The Foxconn suicides and their media prominence: Is the Werther Effect applicable in China?

Article (PDF Available)inBMC Public Health 11(1):841 · November 2011with55 Reads
DOI: 10.1186/1471-2458-11-841 · Source: PubMed
Abstract
Media reporting of suicide and its relationship with actual suicide has rarely been investigated in Mainland China. The "Foxconn suicides" is a description referring to a string of suicides/attempts during 2010, all of which were related to a giant electrical manufacturing company, Foxconn. This study aimed to examine the clustering and copycat effects of the Foxconn suicides, and to investigate temporal patterns in how they were reported by the media in Mainland China, Hong Kong (HK), and Taiwan (TW). Relevant articles were collected from representative newspapers published in three big cities in Mainland China (Beijing (BJ), Shenzhen (SZ), and Guangzhou (GZ)), HK, and TW, together with searching intensity data on the topic conducted using the Baidu search engine in Mainland China. The temporal clustering effects of the Foxconn suicides and their media prominence were assessed using the Kolmogorov-Smirnov test. The media reports of the Foxconn suicides' temporal patterns were explored using a nonparametric curve estimation method (that is, the local linear method). The potential mutual interactions between the Foxconn suicides and their media prominence were also examined, using logistic and Poisson regression methods. The results support a temporal clustering effect for the Foxconn suicides. The BJ-based newspapers' reporting and the occurrence of a Foxconn suicide/attempt are each found to be associated with an elevated chance of a further Foxconn suicide 3 days later. The occurrence of a Foxconn suicide also immediately influenced the intensity of both Baidu searching and newspaper reporting. Regional diversity in suicide reporting tempo-patterns within Mainland China, and similarities between HK and TW, are also demonstrated. The Foxconn suicides were temporally clustered. Their occurrences were influenced by the reporting of BJ-based newspapers, and contagion within the company itself. Further suicide research and prevention work in China should consider its special media environment.
COR R E S P OND E N C E Open Access
The Foxconn suicides and their media
prominence: is the Werther Effect applicable in
China?
Qijin Cheng
1*
, Feng Chen
2
and Paul SF Yip
1,3
Abstract
Background: Media reporting of suicide and its relationship with actual suicide has rarely been investigated in
Mainland China. The Foxconn suicides is a description referring to a string of suicides/attempts during 2010, all of
which were related to a giant electrical manufacturing company, Foxconn. This study aimed to examine the
clustering and copycat effects of the Foxconn suicides, and to investigate temporal patterns in how they were
reported by the media in Mainland China, Hon g Kong (HK), and Taiwan (TW).
Methods: Relevant articles were collected from representative newspapers published in three big cities in
Mainland China (Beijing (BJ), Shenzhen (SZ), and Guangzhou (GZ)), HK, and TW, together with searching intensity
data on the topic conducted using the Baidu search engine in Mainland China. The temporal clustering effects of
the Foxconn suicides and their media prominence were assessed using the Kolmogorov-Smirnov test. The media
reports of the Foxconn suicides temporal patterns were explored using a nonparametric curve estimation method
(that is, the local linear method). The potential mutual interactions between the Foxconn suicides and their media
prominence were also examined, using logistic and Poisson regression methods.
Results: Th e results support a temporal clustering effect for the Foxconn suicides. The BJ-based newspapers
reporting and the occurrence of a Foxconn suici de/attempt are each found to be associated with an elevated
chance of a further Foxconn suicide 3 days later. The occurrence of a Foxconn suicide also immediately influenced
the intensity of both Baidu searching and newspaper reporting. Regional diversity in suicide reporting tempo-
patterns within Mainland China, and similarities between HK and TW, are also demonstrated.
Conclusions: The Foxconn suicides were temporally clustered. Their occurrences were influenced by the reporting
of BJ-based newspapers, and contagion within the company itself. Further suicide research and prevention work in
China should consider its speci al media environment.
Background
The suicide rate in Mainland China has been estimated
to be as high as 23 per 100, 000 individuals and the lat-
est (2009) suicide rates in Hong Kong (HK) and Tai wan
(TW), according to accurate census reports, were 13.8
and 17.6 per 100, 000, respectively [1-3]. Suicides in
these three communities have accounted for more than
a quarter of all such events worldwide [4]. However, the
suicide problem in China was not recognized by the
general public until the Foxconn suicides. In this
paper, this label is used to refer specifically to 13 com-
pleted and 5 attempted suicides in 2010, all of which
were connected to a single company, Foxconn (see
detailed case profile in Figure 1). The description was
coined by the mass media and used thereafter to refer
to the 18 cases.
Foxconn is a Taiwanese-owned company but most of
its factorie s are in Mainland China. It manufactures and
assembles consumer electronic products for companies
like Apple, Nokia, Dell, and so on. The string of sui-
cides/attempts attracted mass media concern about the
suicide problem in modern China in all Chinese com-
munities, and was even discussed by some international
* Correspondence: chengqj@hku.hk
1
Department of Social Work and Social Administration, 13th Floor, KK Leung
Building, The University of Hong Kong, Pokfulam, Hong Kong SAR, China
Full list of author information is available at the end of the article
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
© 2011 Cheng et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the term s of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unre stricted use, distri bution, and reproduction in
any medium, provided the original work is properly cited.
media, such as the New York Times and the Wall Street
Journal [5,6].
