total and afterward cleaned, leading to a cleaned net sample of N = 2724 (387 exclusions in total, 12.4%). Data cleaning considered the following factors: (1) system errors, (2) study processing time, and (3) questionnaire evaluation. First, erroneous cases with only quotation variables being present but no contentual variables and other system errors were excluded (n = 4).¹ Secondly, cases with an unrealistically low study processing time, below five minutes, were excluded (n = 300). Thirdly, cases who reported that they had used a search engine during the study or had not honestly answered the questionnaire were excluded (n = 83). A subsequent quota analysis confirmed very little bias due to the data cleaning process.

Sample Description

Table 2 presents the sociodemographic characteristics of the cleaned net sample (N = 2724). On average, participants took M = 13:09 minutes, SD = 08:58 to complete the study. Just over half, 52.5%, of the participants were female and the mean age was M = 44.1 years, SD = 13.1. According to the quota plan, most respondents were sampled from the denser populated Western federal states of Germany (84.8%). School education and vocational training was balanced with no general school certificate labeled as low (29.4%), middle school or secondary school certificate labeled as medium (32.0%), and A-levels, finished studies labeled as high (38.5%). Additionally, most participants were married (56.2%).

Table 2. Sociodemographic characteristics of participants.

Variables	n	%
Gender
Male	1294	47.5
Female	1430	52.5
Age groups^a
18 to 29 years	547	20.1
30 to 39 years	476	17.5
40 to 49 years	606	22.3
50 to 59 years	749	27.5
60 to 65 years	346	12.7
Federal states of Germany
Old (Western) federal states	2310	84.8
New (Eastern) federal states	414	15.2
School education and vocational training^b
Low	801	29.4
Medium	873	32.0
High	1050	38.5
Family status
Single/unmarried	900	33.0
Married	1532	56.2
Divorced/widowed	292	10.7

N = 2724. ^aAge: M = 44.1 years, SD = 13.1, Mdn = 46. ^blow education: no general school certificate (yet); medium education: middle school or secondary school certificate; high education: A-levels, finished studies. Due to the rounding method, the presented percentages may not add up to exactly 100%.

Table 3 provides further insights into the personal involvement on the topic of CSA for the cleaned net sample. The most recent population-representative study for Germany from 2011 reports a prevalence of about 13% for child sexual abuse across all degrees of severity (Häuser et al., 2011). In our sample, 31.9% of the participants reported some degree of child sexual abuse experience during childhood or youth. This difference in prevalence can be attributed to two main effects. First, participants may have been more willing to disclose past abuse than those surveyed in the 2011 report. Since 2010, the German government has implemented several ongoing, large-scale measures to encourage breaking the silence around CSA and to support victims (The Independent Inquiry into Child Sexual Abuse in Germany, 2020). Second, self-selection bias may have increased the number of CSA survivors among the invited panelists who were particularly interested in participating in this CSA-related study, thus overrepresenting CSA survivors in our sample. The relatively large number of CSA survivors within the sample enabled us to perform comparative analyses between victims and non-victims with adequate statistical power. About one third of the sample reported being frequently exposed to CSA media coverage, and another 43.5% at least sometimes. Overall, most of the participants have high knowledge about CSA (86.5%).

Table 3. Participants’ personal involvement factors in the topic of CSA.

Variables	n	%
CSA victimization^a
Yes	846	31.9
No	1806	68.1
Exposure to CSA reporting^b
Very rarely – rarely	543	21.8
Sometimes	1083	43.5
Often – very often	864	34.7
CSA knowledge^c
Very low – low	44	1.6
Medium	324	11.9
High – very high	2356	86.5

Varying sample sizes because of missing values/prefer not to say. Due to the rounding method, the presented percentages may not add up to exactly 100%.

^aAssessment of CSA victimization with subscale Sexual Abuse of the CTQ (Häuser et al., 2011; Klinitzke et al., 2012). Scale reliability: α = .94, GLB_a = .96.

^bValue range of exposure to CSA reporting: 1: very rarely − 5: very often (M = 3.2; SD = 1.0).

^cAssessment of CSA knowledge with the CSAMS (Collings, 1997; German translation: Bienstein et al., 2019). Scale reliability: α = .88, GLB_a = .90).

