The Review of Economics and Statistics

卷. C

七月 2018

数字 3

SOCIAL COHESION, RELIGIOUS BELIEFS, AND
THE EFFECT OF PROTESTANTISM ON SUICIDE

Sascha O. Becker and Ludger Woessmann*

Abstract—In an economic theory of suicide, we model social cohesion of the
religious community and religious beliefs about afterlife as two mechanisms
by which Protestantism increases suicide propensity. We build a unique
microregional data set of 452 Prussian counties for 1816 到 1821 和 1869
到 1871, when religiousness was still pervasive. Exploiting the concentric
dispersion of Protestantism around Wittenberg, our instrumental variable
model ﬁnds that Protestantism had a substantial positive effect on suicide.
Results are corroborated in ﬁrst-difference models. Tests relating to the two
mechanisms based on historical church attendance data and modern suicide
data suggest that the sociological channel plays the more important role.

我.

介绍

EVERY year, 超过 800,000 people commit suicide world-

宽的, making suicide a leading cause of death, in partic-
ular among young adults (世界卫生组织, 2014).
This creates far-reaching emotional, 社会的, and economic
ramiﬁcations and invokes major policy efforts to prevent
suicides. In the scientiﬁc literature, religious denomination
has long been observed as an important factor related to
suicide. In Le suicide, a classic example of quantitative
investigation of socially framed individual behavior, Émile
Durkheim (1897) presented aggregate indicators suggest-
ing that Protestantism was a leading correlate of suicide
incidence. The proposition that Protestants have higher sui-
cide rates than Catholics has been “accepted widely enough
for nomination as sociology’s one law” (Pope & Danigelis,
1981). Even today, Protestant countries tend to have sub-
stantially higher suicide rates, suggesting that the relation of

Received for publication March 25, 2015. Revision accepted for publica-

tion June 19, 2017. Editor: Asim I. Khwaja.

* Becker: University of Warwick, CEPR, CESifo, IZA, and ifo; Woess-

mann: University of Munich and ifo Institute, CESifo, IZA, and CAGE.

We thank the editor, three referees, Ran Abramitzky, Robert Barro, Gary
Becker, Davide Cantoni, Francesco Cinnirella, Carl-Johan Dalgaard, Angus
Deaton, Luca De Benedictis, Ray Fisman, Jon Gruber, Gordon Hanson,
Dan Hungerman, Larry Iannaccone, Murat Iyigun, Andrew Oswald, Jared
鲁宾, John Sawkins, Fabian Waldinger, and participants at several sem-
inars for helpful discussion and comments. We also proﬁted from the
exchange with Martin Hofmann, who wrote his master’s thesis on an eco-
nomic analysis of confession-speciﬁc suicide rates under our supervision.
Capable research assistance by Martin Hofmann, Laurenz Detsch, and Mar-
tin Hoben; support by Urban Janisch with the remote access to the German
Mortality Statistics; and support by the Pact for Research and Innovation
of the Leibniz Association, the German Science Foundation through CRC
TRR 190, and the ESRC Centre for Competitive Advantage in the Global
Economy (ESRC grant ES/L011719/1) are gratefully acknowledged.

A supplemental appendix is available online at http://www.mitpress

journals.org/doi/suppl/10.1162/rest_a_00708.

religion and suicide remains a vital topic.1 Several contri-
butions have so far revealed the usefulness of investigating
suicide from an economics point of view (Hamermesh &
Soss, 1974; Becker & Posner, 2004; 陈等人。, 2012).2
But religious denomination, a leading established correlate
of suicide in the sociological literature, has received surpris-
ingly little attention in the economics literature, 尽管
recent burst of interest in issues of culture and religion.3
While the economics literature on happiness and subjec-
tive well-being considers suicide as a revealed-preference
outcome measure of utmost unhappiness (Oswald, 1997;
Layard, 2005), these analyses so far have not been linked
to religious denomination.

This paper makes three contributions to the economic
analysis of religion and suicide. 第一的, in section II, we model
social cohesion and religious beliefs as two channels through
which Protestantism may affect suicide in the framework of
an economic theory of suicide. We show how a higher suicide
rate of Protestants relative to Catholics can be understood
as the outcome of denominational differences in community
integration and in theological doctrine. 第二, in section III,
we provide new microregional evidence from Prussia in the
nineteenth century that the effect of Protestantism on sui-
cide may indeed be causal. 第三, in section IV, we use this
empirical setting to devise tests that discriminate between
the sociological and the theological explanations. Our results
suggest that Protestantism is a leading explanatory factor for
suicide rates and that the sociological mechanism plays an
重要的, if not dominant, 角色.

