Article Text

Download PDFPDF

The impact of media reporting of suicide on actual suicides in Taiwan, 2002–05
  1. Ying-Yeh Chen1,2,
  2. Feng Chen3,
  3. Paul S F Yip4,5
  1. 1Taipei City Psychiatric Centre, Taipei City Hospital, Taipei, Taiwan
  2. 2Institute of Public Health and Department of Public Health, National Yang-Ming University, Taipei, Taiwan
  3. 3Department of Statistics, University of New South Wales, New South Wales, Australia
  4. 4Hong Kong Jockey Club Center for Suicide Research and Prevention, University of Hong Kong, Hong Kong
  5. 5Department of Social Work and Social Administration, University of Hong Kong, Hong Kong
  1. Correspondence to Dr Ying-Yeh Chen, 309 Songde Road, Taipei City Psychiatric Center, Taipei City Hospital, Taipei 110, Taiwan; dai73{at}


Objectives To assess changes in the intensity of suicide news reporting in Taiwan's local newspapers after the arrival of a daily tabloid-type newspaper, Apple Daily (AD), and evaluate the impact of suicide news reporting on actual suicides and possible mutual causation.

Methods A counting process was used to estimate the intensity of daily suicide news items reported in the China Times (CT) and United Daily (UD) before and after the arrival of AD (2002–05). Poisson regression models were used to assess the impact of the intensity of suicide news reporting on the actual number of next day suicides. Granger's causation model was used to assess mutual causation between suicide news reporting and actual suicides.

Results There was a significant increase in reporting intensity of suicide news in the UD soon after the entry of the AD into Taiwan's media market, while a delayed increase of approximately 1 year was observed in the CT. After the arrival of the AD, the reporting intensity in the UD was significantly related to the occurrence of actual suicides (p<0.05), even after controlling for social variables, whereas no significant correlation was previously observed. Mutual causation between suicide news reporting and actual suicides was also observed.

Conclusions The presence of the AD in Taiwan has fuelled competitive reporting of suicide news among traditional newspapers. This increase in the intensity of suicide news reporting has consequently had an impact on the actual number of suicides. This provides further empirical support for improving media reporting as a key element in suicide prevention.

  • Media
  • mutual causation
  • reporting intensity
  • suicide
  • epidemiology FQ
  • suicide SI

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Taiwan has recently witnessed a dramatic surge in the suicide rate, from as low as 6.2 per 100 000 (persons) in 1993 to a historical high of 19.3 per 100 000 in 2006. The latest figure is 17.9 per 100 000 population, making it the ninth leading cause of death in 2008.1 In 2007, the potential years of life lost due to suicide in Taiwan has been estimated at US$ 1 billion in lost earnings.2 Suicide has become an urgent and important public health problem in Taiwan.

Widespread media reporting of suicide events has been implicated as a major contributing factor to the rise in the suicide rate in Taiwan.3–7 In particular, media glamorisation of charcoal burning poisoning suicide has played a crucial role in the sudden rash of suicides using this method.3 8 9 The impact of media dissemination of suicides in Taiwan is also evident from a series of recent studies that demonstrated how extensive news coverage of celebrity suicides markedly induced suicide imitations.4–7 In fact, the recent increase in the suicide rate in Taiwan coincides with media proliferation during the same period.10 11

The transformation of Taiwan's mass media market and reporting style is closely related to the arrival of a daily newspaper, the Apple Daily (AD), in May 2003. It is a Hong Kong-based tabloid that highlights celebrities, gossip and scandals.12 13 Grabbing wide readership via sensationalism, exaggerated headlines and flashy graphic images, it came from nowhere but successfully gained a share of the top spot in Taiwan soon after its entry.14 This ‘success’ of the AD has consequently changed the reporting styles of Taiwan's traditional newspapers, with writer-driven journalism giving way to media- and market-driven reporting styles. It has had serious adverse implications on the media industry in Taiwan.12

