Celebrity Suicide and Forced Responsible Reporting in the Nineteenth Century: Crown Prince Rudolf and the Absence of a Werther Effect

ABSTRACT Crown Prince Rudolf, heir to the Imperial throne of the Austro-Hungarian Empire, died by suicide in 1889. His death is a prime example of a historical celebrity suicide. Although news coverage about celebrity suicide has been shown to be linked to increases in suicides – a phenomenon known as the Werther effect –, censorship and/or journalists’ anticipatory obedience back then may have led to a form of “forced responsible reporting” that may have prevented a Werther effect. A content analysis was conducted, and civil death registers were used to identify suicides before and after Rudolf’s suicide. We compared Rudolf’s case with another historical celebrity suicide case (Colonel Redl) for which there is already empirical evidence consistent with a Werther effect. As expected, the press heavily reported on Rudolf’s death, but did not give undue prominence to suicide and rarely gave details on the method. Importantly, there was no evidence of an increase in suicides. This is in stark contrast to the Redl case in which the press reported irresponsibly. The Rudolf case emphasizes the importance of current media guidelines on responsible reporting. Thus, a high amount of news coverage does not necessarily translate into a Werther effect.

There is strong evidence for the existence of the so-called Werther effect, which is conceptualized as a media-induced increase in the number of suicides (Phillips, 1974;Pirkis & Blood, 2001). In particular, the news coverage of celebrity suicides has received intense scholarly attention (Stack, 1987). Celebrity suicide news coverage often gives undue (sensationalized) prominence to suicide and provides vivid details on the suicide method -elements of news coverage that contribute to the Werther effect (see Niederkrotenthaler et al., 2012Niederkrotenthaler et al., , 2020, for meta-analytic evidence). A recent example of a celebrity suicide that elicited an increase in the number of suicides in the immediate aftermath of the event was the death of the actor Robin Williams (Fink et al., 2018).
It is important to note, however, that there are also instances of celebrity suicide that do not seem to have elicited a Werther effect, for example, as documented for the Seattle area following the suicide of Kurt Cobain (Jobes et al., 1996). The absence of a Werther effect was partially explained by the press' adherence to media guidelines for responsible reporting on suicide (RRS, see Bohanna & Wang, 2012) -for example, no undue prominence of suicide and no details on the method employed. However, when taking a more nuanced look, we find that the quality of news coverage for the Cobain case seemed to have made a big difference: In France, where a significant subsequent rise in suicide rates was observed, RRS guidelines were not followed (Queinec et al., 2011). Reporting in the Seattle area -albeit that there was extensive front page news coverage -was much more consistent with media guidelines (Jobes et al., 1996). In short, the adherence to the concept of RRS in this case may be one reason why there was no increase in the suicide rate in the Seattle area. Consistent with this explanation, beneficial effects arising from adherence to media guidelines have been documented by other researchers, for example, by Etzersdorfer and Sonneck (1998), who found that subway suicides in Vienna dropped substantially after media guidelines were communicated to journalists (see also Bohanna & Wang, 2012;Pirkis et al., 2006).
Unfortunately, previous research on the role of the media in the suicide domain has almost exclusively focused on instances of (celebrity) suicide that elicited a pattern of increasing suicide numbers consistent with the Werther effect. Less is known about celebrity suicide that is not related to an increase in the number of suicides. As documented for the Cobain case, a thorough understanding of this substantially adds to our understanding of the news media's role in the suicide domain. The present study contributes to this underrepresented line of research.
As health communication researchers using a historical perspective on the "media and suicide" field, we introduce Rudolf, Crown Prince of the Austro-Hungarian Empire, as a 19thcentury celebrity who died by suicide on January 30 in 1889. We carefully speculated that his suicide, just like Cobain's, might not have elicited a (substantial) Werther effect despite extensive reporting. Of course, we assumed we would find a high level of news coverage about Rudolf's death due to the fact that he was the Crown Prince of the Empire. However, we hypothesized that the press (more or less) adhered to (some) principles of current media guidelines on RRS -although for different reasons. Censorship and/or journalists' anticipatory obedience may have led to a form of what we call "forced responsible reporting" that may have prevented the Werther effect.