Value range: 1 = very low myth acceptance to 5 = very high myth acceptance (M = 1.7; SD = 0.7).

Statistical Analysis

Data cleaning and statistical data analysis were performed using the free statistical programming language R version 4.1.3 and the development environment RStudio version 2022.02.0 Build 443. A total of 21 different CRAN packages were used for the data analysis in this study. To counteract alpha error cumulation, a Bonferroni correction of the alpha error level based on the hierarchical regression and ANCOVA was applied to the inferential statistics (α‘ = .05 / 20 = .003).

To answer RQ1, a linear hierarchical regression was chosen to explain the audience’s overall quality assessment on CSA press articles. Because regression models can only be calculated with complete cases in which all model variables are present, the data set was filtered, which led to n = 2059 complete cases. The criterion to be explained is the overall quality assessment of the either high- or low-quality press article. The regression model was first cleaned by introducing a block of four confounding variables, including the experimental variation and the three personal involvement factors to obtain a more generalized perspective on the audience’s quality assessment. Afterward, the blocks with the normative and subjective quality criteria were introduced. Six statistical assumptions and goodness-of-fit criteria were tested for the regression model and were all sufficiently fulfilled: (1) linearity of the dependent variable/the criterion, (2) independence of errors/no autocorrelation, (3) homoscedasticity, (4) multicollinearity, (5) normal distribution of errors, and (6) outliers (Field, 2018; Field et al., 2014).

For RQ2 and RQ3, a 2 × 2 between-subjects analysis of covariance (ANCOVA) and a moderator analysis were conducted. Like in the regression analysis, a filtered data set was created, resulting in n  = 2353 complete cases. The dependent variable was again the overall quality assessment of the stimulus press article. Two binary independent variables were selected to investigate relevant group differences in the quality assessment. Firstly, the experimental variation with two levels for article Q+ with high normative journalistic quality and article Q- with low normative journalistic quality was investigated. We included CSA victimization as a second experimental factor (two levels: affected, non-affected) because CSA survivors represent an important stakeholder group in the public discourse on the topic that has not yet been explored in terms of their assessment of the quality of CSA reporting. Due to the resulting unequal sample sizes in the four cells of the experimental design, an ANCOVA Type II was chosen to appropriately calculate the sum of squares (Tabachnick & Fidell, 2019, pp. 182–184). The personal involvement factor exposure to CSA reporting was only used as a covariate to not overload the experimental design.

While checking the statistical assumption (1) homogeneity of regression slopes, we found that the personal involvement factor CSA knowledge is not a covariate as initially anticipated but takes the role of a moderator. This variable was therefore investigated separately with a moderator analysis. The other five statistical assumptions for the ANCOVA were sufficiently fulfilled: (2) normal distribution of the dependent variable, (3) normal distribution of residuals, (4) variance homogeneity of residuals, (5) linearity of the dependent variable, and (6) independence of covariates from treatment effects/factors (Field, 2018; Field et al., 2014).

In addition to the ANCOVA, a moderator analysis for the factor experimental variation and CSA knowledge was carried out. To further investigate and interpret this effect, the Johnson-Neyman procedure was applied to calculate the according Johnson-Neyman interval (D’Alonzo, 2004).

Results

Results on RQ1

For RQ1 regarding the overall quality assessment of CSA reporting and the underlying normative and subjective quality criteria, we conducted a hierarchical regression analysis. The regression analysis explains how well the different normative and subjective quality criteria can predict the readers’ overall quality rating of the presented CSA newspaper articles. In order to obtain a more generalized perspective independent from the quality of the stimulus article and personal involvement, the effects of the experimental variation and the three personal involvement factors were included as confounding variables as shown in Table 4. All three blocks of variables of the regression model significantly contribute to the quality assessment.

Table 4. Hierarchical regression analysis of normative and subjective quality criteria on the overall quality assessment of experimental CSA press articles.