However, no prior studies have specifically and empirically explored the changes in the reporting intensity of suicide news after the market shock brought about by the arrival of the AD, and whether suicide news reporting influences actual suicides on the following day or vice versa (ie, mutual causation of suicide news and suicides).15 This study explores the impact of the AD on the intensity of suicide news reporting in two other traditional newspapers, the China Times (CT) and United Daily (UD), which are among the top five newspapers by circulation in Taiwan.11–13 Moreover, this study explores whether reporting suicide news increases suicide incidence on the following day, since the concept of mutual causation is essential to a more comprehensive understanding of the issue. Hence, the causal relationships between media influence on suicides and of actual suicides on news reporting were estimated. Most recent research on the media and suicide has generally focused on the analysis of one or a very limited number of suicide stories that concern entertainment celebrities.16 In contrast, we included all stories, both those concerning celebrities and non-celebrity stories in this study. Following the conclusions of Stack, we argue that suicides of ordinary people also exert societal impact, although the impact is apt to be stronger for celebrities.16 17


Reports of daily suicide news in the CT and UD between January 2002 and December 2005, and in the AD between May 2003 and December 2005, were analysed. Using online search tools, Wisenews (for AD), United Daily News Dataset (for UD), and Knowledge Media Winner (for CT), the daily number of articles with headlines containing Chinese keywords related to suicidal behaviours (eg, ‘suicide’, ‘building jumping’, ‘charcoal burning’ or ‘hanging’) were counted. Editorial commentaries, fictional stories, suicide terrorism and articles related to suicide psycho-education were excluded. Both attempted suicides and completed suicides were included in the analysis. All suicide news items were included regardless of the extent of coverage or the placement of the news article. The CT and UD were chosen because these are two of the longest running newspapers in Taiwan, with good public credibility.11 They were the leading newspapers (in first and second spots, respectively) before 2000.11

Data on the daily number of actual suicides were obtained from official death records in Taiwan. Suicide deaths are defined as mortality coded under ICD-9 E950–959. Since prior studies have shown that deaths classified as ‘undetermined intent’ are often suicide cases,18 we performed our analyses on suicide cases and suicide plus undetermined cases, respectively, and the results did not differ significantly. Hence in this paper we presented the results based on official/confirmed suicide cases only.

A counting process approach was used to estimate the reporting intensity for each of the three newspapers.19 N(t) was defined as the counting process representing the accumulated number of suicide reports in a newspaper up to time t, and Y(t) as the number of cases in the 1-day period prior to time t. Then the reporting intensity function at time t (ie, the expected number of reports per case per unit of time at the time instant t), was defined to be:α(t)=limΔt0+N(t+Δt)N(t)Y(t)Δt

Formulating the question as such, we could estimate the reporting intensity function using a counting process based method, such as the kernel smoothing method or the local polynomial method.20 Here we used the local polynomial method because it does not require the extra effort to correct for boundary effects suffered by the kernel smoothing method.20 More specifically, we used the local linear intensity estimator given by:α^(t)=e1TA(t)101Kb(st)g(st)J(s)Y(s)dN(s),where e1=(1,0)T,A(t)=01Kb(st)g(st)2ds,Kb(x)=1bK(xb), K(x) is a kernel function, such as the Epanechnikov kernel K(x)=0.75(1x2)I{1x1},g(x)=(1,x)T,and J(x)=I{Y(x)>0}. The tuning parameter b, called the bandwidth, was chosen based on data-driven methods as suggested by prior studies.20 21