During the development of the Foxconn suicides, two
assertions were widely reported by the Chinese media
without any empirical evidence being produced; 1) that
the Foxconn suicides formed a cluster [7,8], and 2) that
they were attributable to media reporting [8-10]. Given
that this may be the first time the Chinese public has
had the opportunity to recognize suicide as a public
health problem, it is necessary to examine these asser-
tions scientifically, to avoid people being influenced by
wrong or misleading ideas. Therefore, we conducted an
Figure 1 Time distribution of the Foxconn suicides in 2010. Each case is formatted as date: gender (M = male, F = female)/age/suicidal
behavior (S = completed suicide, A = suicide attempt)/suicide method (J = jumping, C = cutting)/location (SZ = Shenzhen, FS = Foshan, KS =
Kunshan, HC = Hsinchu).
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 2 of 11
empirical study on the Foxconn suicides with the aim of
examining the two propositions outlined above.
For the first assertion, given that the Foxconn suicides
took place in di fferent cit ies but were all relevant to the
same company, we hypothesized that:
H1. There was a temporal clustering effect among the
Foxconn suicides.
For the second assertion, it is known that the inap-
propriate reporting of suicides, especially among celebri-
ties, is related to consequent macro-level increases in
suicide rates. This is commonly ref erred to as the
Werther Effect [11,12]. However, previous studies in this
field have been done mostly in western countries or
developed Chinese communities such as HK and TW
[13,14]. The influ ence of the mass media on r eal sui-
cides has not yet bee n empirically examined in Main-
land China [15]. More importantly, previous studies
often exami ne the relationship between media reporting
and changes of suicide rates in the general population,
rather than in terms of individual cases [11,12].
According to previous studies, suicide clusters can be
divided into two groups: point clusters, which occur
within a small community or ins titution, and mass clus-
ters, which are found in the general population [16].
The Foxconn suicides were all related to one company
so, if they were a cluster at all, should be placed in the
former category. A point cluster can be caused by social
learning within a local network or by the tendency for
similar individuals to share similar risk factors and com-
mit suicide in association with one another. Howeve r,
the relationship between a point cluster and prominent
mass media reporting of suicides is still unproven
[16,17]. Therefore, we hypothesize that:
H2. The mass media focus on the Foxconn suicides
did not significantly influence their occurrences; and
H3. P revious Foxconn suicides significantly influenced
later occurrences.
As well as testing these three hypotheses, we al so used
the Foxconn suicides to investigate one further research
question. We compared the prominent patterns of
media reporting of the same cases in Mainland China,
HK, and TW . The Ma inland Chinese mass mediasstyle
of reporting suicide has rarely been examined. One
recent study demonstrated different styles between
Guangzhou (GZ: a city in Mainland China), HK, and
TW. This encouraged us to ex pand our investigation to
more regions in Mainland China.
In this study, we define mass media to include not
only traditional media (that is, newspapers) but also new
media, specifically online search engines. Internet pene-
tration is over 70% in HK and TW and over 55% in
relatively developed Mainland regions such as Beijing
(BJ) and Guangdong Province [18-20]. With the wide-
spread use of the Internet, more and more media
researchers have turned to its use by the public for
informat ion seeking as a measure o f a news topics pro-
minence [21]. However, as far as we know, no study has
yet examined the influence of such online searching on
suicide occurrences.
We hope our study can narrow the huge margin on
media and suicide research in Mainland China and also
exp lore the prominence of the relationship bet ween the
media (including both traditional media and the Inter-
net) and individual suicide cases. In addition, previous
studies of m edia and suicid e have often been concerned
with single rather than multiple cases, and with the
deaths of celebrities rather than ordinary people [11,22].
Our study on the Foxconn suicides, covering as it does
the suicides or attempts of 18 ordinary people, will
enrich the literature on media and suicide.
Methods
Data collection
The 18 suicide cases, and detailed, relevant information
about them, were identified by reviewing media reports.
Both the company itself, and th e Chinese government,
held general med ia reporting of the Foxconn suicides
responsible for further occurrences [9,10]. Theref ore, we
tried to include a wide range of Chinese newspapers in
our examination. Not only Mainland Chinese newspa-
pers, but also those published in HK and TW were
included to examine t he potential influence of media
reporting. Within Mainland China, not only those news-
paper s distributed in Shenzhen (where most of the Fox-
conn suicides oc curred) were in cluded, but also those
availableinthelargerChinesecitiesandwhosecontent
was accessible online. Nati onal daily new spapers in
Mainland China were not included because they might
not be comparable with their HK and TW compatriots
due to the huge differences in population and
readership.
Specifically, three big cities with mature and active
media markets in Mainland China, namely Shenzhen
(SZ), GZ, and BJ, were selected to compare with HK
and TW. SZ, a special economic zone in Guangdong
Province which borders on HK, is the city in which
most of the Foxconn suicides occurred. GZ is the capital
city of Guangdong Province and is about an hour by
train from SZ. As a provincial capital city, newspapers
in GZ cover stories from around the province and have
correspondents in major Guangdong cities, including
SZ. BJ is the capital and the political, cultural, and
media center of Mainland China. Its local daily newspa-
pers, although they mainly cover local news, are also
concerned with national events.
Following previous studies on HK and TW media,
four popular daily newspapers with si gnificant circula-
tion figures and a wide range of reader profiles in e ach
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 3 of 11
region were included in this study (see Table 1 for
details) [23,24]. All 18 suicides/attempts were com-
mitted by jumping from a height, other than one
attempt where the individual concerned cut his wrists.