Blocks and predictor variables	b	SE B	sr	β	R^2	ΔR^2
(Intercept)	0.59	0.09*
Block 1: Confounding variables					.16*	-
1. Experimental variation	0.06	0.04	.02	.03
2. CSA victimization	< −0.01	<0.01	−.01	−.01
3. Exposure to CSA reporting	0.02	0.01	.02	.03
4. CSA knowledge	<0.01	0.02	<.01	<.01
Block 2: Normative quality criteria from the perspective of science, journalism and CSA survivors					.55*	.39*
NQC 1: Thematic framing	0.05	0.02	.04	.06
NQC 2: Non-sensational reporting	0.09	0.02	.08*	.12*
NQC 3: Use of appropriate terms	0.06	0.02	.04*	.07*
NQC 4: Inclusion of stakeholders	0.06	0.02	.05*	.07*
NQC 5: Non-stereotypical reporting	0.27	0.02	.18*	.31*
NQC 6: Inclusion of prevention and intervention	0.01	0.02	.01	.02
NQC 7: Ethical treatment of survivors in interviews^a	-	-	-	-
NQC 8: Lawful reporting	0.03	0.02	.03	.04
NQC 9: Balance of survivors’ and alleged perpetrators’ interests	0.06	0.02	.05*	.09*
NQC 10: Disclosure and reflection of official sources	0.04	0.02	.04	.05
Block 3: Subjective quality criteria of the media audience					.56*	.01*
SQC 1: Personal relevance	−0.01	0.01	−.02	−.02
SQC 2: Interest in the subject	−0.03	0.02	−.02	−.03
SQC 3: Entertainment	0.09	0.02	.07*	.10*
SQC 4: Voyeurism	0.05	0.02	.04*	.06*
SQC 5: Follow-up communication	0.01	0.02	.01	.01

n = 2059, df = 2040. For their quality assessment of CSA reporting, participants were randomly presented an experimentally varied stimulus press article covering a case of CSA in either high or low normative quality (experimental groups Q+ and Q-).

*Significant on Bonferroni-corrected alpha error level of α‘ = .003.

^aNQC 7 is not related to the press article content but its production process. Therefore, it was not varied in the stimulus articles and not rated by the participants.

The block of confounding variables was introduced first to uncover the net effects of the following contentual blocks. Four confounders explained 16% of the variance, with no single predictor being statistically significant.

Afterward, the normative quality criteria were introduced as a block and explained 39% of the variance of the overall quality assessment. Five out of the 10 predictors in this block were statistically significant: non-stereotypical reporting, non-sensational reporting, the use of appropriate terms, the inclusion of stakeholders, and a balance of survivors’ and alleged perpetrators’ interests. Non-stereotypical reporting was the predictor with the largest effect not only for the block of normative quality criteria, but the entire regression model.

The block of subjective quality criteria was introduced into the hierarchical regression model last because it is the exploratory part of this regression. Despite only explaining 1% of the variance, this block also contributes significantly to the explanation of the overall quality rating of CSA press articles. The two items entertainment and voyeurism were significant as predictors in this block. Overall, the subjective quality criteria, 1% of explained variance, were much less important for the audience’s quality assessment than the normative quality criteria, 39% of explained variance.

Results on RQ2 and RQ3

For RQ2 and RQ3, we first conducted a 2 × 2 between-subjects ANCOVA to examine the readers’ ability to discriminate the normative media quality of the CSA press articles in the experiment as shown in Table 5. Factor 1, the experimental variation, turned out to be statistically significant: F(1, 2348) = 376.57, p < .001, η²_part = .138. Regarding RQ2, this confirms that the media audience is indeed able to distinguish the normative quality between the stimulus articles Q+ and Q-.

Table 5. Analysis of covariance (ANCOVA) for overall quality assessment of experimental CSA press articles.

Main effects and interaction terms	SS	F(1, 2348)	p	η^2part
Factor 1: Experimental variation	268.94	376.57	<.001*	.138
Factor 2: CSA victimization	1.88	2.64	.104	.001
Covariate: Exposure to CSA reporting	12.49	17.49	<.001*	.007
Factor 1 × Factor 2	3.90	5.47	.019	.002

n = 2353. For their quality assessment of CSA reporting, participants were randomly presented an experimentally varied stimulus press article covering a case of CSA in either high or low normative quality (experimental groups Q+ and Q-). ANCOVA Type II was used to calculate the sum of squares. The moderator CSA Knowledge was investigated separately in a follow-up moderator analysis.