To assess the potential influence of newspaper reporting on future occurrence of suicides with the possible confounding effects of time trend, seasonal factors such as day of the week and month of the year, and other exogenous variables such as unemployment and divorce rates, a regression model was fitted to include all of these as independent variables. Since daily suicide counts were counts of rare events, the Poisson regression model was used for analysis. The usual time series model of ARIMA was not used due to the low count of daily suicide incidence. The dependent variable Yt denoted the number of suicides on day t, and the independent variables XtA,XtU,XtC,Xtun,Xtdiv denoted the numbers of suicides reported in the AD, UD, CT, the unemployment rate and divorce rate, respectively. The model was thus as follows:{Yt|σ{XsA,XsC,XsU,s<t;Xtun,Xtdiv}Poisson(μ),log(μ)=β0+β1t+β2Xt1A+β3Xt1U+β4Xt1C+β5Xtun+β6Xtdiv+β7Yt1+i=16β7+iI{DWt=i}+i=212β12+iI{MYt=i},(1)where DW was the variable for day of the week, which equalled 1 if it was Monday and 2 if it was Tuesday, etc., and MY was month of the year, which equalled 2 if it was February, 3 if March, and so on. The parameters βi i=0, 1,…24, measured the contribution of the covariates on the mean of the suicide number on the log scale. They were estimated using the maximum (conditional) likelihood method.22 We also estimated the combined effect by adding an interaction term U×C×A to the model to test whether or not multiple coverage in the three newspapers increased suicide counts.

Similarly, mutual causation23 was also assessed by model (1), with the variable number of suicide reports in a newspaper XtN(N=A,C or U) acting as the dependent variable, having a conditional Poisson distribution with mean μ depending on the independent variables.


Basic statistics (mean, SD and ranges) of the study variables, including daily count of suicide news items and suicide numbers, are given in the appendix. Figure 1 shows daily counts of the number of suicides and news reports in the three newspapers, and figure 2 gives the estimated suicide rates and reporting intensities by the counting process method. The number of news reports increased from May 2003, which would be natural if there were more incidences of suicide leading to more reporting. However, the estimated reporting intensity (per suicide incidence) in the UD increased significantly after the arrival of the AD, whereas the reporting intensity in the CT remained unchanged in 2003, although it seemed to ‘catch up’ with the reporting intensity in the UD starting from 2004 (figure 2). On the other hand, the reporting intensity of suicide news in the AD was high at its inception, but subsequently became relatively low compared to the other two traditional newspapers.

Figure 1

Daily counts of actual suicide and suicide news items in three leading newspapers, United Daily (UD), China Times (CT) and Apple Daily (AD).

Figure 2

Estimated suicide rate (case per day) and reporting intensity (reporting per suicide) of three leading newspapers, United Daily (UD), China Times (CT) and Apple Daily (AD), by the counting process method.

A significant correlation between the two traditional newspapers, the CT and UD, in suicide news reporting was observed throughout the study period (r=0.34 and 0.36 before and after the arrival of the AD, respectively). The correlation between number of suicides and the amount of suicide news reporting differed markedly before and after the entry of the AD (table 1). Before the entry of the AD, neither UD nor CT suicide news reporting was associated significantly with suicide incidence (r=0.05 and 0.07, respectively). After the arrival of the AD, the suicide news reporting for both the UD and CT became significantly related to actual number of suicides (r=0.12 (p<0.0001) and r=0.19 (p<0.0001), respectively), suggesting that the entry of the new AD tabloid might have had an impact on the reporting intensity of traditional newspapers in Taiwan, whereas no significant correlation was found between suicide incidence and the number of suicide news reports in the AD.

Table 1

Correlation coefficients of suicide news reporting for China Times, United Daily and actual number of suicide counts before and after the arrival of Apple Daily, 2002–05

Poisson autoregressive models evaluating the impact of daily suicide news reporting on actual suicides on the following day are presented in table 2. After the entry of the AD, suicide news reporting in the AD significantly predicted suicide incidents the next day. However, after controlling for a series of potential confounders (overall increasing time trend, number of suicides on the previous day, days of the week, months of the year, unemployment rate and divorce rate), the AD suicide news reporting no longer significantly predicted next day suicides. The reportage of the UD, however, turned out to be a potent predictor of next day suicides after similar adjustments (table 2). Turning to the days prior to the AD, suicide news reporting for both traditional newspapers was not related to number of suicides on the following day. There was no correlation between suicide news reporting and suicides in the pre-AD period, and it is reasonable to presume that the entry of the AD changed not only the reporting intensity of suicide news but also the possible copycat relationship of suicide news reporting on actual suicides, particularly for the UD. The interaction between the three newspapers was not significant, indicating that the combined effect of multiple coverage in the different newspapers did not increase the number of suicides.