Accordingly, Foxconn and suicide or jump or fall
or cut in Chinese were entered into the WiseNews
database to search on the titles and content of all 20
newspapers included in the sample. The time period for
the search was set as the full year 2010, from Jan 1 to
Dec 31. A total of 1282 articles returned from the
search, including news articles and commentaries, were
then collected. After excluding articles that were not
actually relevant to the Foxconn suicides, 1279 remained
for content analysis and their full text versions were
downloaded. Some ba sic characteristics, including the
newspapers name, circulation region, and the articles
publication date were collected directly from WiseNews.
To measure the prominence of the Foxconn suicides in
each newspaper, the word count and placement (that is,
whether or not it appeared on the front page) were also
collected for each article.
Baidu searching trends relative to the Foxconn sui-
cides were also collected, as a measure of online user-
generated prominence. While western studies often cite
Google Trend data as an indicator of aggregated online
attention, Baidu, a Chinese search engine company, has
captured over 70% of the Mainland market [25]. It
allows users to enquire how many searches have been
done on a certain term or phrase on Baidu over a set
period (URL: http://index.baidu.com), a function similar
to Google Trend. Following Weeks and Southwell, we
tried several different search queries relevant to the Fox-
conn suicides and noticed that Fu Shi Kang Tiao Lou
("Foxconn jump in Chinese) generated the highest
overall search index rating compar ed to the other tested
terms. Accordingly, this term was ultimately chosen for
the data collection [21].
Data analysis
The dates of the 18 sui cides/attempts are shown in Fig-
ure 1, which immediately suggests a clustering phenom-
enon. To assess the statistical strength of the evidence,
we assumed the Foxconn suicides to have occurred
according to a Poisson process and hence tested the
constancy o f its intensity or its equivalent, the unifor-
mity o f the distribut ion of the event times of t he Pois-
son process over the observation interval. Specifically,
the Kolmogorov-Smirnov test was employed to test the
null hypothesis that the time distribution of the Fox-
conn suicides over the year of 2010 was uniform against
the two-sided al ternative hypothesis that it was not. The
null hypothesis is equivalent to a lack of a temporal
clustering effect for the Foxconn suicide s, and the alter-
native hypothesis to the presence of such an effect. If
the null hypothesis falls to be reject ed, we would then
use the local linear method to estimate the time-varying
Table 1 Descriptions of the 20 selected newspapers and the clustering effects of their reporting intensities of the
Foxconn suicides.
ID Circulation
region
Newspaper Total amount of articles P-value of the test of clustering effect
P1 TW Apple Daily 89 0.00
P2 TW Liberty Times 71 0.00
P3 TW United Daily News 125 0.00
P4 TW China Times 83 0.00
P5 HK Apple Daily 109 0.00
P6 HK Ming Pao 138 0.00
P7 HK Sing Tao 128 0.00
P8 HK Oriental Daily News 68 0.00
P9 GD Southern Metropolis Daily 92 0.00
P10 GD Nanfang Daily 40 0.00
P11 GZ Yangcheng Evening News 50 0.00
P12 GZ Guangzhou Daily 34 0.00
P13 SZ Daily Sunshine 51 0.00
P14 SZ Shenzhen Economics Daily 21 0.02
P15 SZ Shenzhen Special Zone Daily 34 0.00
P16 SZ Shenzhen Evening News 34 0.00
P17 BJ Beijing News 42 0.00
P18 BJ Beijing Youth Daily 37 0.00
P19 BJ Beijing Evening News 19 0.22
P20 BJ Fa Zhi Evening News 14 0.02
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 4 of 11
intensity funct ion, which would help interpret t he clus-
tering [26,27]. The potential clustering effects among
the newspaper reporting and Ba idu searching of the
Foxconn suicides were also assessed using the Kolmo-
gorov-Smirnov te st, and the corresponding time-varying
intensities estimated using the lo cal linear method. At
this point, it is important to clarify that the intensities
of specific types of events (such as the Foxconn suicides,
the newspaper reporting of them, or the resulting online
searching) are the intensity functions of the r espective
Poisson proce sses us ed to model these specific types. So
if N(t) denotes the cu mulativ e number of a specific type
of events from time 0 (the beginning of the year 2010)
up to time t,thenN( t)isassumedtobeaPoissonpro-
cess, and the inte nsity of the event at time t is defined
as:
λ(t) = lim
t
0
E{N( t + t) N(t)}
t
.
(1)
To explore the potential mutual influences between
the media prominence and actual subsequent occur-
rences of the Foxconn suicides/attempts, we performed
the logistic and Poisson regression analyses described
below. Since the potential influence of media coverage
on future suicides may not be immediate, but equally
may not last very long [28], t he time-lagged variables
were set as regressors. As suggested by the time lag
plots (Additional File 1), a lag range from 1 to 5 days
was adopted. In Figure 2, the trend lines were obtained
by the LOWESS scatterplot smoother [29].
To describe the models in more specific terms, let x
t
=
(x
1t
, x
2t
, ..., x
5t
), t = 1, ..., 365, denote the series of vec-
tor-valued counts of interest, where x
1t
, x
2t
,andx
3t
denote the numbers of Foxconn suicide stories pub-
lished by TW and HK papers, GZ and SZ papers, and
BJ papers respectively; x
4t
denotes the amount of Baidu
searching on the keyword; and x
5t
denotes the number
of Foxconn suicides/attempts on day t.Thevariablex
5t
has values 0-1, and we modeled it using a Bernoulli dis-
tribution with probability of success P
t
which, after a
logit transformation, depended linearly on the values of
the series on the previous q days. More specifically, the
model assumed that
x
5t
Bernoulli(p
t
);
logit(p
t
)=log
p
t
1 p
t
= β
0
+
5
i=1
q
j
=1
x
i,tj
β
ji
,
(2)
where b
0
and b
ji
, j = 1, ..., q, i = 1, ..., 5, were model
parameters. Since the variables x
1t
, ..., x
4t
,werelow
counts with zeros, a natural model to consider, when
assessing the influence of other variables on a specific
variable, was the Poisson regression model. Rather than
using the ordinary case, a quasi-Poisson regression wa s
adopted to account for the possible over-dispersion phe-
nomenon and hence reduce the chance of detecting fal-
sely-significant variables. More specifically, the mean of
the response variable μ
it
=E[x
it
], i =1,...,4,afterthe
log transformation, was assumed to be a linear function
of the values of the time series on the past q days, and
the variance of the response was assumed to be equal to
tha t expected for a Poisson distribution up to a positi ve
scale parameter j,knownasthedispersion parameter.