*Significant on Bonferroni-corrected alpha error level of α‘ = .003.

A look at the mean values supports this finding as shown in Table 6. While the stimulus article Q+ with high normative quality was rated with M  = 3.71, SD = 0.74, article Q- was rated significantly lower at M  = 2.93, SD = 0.98, on the five-point Likert scale 1 = very low quality to 5 = very high quality.

Table 6. Descriptive statistics for ANCOVA on overall quality assessment of experimental CSA press articles.

	Factor 1: experimental variation
	Q+ (n = 1175)	Q- (n = 1178)
Factor 2: CSA victimization	M (SD)	M (SD)
yes (n = 773)	3.71 (0.74)	2.93 (0.98)
no (n = 1580)	3.68 (0.74)	3.06 (0.93)

n = 2353. Value range of the dependent variable overall quality assessment: 1 = very low quality to 5 = very high quality.

Regarding RQ3, Factor 2, CSA victimization, was not significant, meaning that both victims and non-victims are equally able to evaluate the quality of media coverage of CSA. Accordingly, looking at the means shows only negligible and insignificant differences. As there was no significant interaction term between the two main factors, they can be interpreted without taking the other factor into account. Exposure to CSA coverage was significant as a covariate, albeit with a very small effect: F(1, 2348) = 17.49, p < .001, η²_part = .007.

Since CSA knowledge turned out to be a moderator rather than a covariate after analyzing the statistical assumptions of the ANCOVA, we conducted an additional moderator analysis using the Johnson-Neyman procedure. A significant interaction term between the experimental variation and the CSA knowledge provides evidence of the moderation effect on the overall quality ratings: t(2349) = −8.23, p < .001, β = −21, sr = -.15. The Johnson-Neyman interval, (1.06, 2.40) for α’ = .003, n = 2353, indicates three zones. In zone 1 with high CSA knowledge, the media quality can be distinguished correctly, meaning that article Q+ was rated significantly better than Q-. Zone 1 included 92.3% of all cases. In zone 2 with a medium level of CSA knowledge, 7.2% of the sample, media quality could no longer be distinguished between the two press articles. In zone 3 with low knowledge about CSA—only 0.5% of the sample, indicative of mostly outliers—article Q- with the low normative quality is even rated significantly better than Q + .

Discussion

The results of the present study extend the state of research on journalistic quality research by examining audience evaluations of CSA press coverage in the following ways.

RQ1 examined how readers generally assess the overall quality of a CSA newspaper article and what types of quality criteria play a role in this assessment. Consistent with previous research, both normative and subjective quality criteria are relevant for audience quality assessment (Jungnickel, 2011; Urban & Schweiger, 2014). Normative quality criteria explained 39% of the variance in the quality judgments, showing that lay judgments are largely consistent with expert evaluations (Costera Meijer, 2013). Non-stereotypical reporting was the normative quality criterion that best predicted the audience’s overall quality assessment of CSA reporting. Subjective quality criteria are also relevant but play a much less important role with only 1% explained variance. Nevertheless, media audiences seem to value certain elements of entertainment and voyeurism in CSA reporting, probably due to the excitement and curiosity associated with crime reporting.

With respect to RQ2, our study shows that readers are indeed able to discriminate between high and low normative quality in CSA-related newspaper articles. This finding is consistent with the current state of research investigating differences in the quality assessment of newspaper articles about reputable brands versus yellow press brands (Urban & Schweiger, 2014; Voigt, 2016). However, looking at the means of the factor experimental variation, the participants rated the quality of the press article with low normative quality as mediocre rather than really bad, indicating a positive bias in their evaluation. Two possible explanations for this positive bias seem plausible but need further confirmation. First, some violations of normative quality criteria (e.g., sensational reporting with juicy details of the crime) serve an entertainment value of CSA coverage that some readers evaluate positively due to sensation-seeking and voyeurism. Second, some violations of normative quality criteria (e.g., use of inappropriate terms such as “monster” or “devil” to refer to perpetrators) appeal to the moral outrage of some readers and may therefore be viewed as appropriate and positive.