Forced responsible reporting
We illustrate the concept of forced responsible reporting in detail in the context of Rudolf's death: Numerous conspiracy theories circulated about the circumstances of his death (Bled, 2006); however, we know now that Rudolf had previously expressed suicidal thoughts and had finally ended his life in a suicide pact with his extramarital lover, Mary Vetsera, in Mayerling, a small town near Vienna, then the capital of the Austro-Hungarian Empire (Hamann, 1988). When Rudolf was found dead accompanied by his mistress, everyone involved had to remain resolutely silent about her presence at the scene (Hamann, 2005;Ringel, 1989). A cover-up story deemed necessary by the Imperial Court (Bled, 2006) and the authorities had the press report that Rudolf had died of a stroke (Hamann, 2005). Rumors started to spread, and some foreign newspapers reported on all sorts of opposing explanations for Rudolf's death. At the time, public defamation of the emperor or his family was a prosecutable offense in the Austro-Hungarian Empire (Czech, 2010, p. 101) and it was still obligatory to send one copy of any publicized newspaper issue to the authorities upon distribution of the issue (pp. 315). Therefore, some domestic newspapers that stated murder and/or suicide were confiscated as soon as they were distributed (Hamann, 2005, p. 472;Bled, 2006, p. 216) and officials were ordered to prevent the illegal spread of these pamphlets with due rigor (Czech, 2010, p. 331).
Once the Emperor had learned about the real cause of death, amid growing public concern, he made the autopsy findings public, albeit ensuring they mentioned deformations of the brain to justify a proper Catholic burial (Ringel, 1989). Although suicides had not been prohibited by law since 1850, suicide was still frowned upon, especially within the Catholic Church, where most clerics did not offer a proper Catholic burial unless the suicide could be explained by reason of insanity or mental illness, often declared posthumously by stating that the individual had a deformation of the brain (Kuttelwascher, 1912;Ortmayr, 1990).
While the intention was not to prevent copycat suicides, the press may still have predominantly adhered to (some) principles of the current RRS standards to avoid prosecution or confiscation of the paper. This may have contributed to the prevention of imitation suicides (i.e., no Werther effect). To test this idea in the present study, we first conducted a content analysis (study 1) that investigated elements of newspaper reports of Crown Prince Rudolf's suicide that are known to be important factors in Werther effect research (e.g., giving undue prominence to suicide, details about the method). We cautiously speculated that the mere lack of Werther-related elements in the news coverage may have sufficed to prevent (at least) a (strong) rise in suicides following Rudolf's death. Therefore, in a second step, we conducted archival research (study 2) to test whether there was an increase in the number of suicides following Crown Prince Rudolf's suicide.

Study 1: Content analysis
We conducted a content analysis of the news coverage about Crown Prince Rudolf's suicide by focusing on two elements that are already known to be contributing factors for Werther effects (see above): giving undue prominence to the suicide in news reports and providing details about the method. Our approach is built upon one important methodological decision: Due to the fact that a study on the press' portrayal of Crown Prince Rudolf's death alone would lack explanatory power, and to add a contrasting aspect, we compared the Rudolf case with another Viennese celebrity suicide from around the same time that is already known to have generated a high amount of news coverage but did not adhere to modern standards of RRS. Austrian military Colonel Alfred Redl died of suicide on May 25, 1913, after being confronted and eventually admitting to betraying his homeland and spying on the Austro-Hungarian Empire. Importantly, previous research already documented increasing suicide numbers in the immediate aftermath of his death, a pattern that is consistent with a Werther effect (Arendt & Mestas, 2021). Therefore, study 1 involved a content analysis of both the Rudolf and Redl cases. We assumed that both celebrity suicides received a high amount of news coverage but that only the Rudolf case was consistent with the concept of "forced responsible reporting."