That is,
log(μ
kt
)=β
0
+
5
i=1
q
j=1
x
i,tj
β
ji
,
v
kt
=var
(
x
kt
)
= ϕμ
kt
, k =1,...,4
,
(3)
where the par ameters b
0
, b
ji
,j=1, ..., q, i = 1, ..., 5,
and j depend on which variable acts as the response
variable.
The logistic (2) and log-linear (3) models were all
fitted using the General Linear Model (GLM) function
of the statistical package R [30]. When fitting these
regression models, we only used data from day 14
onward. This was because there were no suicides/
attempts during the first 14 days of 2010. T here were
Figure 2 Estima ted daily suicide intensity at Foxconn in 2010
and its pointwise 95% confidence intervals.
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 5 of 11
two newspa per articles over this period about one sui-
cideataFoxconnfactoryinJuly2009.Sincethatcase
was otherwise not included in the Foxconn suicides by
the mass media, it was also excluded from this study.
Results
Clustering effect
The P value of the Kolmogorov-Smirnov test of unifor-
mity of the Foxconn suicides time distribution was
0.002 < 0.05, so the null hypothesis (no clustering of
suicide times) can b e firmly rejected in favor of the
alternative hypothesis that there was a temporal cluster-
ing of the Foxconn suicides. Figure 2 pres ents the local
linear estimates of the suicide intensity together with
point wise 95% confidence intervals using the Epanech-
nikov kernel and the rule of thumb bandwidth calcu-
lated from the data [27]. Figure 2 shows a clear peak in
the suicide intensity occurring roughly b etween day 70
to day 150, or mid-March to May 2010. This
corresponds well with the clustering of most of the Fox-
conn suicides in this period. Therefore, H1 is supported.
Prominence tempo-pattern
Figure 3 shows the estimated reporting inte nsities of the
20 selected newspapers a nd the estimated searching
intensity on Baidu. These tend to suggest that while dif-
ferent newspapers or media types had clearly different
scale of reporting/search intensities, their tempo-pa t-
terns were more or less consistent, with all showing
clear peaks around clusters of suicid es or isolated cases.
A closer inspection shows that the peaks of the newspa-
pers reporting intens ities in the same region were very
well aligned, but those of newspapers from different
regions tended to differ. This phenomeno n can also be
seen from Figure 4, which shows the daily amounts of
news reporting and online searching running in parallel.
According to their reporting patterns, the 20 newspa-
pers can be divided into three categori es: HK- & TW-
0 100 200 300
01234
P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
P13
P14
P15
P16
P17
P18
P19
P20
baidu
Da
y
s in 2010
Reporting/searching intensities
Figure 3 Estimated newspaper reporting intensities and online searching intensity for the Foxconn suicides in 2010. The descriptions of
P1-20 can be found in Table 1. The Baidu searching intensity curves on a log transformed scale.
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 6 of 11
based, GZ- & SZ-based, and BJ-based. The time period
can also be divided into three phases. F rom January to
mid-May 2010, the GZ- & SZ-based papers reporting
intensities reflected the time distribution of the Foxconn
suicides, whereas those of other newspapers under-
reflected it. From mid-May to early June 2010, all the
newspapers reporting intensities increased strongly,
reflecting the peak of the Foxconn suicides intensity.
Figure 4 Daily amount of newspaper reporting and online searching on the Foxconn suicides in 2010. The descriptions of P1-20 can be
found in Table 1. The broken grey lines indicate occurrences of the Foxconn suicides.
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 7 of 11
After that, the HK- & TW-based papers reporting
intensities continued to mirror the pattern of the Fox-
conn suicid es, whereas those of the Mainland Chinese
newspapers all dropped to a much lower level. D uring
the whole of 2010, the BJ-based papers showed a very
low reporting intensity for the Foxconn suicides. No sig-
nificant searching for the topic on Baidu took plac e
until 29 March 2010, the date of the fourth Foxconn
suicide. However, from 30 March 2010 on wards, the
temporal pattern of Baidu searching intensity was fairly
consistent with that of the reporting intensity of the
newspapers, especially those in TW and HK.
Interactions between the Foxconn suicides and their
media prominence
The results of the regression analysis are shown in Table
2. Of all the 20 newspapers and one search engine stu-
died, only the co ver age in the BJ-based newspapers sig-
nificantly contributed to the occurrences of Foxconn
suicides 3 days later. A Foxconn suicide/attempt on a
certain day also significantly contributed to a subsequent
occurrence 3 days later. Therefore, H3, but not H2, is
supported.
In addition, occurrences of Foxconn suicides had an
immediate influence on the amount of Baidu searching
at lag 1 and a continuous influence until lag 4. They
had a relatively slower influence on the amount of GZ,
SZ, HK, & TW newspapers reporting at lag 2 and 3,
and the slowest influence on all the BJ-based newspa-
pers at lag 4.