Our empirical setting is Prussia in the nineteenth century.
Apart from mirroring the perspective of Durkheim’s (1897)
工作, the nineteenth century has the advantage that virtu-
ally everybody was a member of a religious denomination

1 Among the ten OECD countries in which either Protestants or Catholics
make up over 85% of the population in 2000, the average suicide rate
among the four Protestant countries is 15.5 suicides per 100,000 因哈比-
短裤, whereas it is 8.9 among the six Catholic countries (suicide data from
经合组织, 2009; religion data from Barrett, Kurian, & 约翰逊, 2001). 也可以看看
黄 (1996) and Helliwell (2007) for cross-country studies of religion and
suicide.

2 卡特勒, Glaeser, and Norberg (2001), Daly and Wilson (2009), Daly et al.

(2011), and Daly, Wilson, and Johnson (2013) are further examples.

3 The economics literature on culture and religion (Iannaccone, 1998;
Guiso, Sapienza, & Zingales, 2006; Iyer, 2016) does not emphasize suicide
as a possible outcome.

The Review of Economics and Statistics, 七月 2018, 100(3): 377–391
© 2018 由哈佛大学和麻省理工学院的校长和研究员撰写. 根据知识共享署名发布 4.0
Unported (抄送 4.0) 执照.
土井:10.1162/rest_a_00708

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

378

THE REVIEW OF ECONOMICS AND STATISTICS

and that religion pervaded all aspects of life. The Prussian
perspective offers the opportunity to compare nonminority
occurrences of the two religious denominations within a
common setting of political governance, 机构, juris-
措辞, 语言, and basic culture. Combining suicide data
administered by local police departments from 1869 到 1871
with rich census data, we build a unique new microregional
data set on suicide, religion, and relevant covariates for 452
Prussian counties. We also use data from 1816 到 1821.

A fundamental challenge for empirical identiﬁcation is
self-selection of more suicide-prone people into Protes-
tantism, as hypothesized already in 1919 by the neurolo-
gist Kollarits (1919). 然而, the endogeneity bias may
also go in the opposite direction; 例如, during the
Reformation, Protestantism may have spread more easily
to regions where people were willing to take matters into
their own hands and change their lives, which may be nega-
tively related to suicide proneness. While manifold, existing
studies do not address this fundamental endogeneity prob-
莱姆. To identify the causal effect of Protestantism, 我们用
distance to Wittenberg as an instrumental variable tracing
the initial spread of the Reformation from its epicenter
(Becker & Woessmann, 2009). We vindicate the validity of
the instrument with evidence that it is orthogonal to impor-
tant correlates of suicide rates in 1517, before the start of
the Reformation.

Our results show that Protestantism had a signiﬁcant pos-
itive effect on suicides in Prussia in both the early and late
十九世纪. Protestantism increased the annual sui-
cide rate per 100,000 inhabitants, which has a mean of
thirteen suicides from 1869 到 1871, by about ﬁfteen to
twenty suicides. Channels such as economic moderniza-
tion and literacy, which are also affected by Protestantism,
seem to play only a minor role in this effect, suggesting
that it is an effect of the Reformation per se rather than
of its nonreligious outcomes. The empirical result proves
very robust to a large set of robustness tests. 此外,
exploiting differential changes in Protestant shares across
counties between 1816 和 1871, a ﬁrst-difference model
corroborates a positive effect of Protestantism on suicide.
因此, Protestantism seems to bring positive effects for some
people and negative effects for others: For the majority of
人口, it raises economic prosperity through higher
人力资本 (Becker & Woessmann, 2009), but for the
select group of people in a suicidal state of mind, it may tip
the balance toward ending their lives.

We devise several tests to tentatively discriminate between
the two classes of theoretical models. All turn out to speak
in favor of the empirical importance of sociological com-
pared to theological channels. Among others, the effect of
Protestantism on suicide is lower in counties where church
attendance is high, implying closer community integration.
相比之下, according to the theological channel, 更高
church attendance would indicate a more devout belief in
Protestant doctrine, consistent with higher suicide rates.
Modern individual-level data show that in 1992, suicides

are higher among Protestants than among Catholics, 但
even higher among those without a religious afﬁliation.
经过 2009, the Protestant-Catholic difference is substantially
reduced and only the religiously nonafﬁliated have substan-
tially higher suicide proneness. 再次, this speaks against the
theological channel because those who remain in the church
are presumably the most devout believers in Protestant
doctrine.

Our economic analysis of religion and suicide contributes
to the major debate in sociology since Durkheim (1897).
Several contributions have questioned the empirical regular-
ity that Protestants have higher suicide rates than Catholics.4
Durkheim’s theoretical hypotheses have also been subjected
to major criticism (see Pope & Danigelis, 1981). Stark,
多伊尔, and Rushing (1983) go as far as ﬁnding Durkheim’s
argument “inconsistent and unconvincing” and “amazingly
uninformed and misleading about elementary features of
religion in 19th century Europe” (p. 120). 相比之下, 两个都
our evidence of a strong causal effect of Protestantism on sui-
cide in nineteenth-century Prussia and the indications that
social cohesion may have been a stronger mechanism in
this than religious doctrine corroborate Durkheim’s (1897)
original contribution.