Method
We content-analyzed three different types of newspapers per case, each including one representative of the liberal press (Neue Freie Presse), the local papers (Neues Wiener Tagblatt), and the (developing) mass press (for Rudolf we investigated the Illustrirte Wiener Extrablatt and for Redl we examined the Illustrierte Kronen-Zeitung -the latter, very important newspaper first appeared on January 2, 1900, and thus was not available for Rudolf). For each celebrity suicide, we contentanalyzed a baseline period (before) and a follow-up period (after the celebrity's suicide). Since Crown Prince Rudolf's suicide occurred on January 30, 1889, we analyzed January 20-February 19, 1889, and since Colonel Redl's suicide occurred on May 25, 1913, our period of interest was May 15-June 14, 1913.

Variables
We developed a measurement instrument that focused on important elements found in modern-day media guidelines (see Bohanna & Wang, 2012). The coding unit was one issue of a given newspaper that appeared on a given day during the observation period. We coded a total of N = 185 issues (= 2 celebrity suicides × 3 newspapers issues × 31 days; minus 1 issue that did not appear on one day). In a next step, we summed up the values that were coded for each individual newspaper issue to get a score across all three papers for data analysis. Given that we used representative newspapers for the liberal press, the local papers, and the (developing) mass press, these measures were interpreted as (admittedly rough) indicators for news coverage back then. The following variables were assessed: Prominence of person. We measured the number of pages that reported on the celebrity within the three newspapers for each day of the observation period -regardless of the context.

Prominence of death.
We identified the number of pages that reported on the death of the celebrity within the three newspapers for each day of the observation period -regardless of whether or not the death was labeled as a suicide.
Explicit prominence of suicide. To assess the salient (undue) prominence given to suicide, we followed a procedure used by Niederkrotenthaler and Sonneck (2007) and assessed how often the term "suicide" (German: Selbstmord) appeared within the three newspapers for each day of the observation period. Thus, we simply counted the number of times the word Selbstmord appeared. This measure assesses explicit prominence.
Implicit prominence of suicide. We also assessed how often suicide was mentioned in a subtle way that allowed readers to infer that it was a death by suicide. We assessed the number of times suicide was paraphrased, which was very broadly defined as any mention of suicide without using the actual word "suicide" (e.g., "hat sich selbst gerichtet" -executed himself; "des Lebens überdrüssig" -weary of life; "er hat es selbst gethan" -he did it himself; "Der Wahnsinn ließ ihn zur todbringenden Waffe greifen." -Madness made him pick the death-bringing weapon.).

Suicide method.
We measured a total of six indicator variables related to the depiction of the suicide method (i.e., method reported on front page, method details reported on front page, graphical depiction of suicide method on front page, andrelated to the remaining news coverage excluding the front page -method reported, method details reported, and graphical depiction of suicide method). The six variables were coded dichotomously (0 = not present, 1 = present) and summed up to produce an overall method score. Although this is a rather rough measure, a higher score indicates a more vivid (graphical) depiction, including details about the method being prominently reported. Again, this measure was calculated across the three newspapers for each day of the observation period. As an analysis in the context of the reliability test (see below) indicated, internal consistency among these six indicator variables was high (Cronbach's α = .806).

Test of intercoder reliability
We assessed whether the measurement was reliable. The coding of the main study was done by one of the authors (MM). For the reliability test, a second person, not part of the research team, coded a subsample of N = 36 issues (= 19.5% of all coded issues) of news coverage. The analysis indicated a reliable measurement (all Krippendorff's alphas > 0.827). Details can be found in the Electronic Supplementary Material File.