Table 2

The impact of suicide news reporting on actual suicides before and after the arrival of Apple Daily, 2002–05

The estimated model suggests that all things being equal, every single UD suicide report inflated the expected number of suicides on the following day by a factor exp(0.016)=1.02 on average. The effects of the other two papers did not seem to be statistically significant. For the period from May 2003 to December 2005, the total inflation effects due to the reporting of the UD varied in the range 1–1.17. If the UD did not report suicide news at all, the expected number would have been discounted by a factor in the range 1–0.85. Our estimation shows that the UD alone accounted for 0–15% of the suicide counts during the post AD period.

Given the adverse impact of mass media reporting on suicide incidence, the causation from the opposite direction—that is, the influence of suicide events on suicide news reporting, could not be neglected. In table 3, we show the results with the reporting of suicide news for each newspaper as dependent variables, and number of suicide events on the previous day and prior day reporting of suicide news as predictors to evaluate the mutual influence of actual suicides and media reporting. Suicide events on the previous day significantly affected next day suicide news reporting for all the newspapers examined. The reportage of suicide news in the AD significantly predicted the reportage of suicide news in the CT the next day (table 3). The complex interactions between suicide news reporting, actual suicides and reportage of different newspapers are presented in table 3 and figure 3. The occurrence of a suicide increased the reporting of suicide news for all newspapers studied. However, only the reporting of the UD significantly correlated to elevated suicide rates.

Table 3

Evaluating the impact of actual suicide and suicide news reporting of other newspapers on the reportage of suicide news of United Daily, China Times and Apple Daily, respectively, in the following day

Figure 3

Granger causations among number of suicide and suicide news reporting for the three leading newspapers.


Main findings

There was a prominent increase in reporting intensity of suicide news in the UD soon after the entry of the AD, while the increase in the CT was delayed for nearly a year (after 2004). The change in reporting intensity in the UD after the entry of the new competitor was shown to be significantly related to the increasing trend of the suicide rate in Taiwan by the Poisson regression model. When the AD joined the media market, it had an impact on the reporting intensity of suicide news in local traditional newspapers, and may have consequently fuelled the suicide epidemic in Taiwan. Even after considering possible mutual causation (ie, escalated number of suicides resulting in greater reporting intensity), suicide news reporting in the UD remained a significant predictor of suicide incidence.

Response of local newspapers to the entry of AD

In face of fierce competition, the two traditional newspapers, CT and UD, adopted different strategies in terms of reporting suicide news, at least initially. One prior study investigating newspaper diversity before and 1 year after the entry of the AD found that Taiwan's traditional newspapers reacted differently towards the new competitor.13 The study showed that the amount of political news and reporting of social events increased significantly in the UD after May 2003, whereas the CT responded by increasing the reporting of ‘soft news’, such as leisure and lifestyle-related topics.13 The increase in reporting of political news as well as news in relation to social events in the UD reflects the newsmakers' intention of competing with the AD by enhancing its content diversity (ie, increase reporting of social events) as well as by offering a higher level of specialisation (ie, increased political news).10 13

Also aiming to increase content diversity, the initial strategy adopted by the CT was quite different from that of the UD. It added previously underemphasized news elements—soft news—to their product, rather than focusing on extending the reporting of social events. It is therefore not surprising to observe a steep rise in reporting intensity of suicide news in the UD in response to the AD, whereas the reporting intensity of suicide news items remained unchanged in the CT in 2003. However, the CT made substantial changes to its editorial policy and direction and caught up with the UD in reporting suicide news after 2004. It is obvious from the Granger causation model that the AD played a significant role in affecting the reporting intensity of suicide news in the CT during the study period.