Discussion
The study echoes previous work showing that a point
cluster can be caus ed by a copycat eff ect within a small
community or institution. T he influence of previous
Fox conn suicides on later such events can be attributed
to interpersonal communication within the company.
Foxconn provides workers with a huge campus contain-
ing dorms, canteens, and other facilities, which results
in a close community [7]. In addition, mobile phone
penetration is over 64% in Mainland China and almost
100% in Guangdong Provin ce [31], whic h can lead to
the fast spread of information within a group.
It is not surprising that most of the mass media pro-
minence had little influence on the occurrences of Fox-
conn suicides. The Foxconn suicides were all committed
by ordinary people, and the evidence s uggests that
media reporting of such suicides rarely leads to a copy-
cat effect [22,32]. Strikingly, the coverage of the Fox-
conn suicides in the BJ-based newspapers did have a
sig nificant influence on future occurrences. More inter-
estingly, the BJ-based newspapers actually showed the
lowest interest in, and the slowest follow up of, the Fox-
conn suicides among all the media studied here. This
Table 2 Interactions between the occurrences of the
Foxconn suicides and their media prominence.
Independent variables Estimate Std.
Error
Est./S.
E.
P
value
Sig.
The Foxconn suicides occurrences as dependent variable
(Intercept) -3.11 0.37 -8.36 0.00 Y-
TWHK newspapers.lag1 -0.32 0.24 -1.32 0.19
TWHK newspapers.lag2 0.10 0.20 0.47 0.64
TWHK newspapers.lag3 -0.42 0.24 -1.74 0.08
TWHK newspapers.lag4 -0.12 0.23 -0.53 0.60
TWHK newspapers.lag5 -0.26 0.22 -1.18 0.24
GZSZ newspapers.lag1 -0.27 0.26 -1.01 0.31
GZSZ newspapers.lag2 0.30 0.20 1.47 0.14
GZSZ newspapers.lag3 -0.06 0.20 -0.29 0.77
GZSZ newspapers.lag4 0.01 0.20 0.03 0.98
GZSZ newspapers.lag5 -0.01 0.18 -0.04 0.97
BJ newspapers.lag1 -0.64 0.71 -0.90 0.37
BJ newspapers.lag2 0.08 0.49 0.16 0.88
BJ newspapers.lag3 1.13 0.53 2.14 0.03 Y+
BJ newspapers.lag4 0.41 0.66 0.63 0.53
BJ newspapers.lag5 -0.57 0.70 -0.82 0.42
Baidu.lag1 0.00 0.00 1.29 0.20
Baidu.lag2 0.00 0.00 0.57 0.57
Baidu.lag3 0.00 0.00 0.67 0.50
Baidu.lag4 0.00 0.00 -0.11 0.91
Baidu.lag5 0.00 0.00 0.64 0.52
Foxconn Suicides
occurrences.lag1
1.26 1.02 1.23 0.22
Foxconn Suicides
occurrences.lag2
-4.26 3.42 -1.25 0.21
Foxconn Suicides
occurrences.lag3
2.31 1.10 2.11 0.04 Y+
Foxconn Suicides
occurrences.lag4
-2.13 2.36 -0.91 0.37
Foxconn Suicides
occurrences.lag5
-0.14 1.60 -0.08 0.93
Hong Kong and Taiwan newspapers daily reporting amount as dependent
variable
Foxconn Suicides
occurrences.lag1
-1.38 0.58 -2.39 0.02 Y-
Foxconn Suicides
occurrences.lag2
1.27 0.29 4.32 0.00 Y+
Foxconn Suicides
occurrences.lag3
1.37 0.31 4.48 0.00 Y+
Foxconn Suicides
occurrences.lag4
0.76 0.44 1.74 0.08
Foxconn Suicides
occurrences.lag5
0.29 0.42 0.69 0.49
Guangzhou and Shenzhen newspapers daily reporting amount as
dependent variable
Foxconn Suicides
occurrences.lag1
-0.13 0.51 -0.25 0.80
Foxconn Suicides
occurrences.lag2
1.93 0.29 6.54 0.00 Y+
Foxconn Suicides
occurrences.lag3
1.32 0.35 3.73 0.00 Y+
Foxconn Suicides
occurrences.lag4
0.43 0.54 0.79 0.43
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 8 of 11
somewhat unexpected phenomenon might be explained
by the special nature of BJ newspapers. Unlike t he HK
and TW media, Mainla nd Chinese media are wide ly
considered as the mouthpiece of the Chinese govern-
men t. BJ newspapers, located as they are in the nation s
capital, are often considered by the public to be the
mouthpiece of both the local and central government.
BJ newspapers concernwiththeFoxconnsuicides
might have been perceived as an authoritative message
from the Chinese government, which may unfortunately
have encouraged some vulnerable individuals to commit
suicide to attract public attention. If such an explanation
is plausible, we should pay more attention to the special
media environment in Mainland China when conducting
future studies on media and suicide.
As part of the exploration carried out in this study,
theonlinemedia s role in spreading the news of t he
Foxconn suici des is also noteworthy. Although the Fox-
conn suicides were associated with a subsequent
increase in intensity for all the media coverage studied
here, the impact of a Foxconn suicide on Baidu search-
ing was the most immediate and longest las ting. This
demonstrates the speed and duration of the response
generated by online users attention to the phenomenon.