Results
As can be seen in Figure 1, the news heavily reported about Rudolf and Redl (prominence of person and death). Conversely, the analysis indicated a huge difference regarding the prominence of suicide: Whereas the press heavily reported about suicide in the Redl case by explicitly mentioning suicide numerous times, the press rarely explicitly mentioned suicide in the Rudolf case (explicit prominence of suicide). Nevertheless, this does not mean that readers could not be aware of the fact that Rudolf's death was a suicide, given that the press provided other, more subtle paraphrases of suicide (implicit prominence of suicide). These subtle descriptions were used in both cases, but again, more frequently in the Redl case. Furthermore, the press heavily reported about the suicide method in the Redl case and provided less details on the suicide method in the Rudolf case.

Discussion
The content analysis of Redl's suicide provides evidence for an irresponsible style of reporting on suicide in terms of a higher amount, a longer duration, and a higher prominence of suicide, including more detailed descriptions of the method. Based on Werther effect research, "media and suicide" scholars would presumably hypothesize that Redl's death might have elicited an increase in the number of suicides in the immediate aftermath of his suicide. In fact, a pattern consistent with this hypothesis is exactly what a previous study has already found (Arendt & Mestas, 2021). Thus, there was an increase in the number of suicides after the heavy, irresponsible news coverage about Colonel Redl's death.
Conversely, although the content analysis for Crown Prince Rudolf equally shows a high amount, a long duration, and a high prominence of news coverage related to the person and his death, suicide per se was not a substantial (explicit) part of the story and details on the method were rare. In other words, factors that had been identified to be related to Werther effects (i.e., undue prominence, detailed depictions of the method) by previous Werther effect research (see above) could be found to a substantially lesser extent when compared to the Redl case. According to our interpretation of the findings, it seems to be plausible that some form of anticipated censorship, obedience or decorum account for the lack of detailed and explicit (sensational) suicide reporting.
This evidence led us to carefully hypothesize that Rudolf's suicide might not have elicited a Werther effect similar in size as already documented for the Redl case. Accordingly, we did not expect a (substantial) Werther effect-consistent pattern of increasing numbers of suicides in the aftermath of Rudolf's suicide.

Study 2: Werther effect study
We conducted an observational study and used civil death registers to identify suicides before and after Rudolf's suicide. We tested whether Rudolf's suicide elicited an increase in the number of suicides. We followed the cited previous study's logic of analysis (Arendt & Mestas, 2021) to increase the comparability of results between the Redl and Rudolf cases.

Method
Due to the fact that daily suicide statistics for the City of Vienna were not available for that time period, we conducted archival research using the handwritten public civil death records of Vienna to obtain daily suicide numbers. We identified all suicides that explicitly mentioned Selbstmord (German for "suicide") as the official cause of death. We defined a total of six periods, each with a length of four months. Although the length of four months is longer than the period in the previous study on Colonel Redl (Arendt & Mestas, 2021), it was necessary to increase the period under study given that the absolute number of (documented) suicides in the civil death register was much lower compared to the Redl periods. This corresponds with the substantial difference in general suicide numbers between both periods (see Ortmayr, 1990).
One of the periods was the target period (termed Post-Rudolf 1889), representing the aftermath of Rudolfs's suicide (i.e., four months after January 30, 1889). Consistent with the target period, we used the 30th of each month for the definition of five control periods. We used control periods of the same year of Rudolf's suicide (i.e., four months before, termed Pre-Rudolf 1889), control periods of the preceding year (i.e., the same months before and after January 30 in 1888, termed Pre-Preceding Year and Post-Preceding Year), and of the subsequent year (i.e., the same months before and after January 30 in 1890, termed Pre-Subsequent Year and Post-Subsequent Year).
Consistent with the previous study (Arendt & Mestas, 2021), we took a person/suicide-centered perspective and relied on a survival analysis to test for a Werther effect. We followed this previous study's data-analytic approach to increase the comparability of the results between Redl and Rudolf. In fact, we used the Kaplan-Meier method (Kaplan & Meier, 1958). We categorized the identified suicides into the six periods and calculated a one-step function for each period. The log-rank test was used to compare the total survival experience for the six periods and to test whether they were significantly different.
In a first step, we identified the number of suicides for all six periods. As we report in more detail below, Pre-Rudolf 1889 had the lowest number of suicides (N = 24). In a second step, we identified the first 24 suicides that occurred in each of the remaining five periods, beginning on the 30th of each period's first month. This ensures that survival is zero in all periods after the same number of deaths had occurred, allowing for a straightforward assessment of the Werther effect. Stated differently, the question was how long it took for 24 individuals to die by suicide within each period. A Werther effect would be indicated by a significant "faster" survival experience in the target period compared to the control periods.