A decline in intensity of suicide coverage in the AD was found after 2004. According to a research report, the AD continued to adjust its reporting style to attract Taiwanese consumers.24 Compared to the content in the early stages of its entrance, which was heavily loaded on the reporting of social events, the AD also increased the coverage of political and area-specific life/leisure news, which would appeal to Taiwanese readers.24 In view of the competition for space in any newspaper, the AD's modification of reporting style might be consequently related to the decrease in reportage of suicide news.

Echoing ample evidence from prior research that media reporting of suicide events may induce imitative suicides, especially among vulnerable individuals,25–27 this study again demonstrates the powerful role of mass media in suicide incidence. The media influence, however, only became significant after 2003 (entry of the AD). Before that, reporting of suicide news was not substantially related to suicide incidence (ie, the correlation was small and not statistically significant). In other words, the seemingly knee-jerk response or unintentional marketing strategic move of traditional newspapers in facing their competitors turned out to be a potential threat to vulnerable members of society at risk of suicide, something the editors of the UD and CT may not be aware of.

Impact of news reporting on suicides

Although the amount of reporting of suicide news in both the CT and UD were potent predictors of actual suicides during the post-AD period, only the reporting of the UD remained significant after making adjustments to a series of socially related variables. This suggests that the UD was independently related to the increase in number of suicide cases after controlling for other covariates. This could relate to UD readers taking what is reported in the newspaper more seriously compared to CT and AD readers. In fact, a previous study that examined differences in characteristics of readership among different newspapers in Taiwan between 1992 and 2004,28 showed that despite certain levels of overlap for the readership of these two traditional newspapers, for example their readers were more likely to be middle-class elite with a nationalist political orientation (ie, pro-unification with mainland China), there were some marked differences.28 The middle-aged group (aged 40–59 years) was over-represented among UD readers compared to those of the CT and AD (46.2% for UD vs 38.1% for CT and 23.8% for AD). Since the majority of the increase in the suicide rate in Taiwan in the past decade has occurred in the middle-aged group,1 the readers' age can be a potential explanation for the greater impact of the UD on the suicide incidence. In addition, we did not analyse the content and the length of the news articles. The possibility that the UD tends to report longer or more sensational suicide stories cannot be ruled out.

Although accurate newspaper circulation data in Taiwan are not available for all newspapers, a survey report revealed that after the AD joined the market, the readership for the CT decreased slightly; conversely, however, the readership for the UD increased.29 This could be another possible reason for the greater impact of the UD on the increase in the suicide rate.

Although the entry of the AD changed the intensity of suicide news reporting of Taiwan's traditional newspapers, its suicide news reporting per se was not shown to be related to actual suicides. This finding might be counter-intuitive at first glance, as the AD tends to attract young readers (ie, 50% are younger than 30 years old), the age group that is thought to be most vulnerable to a media copycat effect.30 However, it should be noted that prior studies that found a stronger copycat effect among youth were mostly derived from studies of celebrity suicides. Research that evaluated a wide range of suicide stories (including celebrity and non-celebrity suicides) did not necessarily observe a greater modelling effect in youths; conversely, some of them even revealed a lower risk of suicide modelling in young age groups.16 31 Since our analyses included all types of suicide stories, the null findings for the young-reader oriented tabloid journalism may not be that surprising. Future studies should explore the elements that facilitate the ‘match’ or ‘mismatch’ between viewers and suicide stories in order to better understand the age–gender-specific copycat effect. There are several other possible explanations for the null findings of the AD. First, the AD is not very selective in their choice of news stories and they essentially pick anything (including suicide reports) which is proven to be appealing to the public, thereby increasing their circulation. They are not necessarily interested in suicide news itself, but will publish anything that is most appealing to the community each day. The selection is totally based on what it sells rather than anything else. Second, the tabloid nature of the product may attract readers who prefer ‘soft news’ and therefore, may not take suicide news seriously. Third, the age structure of AD readers is generally not the high-risk age group for suicide in Taiwan. Lastly, visual presentations, such as photographs, graphics, and enlarged news headers, are usually key elements of tabloid journalism and simply by counting the number of news items may not fully reflect the impact of the AD.