Public health implications
Of the two assertions reported by the media, the one
tha t the Foxconn suic ides formed a cluster is supported
by our empirical study. However, the other one that the
Foxconn suicides were attributable to media reporting is
only supported under ce rtain conditions. The finding
that the reporting by the BJ newspapers, but not the
newspapers in other regions, influenced further Foxconn
suicidesremindsustobemorecarefultobalancethe
competing concerns of press freedom and the minimiza-
tion of harm. It is therefore necessary to report our
findings to professionals and policymakers who are
working on, or interested i n, suici de prevention in
China, as well as the general public.
Suicide prevention efforts in China should be pursued,
including introducing media guidelines on reporting sui-
cides. How ever, the current guidelines are based largely
on western studies [33]. More China-specific studies on
media and suicide are needed to optimize these guide-
lines for the special media environment in China. In
addition, the diversity of m edia reporting patterns in
Mainland China is noteworthy. Those who seek to
cooperate with the media to promote suicide prevention
in China should pay attention to these features and
design a flexible working strategy. To promote and prac-
tice suicide prevention in the new media environment,
they should also consider contributing more suicide pre-
vention content to the Internet for online searchers.
Limitations
The study al so has some limit ations, which call for
further research. Additional studies seeking to explain
why the BJ-based newspapers influenced the Foxconn
suicides are required. In-depth content analysis and qua-
litative inter views with relev ant journalists and Foxconn
workers may be helpful in understanding this phenom-
enon. In addition, due t o the lack of daily suicide rate
data in Mainl and China, this st udy has only examined
the influence of media prominence on the 18 Foxconn
suicides. Nevertheless, it has explored an a lternative
approach to studying the media s effect on copycat sui-
cides in a country where census data on suicide is not
available.
Conclusions
This study, utilizing advanced statistical methods, sup-
ports the hypoth esis that the Foxconn suicides formed a
temporal cluster. Of the 20 newspapers and one online
search engine studied here, only the BJ-based newspa-
pers reporting of the Foxconn suicides significantly
influenced subsequent occurrences. At the same time,
previous Foxconn suicides also significantly contributed
to later ones, which suggest a social learning phenom-
enon at work within the company.
The dynamics of the medias influence on copycat sui-
cide is complex and its investigation is just beginning.
Our study is pioneering in its examination o f the
Table 2 Interaction s between the occurrences of the Fox-
conn suicides and their media prominence. (Continued)
Foxconn Suicides
occurrences.lag5
0.25 0.50 0.50 0.62
Beijing newspapers daily reporting amount as dependent variable
Foxconn Suicides
occurrences.lag1
0.32 0.74 0.43 0.67
Foxconn Suicides
occurrences.lag2
0.88 0.59 1.50 0.13
Foxconn Suicides
occurrences.lag3
0.00 0.61 -0.01 0.99
Foxconn Suicides
occurrences.lag4
1.45 0.65 2.25 0.02 Y+
Foxconn Suicides
occurrences.lag5
0.79 0.59 1.34 0.18
Baidu searchings daily searching amount as dependent variable
Foxconn Suicides
occurrences.lag1
0.93 0.20 4.60 0.00 Y+
Foxconn Suicides
occurrences.lag2
1.22 0.19 6.56 0.00 Y+
Foxconn Suicides
occurrences.lag3
0.78 0.23 3.42 0.00 Y+
Foxconn Suicides
occurrences.lag4
0.92 0.26 3.49 0.00 Y+
Foxconn Suicides
occurrences.lag5
0.47 0.24 1.96 0.05
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 9 of 11
Mainland Chinese medias tempo-patterns of reporting
on suicides and their relationship with a suicide cluster.
It highlights more questions and possibil ities for this
field of study. The Mainland China media environment
is special due to the Chinese governments censorship
and the mediasstrugglewiththisinthecontextofthe
emerging market economy. Previous findings from wes-
tern countries, and even other Chinese societies like HK
and TW, may not be directly applicable to Mainland
China. To work with the media to prevent suicide in
China, the relevant professionals and policymakers need
to pay attention to its special media environment and
develop flexible strategies.
Additional material
Additional file 1: Lag plots of the count time series with the
occurrences of the Foxconn suicides as target variables. The file
contains a figure showing lag plots of the count time series with the
occurrences of the Foxconn suicides as target variables. Figure legend:
Suic = the Foxconn suicides, TWHK = Taiwan and Hong Kong
newspapers daily reporting amount, GDSZ = Guangzhou and Shenzhen
newspapers daily reporting amount, BJ = Beijing newspapers dailiy
reporting amount, Baidu = Baidus daily searching amount.
Acknowledgements
The study was partly supported by the General Research Grant from the
Hong Kong Research Grant Council (recipient: PSFY) and the University of
New South Wales ECR grant (recipient: FC).
Author details
1
Department of Social Work and Social Administration, 13th Floor, KK Leung
Building, The University of Hong Kong, Pokfulam, Hong Kong SAR, China.
2
School of Mathematics and Statistics, University of New South Wales,
Sydney, Australia.
3
HKJC Centre for Suicide Research and Prevention, No. 2,
University Drive, The University of Hong Kong, Hong Kong SAR, China.
Authors contributions
The study was initiated by QC and PSFY. QC designed the study, then
collected and coded the data. FC helped with the data analysis and
processed the figures. PSFY supervised the study and acted as coordinator.
All three authors participated in the explanation and discussion of the
results. The submitted manuscript was drafted by QC and discussed and
agreed upon by the other two authors. All authors read and approved the
final manuscript.
Competing interests
The authors declare that they have no competing interests.
Received: 7 July 2011 Accepted: 2 November 2011
Published: 2 November 2011
References
1. Phillips MR, Li XY, Zhang YP: Suicide rates in China, 1995-99. Lancet 2002,
359(9309):835-840.