Results
Using the log-rank test (Mantel-Cox), we found a significant difference among the survival times for the six observation periods, χ(5) = 27.22, p < .001. Figure 2 indicates that this effect can largely be attributed to the Pre-Rudolf 1889 and Pre-Subsequent Year periods, in which deaths occurred more slowly. When comparing the Pre-Rudolf 1889 with the Post-Rudolf 1889 period, a significant difference emerged, χ(1) = 6.70, p = .010. Although this pattern, when viewed in isolation, would be consistent with the Werther effect, individuals generally died slower in the post-periods, consistent with a wellknown seasonal pattern (i.e., more suicides occurred in spring, which corresponds to the post-periods in the present study; see Kuttelwascher, 1912, p. 334). Indeed, similar patterns related to the difference between the pre-and post-periods could be revealed for the preceding year, χ(1) = 2.24, p = .135, and for the subsequent year, χ(1) = 11.32, p = .001. Most importantly, when comparing the three post-periods (i.e., 1888, 1889, 1890) and thus holding the seasonal variations constant, the Post-Rudolf 1889 period was not significantly different from the Post-Preceding Year, χ(1) = 2.67, p = .102, or the Post-Subsequent Year, χ(1) = 1.10, p = .294. These findings do not support the Werther effect. As already noted above, a Werther effect would be indicated by a significantly "faster" survival experience in the Post-Rudolf 1889 period compared to the control periods. This was not the case.

Discussion
Taken together, when looking at all six periods, the analysis does not provide convincing supporting evidence for a Werther effect. Although there were significant differences in the survival experience among the six periods, the period in the aftermath of Rudolf's suicide did not show a Werther effect-consistent pattern. Of interest, the Pre-Preceding Year period showed a comparably faster declining step function (see Figure 2) compared to the Pre-Rudolf 1889 and Pre-Subsequent Year periods, as one would expect based on seasonal effects: As already noted, it comes as no surprise that the two latter pre-periods showed slower decreasing step functions (i.e., less deaths by suicide within a given time interval). However, we could not explain why the Pre-Preceding Year did not show the same pattern. Although this could be down to pure chance, a systematic contributing factor is likely. Given that the present study did not focus on this time period and that the Pre-Preceding Year is a less important control period for the target period, we refrain from providing post hoc interpretations of possible causes.