RQ3 focused on the influence of personal involvement factors on readers’ quality ratings. Regarding CSA victimization, no significant difference in the quality ratings was found between CSA victims and non-victims, meaning that both groups are equally able to evaluate the media quality of CSA press articles. Exposure to CSA coverage was found to be a rather negligible covariate with only a small impact, and thus it does not play an important role in the quality ratings of participants. Finally, knowledge about CSA plays a crucial moderating role in the assessment of CSA reporting quality. Participants with high knowledge of the topic were able to discern the difference, while medium to low knowledge of CSA impaired the ability of media audiences to correctly assess the overall quality of CSA reporting. From a normative perspective, judging the quality of an object always requires extensive background knowledge on the respective object. Accordingly, this also applies to the media audience and is in line with the state of research on both news media literacy (Maksl et al., 2015) and subjective news media quality criteria (Voigt, 2016). Beyond the pronounced and clear role of knowledge in evaluating news media quality, our heterogeneous findings on personal involvement factors underscore the need for further research to gain a more nuanced understanding of their various influences.

Strengths and limitations of the study

The present study provides new insights into an understudied topic by first investigating how readers make judgments about the media quality of CSA press articles and, in addition, exploring differences in these quality judgments based on personal involvement factors. It draws strength from its solid theoretical foundation by using an established model of normative quality criteria. The two self-designed stimulus articles based on the same case were systematically varied across all quality criteria to ensure a high degree of comparability.

However, the study has several limitations. The non-representative quota sample focused on Internet users and the disproportionate number of CSA survivors in the sample do not allow for a generalization of the results to the general population in Germany. Due to research economics and study design, only two exemplary press articles of low and high journalistic quality could be compared. How the general public evaluates medium-quality press coverage of CSA therefore remains an open question. For our stimulus articles, we created a fictional case of girls being abused by a single male perpetrator in a public youth center. Other CSA case scenarios involving intra-familial abuse and different institutions or survivor-perpetrator constellations might also have different effects on the audience’s assessment of the quality of press coverage, which should be investigated in future research. With a German audience in mind, the high-quality article Q+ of this study was designed to respect German legal requirements and journalistic practices, including claims of attribution. These environmental conditions might differ for other countries and their national media audiences.

Outlook on future research and practice

Our study shows that normative quality criteria are not only relevant from a scientific point of view but are also largely adopted and applied by readers to evaluate the quality of the CSA press coverage. Sensation-seeking and voyeuristic details play only a minor role for the audience when judging CSA press articles. According to communication experts, journalists should strive to improve the quality of their coverage and thus contribute to the public debate on CSA. Our findings indicate that readers can recognize and appreciate these improvements. Further educating the public on the issue of CSA could also raise the general awareness for media quality and adequate reporting. This would allow readers to deny attention and monetarization to sensational and stereotypical media coverage that does not reflect the empirical reality of CSA.

In the present study, we deliberately excluded images from the audience’s quality assessment and focused the experiment exclusively on text-based press coverage. Future research could also investigate the possible effects of CSA iconography and commonly used stock photos (Döring & Walter, 2021). Additionally, other audiovisual news formats such as radio news, TV news, and online reporting should be investigated. The implications of CSA news media literacy for CSA media selection and CSA media effects merit further research.

Open scholarship

This article has earned the Center for Open Science badges for Open Data, Open Materials and Preregistered. The data and materials are openly accessible at https://osf.io/pwth5/, and https://osf.io/n2v7m.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Nicola Döring

Nicola Döring (Ph.D.) is professor of Media Psychology and Media Design in the Department of Economic Sciences and Media at Technische Universität Ilmenau. Her research interests include media psychology, communication science, and sex research.

Roberto Walter

Roberto Walter (Ph.D.) was a doctoral researcher in the Department of Economic Sciences and Media at Technische Universität Ilmenau. His research interests include health communication, journalistic quality, and quantitative methods.

Notes

¹ Contentual describes a content-related variable, as opposed to a formal variable.

Appendix

Figure A1. Stimulus article Q+: high normative journalistic quality.

Figure A2. Stimulus article Q-: low normative journalistic quality.