2. Suicide rates and deaths by age group. [http://csrp.hku.hk/WEB/eng/
statistics.asp].
3. An analysis of the 2009 cause-of-death statistics in Taiwan. [http://www.
doh.gov.tw/CHT2006/DM/DM2_p01.aspx?
class_no=25&level_no=1&doc_no=76013].
4. Yip PSF, Liu K, Hu J, Song X: Suicide rates in China during a decade of
rapid social changes. Social Psychiatry and Psychiatric Epidemiology 2005,
40(10):792-798.
5. Barboza D: String of suicides continues at electronics supplier in China.
New York Times New York edn; 2010, B10.
6. The Foxconn suicides. The Wall Street Journal 2010.
7. Liu Z, Yang J: The puzzling case of eight continious jumping suicides at
Foxconn (in Chinese). Southern Weekend Guangzhou; 2010, A1.
8. How can we break the curse of Foxconns jumping cluster? (in Chinese).
Guangzhou Daily Guangzhou; 2010, A18.
9. Reporting on the Foxconn suicides was restricted (in Chinese). Ming Pao
Hong Kong; 2010, A03.
10. We are not running blood and sweat factories, says Foxconn founder (in
Chinese). Oriental Daily News Hong Kong; 2010, A30.
11. Stack S: Media coverage as a risk factor in suicide. Journal of Epidemiology
and Community Health 2003, 57:238-240.
12. Blood RW, Pirkis J: Suicide and the media. Crisis: Journal of Crisis
Intervention & Suicide 2001, 22(4):163-169.
13. Cheng ATA, Hawton K, Lee CTC, Chen THH: The influence of media
reporting of the suicide of a celebrity on suicide rates: a population-
based study. International Journal of Epidemiology 2007, 36(6):1229-1234.
14. Fu KW, Yip PSF: Estimating the risk for suicide following the suicide
deaths of 3 Asian entertainment celebrities: A meta-analytic approach.
Journal of Clinical Psychiatry 2009, 70(6):869-878.
15. Cheng Q, Fu KW, Yip PS: A comparative study of online suicide-related
information in Chinese and English. Journal of Clinical Psychiatry 2011,
72(3):313-319.
16. TEJ Jr: The clustering and contagion of suicide. Current Directions in
Psychological Science 1999, 8(3):89-92.
17. Mesoudi A: The cultural dynamics of copycat suicide. PLoS ONE 2009,
4(9):e7252.
18. Census and Statistics Department of Hong Kong SAR: Thematic
household survey report No. 43. Information technology usage and
penetration In.
Hong Kong; 2009 [http://www.censtatd.gov.hk/
hong_kong_statistics/key_economic_and_social_indicators/index.jsp].
19. Latest indicators. [http://www.internetworldstats.com/asia/tw.htm].
20. China Internet Network Information Centre [CNNIC]: Statistical survey
report on the Internet development in China. China Internet Network
Information Centre 2011.
21. Weeks B, Southwell B: The symbiosis of news coverage and aggregate
online search behavior: Obama, rumors, and presidential politics. Mass
Communication and Society 2010, 13(4):341-360.
22. Stack S: A reanalysis of the impact of non celebrity suicides. Social
Psychiatry and Psychiatric Epidemiology 1990, 25(5):269-273.
23. Fu KW, Chan Y-Y, Yip PSF: Newspaper reporting of suicides in Hong
Kong, Taiwan and Guangzhou: compliance with WHO media guidelines
and epidemiological comparisons. Journal of Epidemiology and Community
Health 2011, 65:928-933.
24. Chen Y-Y, Chen F, Yip PSF: The impact of media reporting of suicide on
actual suicides in Taiwan, 2002-05. Journal of Epidemiology and
Community Health 2011, 65(10):934-940.
25. China Internet Network Information Centre [CNNIC]: Statistical Survey
Report on the Internet Development in China (in Chinese). In: China
Internet Network Information Centre; 2011.
26. Chen F: Maximum local partial likelihood estimators for the counting
process intensity function and its derivatives. Statistica Sinica 2011,
21(1):107-128.
27. Chen F, Yip PSF, Lam KF: On the local polynomial estimation of the
counting process intensity function and its derivatives. Scandinavian
Journal of Statistics 2011.
28. Tousignant M, Mishara BL, Caillaud A, Fortin V, St-Laurent D: The impact of
media coverage of the suicide of a well-known Quebec reporter: the
case of Gaëtan Girouard. Social Science & Medicine 2005, 60(9):1919-1926.
29. Cleveland WS: Robust locally weighted regression and smoothing
scatterplots. Journal of the American Statistical Association 1979, 74:829-836.
30. R Development Core Team: R: A language and environment for statistical
computing. In. Vienna: R Foundation for Statistical Computing; 2010.
31. Ministry of Industry and Information Technology of PRC: 2010 Statistical
communique of the national telecom industry (in Chinese). In: Ministry of
Industry and Information Technology of PRC; 2011.
32. Stack S: Suicide in the Media: A quantitative review of studies based on
nonfictional stories. Suicide and Life-Threatening Behavior 2005,
35(2):121-133.
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 10 of 11
33. Pirkis J, Blood RW, Beautrais A, Burgess P, Skehan J: Media guidelines on
the reporting of suicide. Crisis: The Journal of Crisis Intervention and Suicide
Prevention 2006, 27(2):82-87.