General discussion
Despite the enormous amount of press coverage of Rudolf's death, reporting did not give undue prominence to his suicide and details on the method were rare. Importantly, we did not find evidence of a subsequent increase in suicides in the aftermath of Rudolf's death. Conversely, in the Redl case, the press heavily reported in an irresponsible manner, as the content analysis of the present study revealed. Importantly, previous research has already documented a Werther effect-consistent pattern; that is, an increasing number of suicides following Redl's suicide (Arendt & Mestas, 2021).
The findings of the present study related to the Rudolf case emphasize the importance of current media guidelines on responsible reporting. In fact, a high amount of news coverage does not necessarily translate into a Werther effect. Importantly, the aim of today's guidelines on RRS is not to discourage journalists' reporting of suicides in general nor to suppress the media; instead, the aim is not to give undue prominence to suicide nor report on vivid details about the method employed. The present study thus adds to previous research investigating celebrity suicide cases that were reported on but were not related to a Werther effect-consistent pattern of increasing suicide numbers, such as the death of Kurt Cobain in the Seattle area (Jobes et al., 1996). The latter case is, in our view, roughly comparable to the Rudolf case. Conversely, the Redl case in which the press heavily reported in an irresponsible way may be roughly compared to many of today's celebrity suicides, such as the death of Robin Williams (see above) for which a Werther effect-consistent pattern had also been revealed (Fink et al., 2018). Future research should more often study (celebrity) suicides for which no Werther effect-consistent pattern could be observed. This is a supplementary perspective to the standard methodological paradigm that focuses on the existence of effects. It is our hope that we can also learn from the absence of an effect, thus contributing to our knowledge about the role of the media in the suicide (prevention) domain.

Limitations
The study has a number of limitations. First, this study relied on handwritten death records of the individual districts of the City of Vienna. These data come with a set of limitations. Scholars have pointed out that suicides were massively underreported in the 19th century per se, based in part on the views of the Catholic Church on suicides and the subsequent pressure from family members of the deceased to ensure a proper burial (Kuttelwascher, 1912). Although this underreporting presumably occurred in all six periods to the same extent, underreporting reduces the number of documented cases. For that reason, we had to use periods that were longer than usual (i.e., four months). Second, this study focused on the City of Vienna, the capital of the Austro-Hungarian Empire. It would be interesting to find out if there was an observable Werther effect in the rest of the Empire and even beyond country lines. The German Empire, for example, did not censor the specific details on Rudolf's death. This aspect would be especially interesting in comparison with data from Cobain's suicide that we mentioned above (i.e., the existence and absence of a Werther effect-consistent pattern in the Seattle area versus France). Third, we only relied on (officially documented) suicides identified in civil death registers. Of note, there are anecdotal reports of attempted suicides at the chapel where Rudolf's embalmed body was laid out for the public to grieve (Hamann, 1988). However, official data regarding these suicide attempts are not available, thus a further investigation of this aspect was not possible. Fourth, we want to emphasize that our ideas related to "forced responsible reporting" represent a posthoc interpretation. Although there were differences between reporting on the Redl and Rudolf cases, we were not able to provide supporting empirical evidence that confirms this "forced responsible reporting" interpretation. Other differences between both cases may have contributed to the differences in reporting. For example, we were not able to contentanalyze newspapers pre-and post confiscation (i.e., subsequently revised news articles) due to the fact that there are no public records on confiscated newspaper issues. Nevertheless, this is a valuable starting point for a future study that will hopefully allow us to paint a more exact picture. Related to this point, we acknowledge that the time span between Crown Prince Rudolf's (1889) and Colonel Redl's (1913) death itself may account for some of the difference in reporting on these particular suicides as the fin de siècle constituted a time of rapid societal, economic, political and media change (Melischek & Seethalter, 2016). Nevertheless, both suicides occurred in the advent of World War I and amidst the inevitable fall of the Habsburg empire and share a similar amount of public attention which made the cases comparable. Furthermore, Colonel Redl's suicide is the only well documented celebrity suicide of that time in Werther effect research, providing empirical evidence.

Conclusion
Despite its limitations, the present study provides supporting empirical evidence that a high amount of news coverage does not necessarily translate into a Werther effect. The findings emphasize the importance of responsible suicide reporting and current media guidelines. Even though "forced responsible reporting" back then seemed to have had a positive effect in this particular case, it is crucial to note that we, in no way, shape, or form, condone censorship and/or anticipated sanctions. However, what the present study emphasizes is that we can conduct research on a different time and apply these findings in the present to paint a more nuanced picture of today. This will hopefully contribute to knowledge generation and thus increase our understanding of the role of the media in the suicide (prevention) domain.