Mutual causation

The AD entered Taiwan at a time when the suicide rate was rising quickly. The estimation of media influence on suicide would not be complete without considering mutual causation. The analyses here support the hypothesis of mutual causation: actual suicides increased suicide news reporting the next day and conversely, the number of news reports in the UD affected suicide incidence. One prior study in Japan that estimated mutual causation between suicide news reporting and actual suicides, revealed a one-way causation from the direction of news reporting to suicide, with no effect found for the impact of actual suicides on news reporting.15 The reason that this study failed to discover the causation from actual suicides to news reporting may be due to the aggregation of the daily data into monthly data, so the timeliness of suicide news on a daily basis would be masked, disrupted and distorted. In other words, the newspapers involved in Ishii's analysis were issued on a daily basis, but the news items needed to be aggregated to a monthly basis to match the sampling frequency of the suicide data. However, newspapers obviously value the timeliness of their news stories, and the possible causal effect of suicides on suicide news would be masked based on monthly data. It is a conjecture that if Ishii (1991) had done the analysis based on daily data, then the causal effect of suicides on suicide news might be discovered. The reporting practice of media professionals in different countries may be another explanation. To publish or not to publish suicide news is essentially a decision made by the editors of the relevant newspapers. Sometimes, professional accountability and responsibility of the mass media has been compromised by market-driven considerations only, and the protection of vulnerable members of the community has consequently been undermined. Indeed, the results of this study are a wake-up call for the media professionals in Taiwan to exercise professional judgement and to at least adopt a very basic ‘no harm rule’ as outlined by the code of ethics of the Society of Professional Journalists in the USA, which states, ‘Seek truth and report it’, followed by, ‘Minimise harm’. Certainly, not all of the blame should be placed on the media, and the causes of suicide are likely to be multi-factorial. However, it is a good time to reexamine the current practices of media reporting on suicide news objectively, and minimize possible harm, especially for vulnerable members of the community who are not able to protect themselves.


Several limitations need to be considered in interpreting the current results. First, suicide news reporting was measured by counting news items. Specific content analyses were not performed and other variables that may be related to media influence, such as position (eg, front page or not) or layout (eg, photos, graphics, headers) of the news articles were not considered. As mentioned earlier, the impact of the AD may be underestimated, as the visual presentation of the news article was not taken into account. Second, there is no accurate newspaper circulation data available for all the newspapers in Taiwan; reported amounts of sales were based on telephone surveys of a limited sample of Taiwanese. It should be recognised that the estimated impact of news reporting did not consider newspaper sales. Third, the impact of news reporting on actual suicides may last for more than one day. As newspapers can be saved and re-read, using daily count data may underestimate the impact of the media. The electronic media, including TV stations, were not included in the analyses. Stories covered in all media channels might be expected to have the greatest chance of significantly affecting the suicide rate. Lastly, the control covariates, divorce and unemployment rates, were only available on a monthly basis, assuming daily rates for these variables approximated to the monthly rates.

What is already known on this subject

  • Media proliferation in recent years in Taiwan is closely related to the arrival of a daily tabloid-type newspaper—the Apple Daily.

  • Widespread media glamorisation of suicide events has been implicated as a major contributing factor to the rise in the suicide rate in Taiwan.

What this study adds

  • The presence of the Apple Daily has fuelled competitive reporting of suicide news among Taiwan's traditional newspapers.

  • The increased suicide news reporting among Taiwan's traditional newspapers has consequently increased the actual number of suicides.


Descriptive statistics of daily counts of suicide news items and actual numbers of suicides

Number of suicides
United Daily
China Times
Apple Daily


View Abstract


  • Funding This study was partially supported by a grant from Department of Health, Taipei City Government (Grant No: 99001-62-016), GRF grant of the Research Council of the Hong Kong SAR Government and UNSW SFRG/ECR grant.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.