Pre-publication history
The pre-publication history for this paper can be accessed here:
http://www.biomedcentral.com/1471-2458/11/841/prepub
doi:10.1186/1471-2458-11-841
Cite this article as: Cheng et al.: The Foxconn suicides and their media
prominence: is the Werther Effect applicable in China? BMC Public Health
2011 11:841.
Submit your next manuscript to BioMed Central
and take full advantage of:
Convenient online submission
Thorough peer review
No space constraints or color figure charges
Immediate publication on acceptance
Inclusion in PubMed, CAS, Scopus and Google Scholar
Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit
Cheng et al . BMC Public Health 2011, 11:841
http://www.biomedcentral.com/1471-2458/11/841
Page 11 of 11
    • "The advantage of mental health in migrant workers compared to urban workers was found mainly in the older group in our study, whilst younger migrants had lower level of mental wellbeing than their older counterparts . This is in keeping with the recent concern about poor mental health in young Chinese migrants [17,37]. There is no prior systematic investigation of mental health amongst " new generation " or younger Chinese migrant workers in comparison to older migrants [38]. "
    [Show abstract] [Hide abstract] ABSTRACT: There has been a dramatic increase in internal migrant workers in China over recent decades, and there is a recent concern of poor mental health particularly amongst younger or "new generation" migrants who were born in 1980 or later. A cross-sectional study was conducted in Guangzhou city between May and July in 2012. Mental wellbeing was measured using the World Health Organization Five-item Well-Being Index Scale and the 36 Item Short Form Health Survey mental health scale. Linear and logistic regression models were used to investigate the differences between migrant workers and their urban counterparts and between younger and older migrants. Migrant workers (n = 914) showed a small but significant advantage in mental wellbeing compared to their urban counterparts (n = 814). There was some evidence for age modification effect (p for interaction = 0.055-0.095); better mental wellbeing in migrants than urbanites were mainly seen in the older compared to the younger group, and the difference attenuated somewhat after controlling for income satisfaction. Older migrants showed better mental health than younger migrants. Factors that were independently associated with poor mental health in migrants included being male, longer working hours, and income dissatisfaction, whilst older age, factory job, high income, and increased use of social support resources were associated with reduced risk. Efforts to promote mental health amongst migrant workers may be usefully targeted on younger migrants and include measures aimed to improve working conditions, strengthen the social support network, and address age-specific needs.
    Full-text · Article · Dec 2014
    • "Other possible indicators of prominence are based on the physical placement of the article in the news, but as we also use online news sources, this is not feasible in our study. Nonetheless, numerous former studies focused on word count as a base to measure media prominence (Cheng, Chen, & Yip, 2011; Hartley & Coleman, 2008; Hayes et al., 2007; Lee, 2007; Stroud & Kenski, 2007). Based on the statements we also identified the policy positions, one of our key independent variables, in terms of whether (a) actors aimed to block or shape most of the legislative proposal, (b) actors aimed to shape specific parts of it, (c) actors were supporting it, or (d) no clear policy position could be adopted. "
    [Show abstract] [Hide abstract] ABSTRACT: This article examines the coverage of legislative lobbying in European news media. The starting point thereby is that lobbying in the crowded EU-level interest community is not only a struggle for direct access to policymakers, but that in order to realize policy goals many interest groups rely on political attention generated by the media. Our main research question is how media attention is skewed towards particular interests and which factors explain these varying levels of prominence. Our empirical analysis is based on a set of 125 legislative proposals adopted by the European Commission between 2008 and 2010. For all these cases we identified 379 interest organizations that made public statements, we coded the amount of media attention these organized interest gained, the type of statements they made as well as some key organizational features. While the aggregate levels of attention look pretty balanced, our evidence shows that media prominence is skewed towards particular types of interests, more in particularly that organized interests which oppose a proposed policy gain significantly higher levels of media attention.
    Full-text · Article · Jan 2014
    • "Presumptions about the time span in which media reports about fatal suicidal acts are potentially followed by copycat behaviour are diverse. While the effect seems to be greatest three days after the incident (Cheng et al., 2011), in the case of Robert Enke it lasted throughout four weeks (Ladwig et al., 2012). Little is known about the influence on time-spans exceeding this period. "
    [Show abstract] [Hide abstract] ABSTRACT: BACKGROUND: Following the railway suicide of Robert Enke, a famous German football goal keeper, short-term copycat effects have been found. Main aims of the present study were to analyze long-term effects of this incidence and to compare them with overall national suicide data, as well as to investigate possible "anniversary effects". METHODS: For long-term effects, the number of railway suicidal acts in the two years before and after Robert Enke's suicide (10th November 2009) were compared. For anniversary effects, the corresponding 2-week-periods in 2009, 2010 and 2011 were analyzed. Incidence ratios with 95% confidence intervals were computed. RESULTS: Compared to the two years before Enke's suicide the incidence ratio of the number of railway suicidal acts in the 2-year-period following this event increased by 18.8% (95% confidence interval (CI)=11.0-27.1%; p<0.001). The median number of suicidal acts per day increased from 2 to 3 (p<0.001). This effect remains significant after excluding short-term 2-week effects of Enke's suicide. An anniversary effect was not present. The increase of fatal railway suicides between 2007 and 2010 (25%) was significantly different from that for the total number of suicides in Germany (6.6%) (p<0.0001). LIMITATIONS: Due to missing data, analyses regarding gender were limited and regarding age not feasible. CONCLUSIONS: Long-term effects of Enke's suicide on railway suicidal acts in Germany in the sense of copycat behavior are probable as this increase cannot be explained by corresponding changes of the total number of suicides in Germany.
    Full-text · Article · Oct 2012
Show more

Supplementary resources