二. A Theory of Religion-Speciﬁc Suicide

We see two classes of theoretical reasoning—one related
to social cohesion (sociological channel for short),
这
other to individual religious beliefs (theological channel, 为了
short)—that have a bearing on the rationality of the act of
suicide in Catholicism and in Protestantism. We model these
denominational differences in the framework of an economic
theory of suicide. We brieﬂy sketch the mechanisms here;
online appendix A presents the model in detail.

The scientiﬁc study of suicidal behavior and its preven-
的, or suicidology, is the topic of several disciplines (在线的
appendix A.1). As stressed by the psychology of suicide,
most suicides are committed in a depressed mental state that
is transient and diverges from a person’s usual state of pref-
erences. But even in this state, suicidal persons may take the
costs and beneﬁts of their action into account. 在这个意义上,
an intertemporal utility-maximizing framework of standard
economic theory may provide insight into an understanding
of what might lead these people to carry out their suicide or
not and exit the suicidal state.

Our model framework extends the economic theory of sui-
cide developed by Hamermesh and Soss (1974) and Becker
and Posner (2004). Suicide is modeled as forward-looking
utility-maximizing behavior. We view this model as applying
to the suicidal state of mind depicted in the psychological
研究, where preference parameters may differ from those

4 See Pope and Danigelis (1981), Bankston, 艾伦, and Cunningham
(1983), van Poppel and Day (1996), and Simpson (1998) for leading
examples of the controversy.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

THE EFFECT OF PROTESTANTISM ON SUICIDE

379

during normal mental states. In a process of rational deci-
sion making, individuals in the suicidal state compare the
expected utility from living with that from death. If the latter
is greater than the former, committing suicide will maximize
公用事业 (see online appendix A.2 for a formal derivation).

To understand how religious differences may affect the
propensity to commit suicide, we start by modeling a soci-
ological aspect of denominational differences. Durkheim
(1897) emphasized that Protestant doctrine encourages inde-
thought and religious individualism, decreasing
pendent
social cohesion relative to a more uniﬁed Catholic commu-
本质, which tends to protect people from committing suicide.
If there is mutual interdependence in preferences, 事实
that there are others who would suffer from a person’s sui-
cide will tend to discourage people from committing suicide.
In terms of our simple economic model of suicide, 较低的
cohesion of the Protestant community leads to the prediction
that suicide rates would be higher in Protestant communities
than in Catholic communities (see online appendix A.3 for
细节).

To this sociological mechanism, we add a couple of the-
ological explanations. When afterlife is added to the model,
Protestant-Catholic differences rooted more deeply in reli-
gious doctrine affect suicidal behavior through the utility or
disutility of afterlife (see online appendix A.4 for details).
尤其, Protestantism tends to stress that human sal-
vation is by God’s grace alone, and not by any merit of
a person’s own work, whereas Catholicism allows human
deeds and sins to affect God’s judgment. Committing suicide
thus entails the disutility of foregoing paradise for Catholics
but not for Protestants. 此外, the confession of sins
is a holy sacrament in Catholicism but not in Protestantism.
Since suicide is the only sin that (by deﬁnition) can no longer
be confessed, this additionally creates a substitution effect
that diverts Catholics from committing suicide toward other
forms of behavior considered in times of desperation.

三、. Evidence on the Effect of Protestantism
on Suicide in Nineteenth-Century Prussia

A. Data and Descriptive Statistics

时间,

Prussia provides uniquely rich census-based data to study
the relation of suicide, religion, and covariates at the county
level in the nineteenth century. The focus on the nineteenth
century has the advantage that religiosity was pervasive
in the sense that almost everybody had a
在
religious afﬁliation and that religion affected virtually all
dimensions of everyday life. The focus on Prussia allows
us to exploit variation between counties with nonminor-
ity Protestant and Catholic populations within the setting
of one country. 尤其, the Prussian population was
about two-thirds Protestants and one-third Catholics, 和
a majority of counties were close to having a uniformly
Catholic or uniformly Protestant population, so that no

denomination was an extreme minority. This may be impor-
tant to exclude that religious factors are confounded with
particular behavior in religious minorities. The religious
division of Prussian territory goes back to Reformation
times and was solidiﬁed by the exceptional individual reli-
gious freedom granted in Prussia at least since Frederick the
Great in the mid-eighteenth century. In its nineteenth-century
shape, Prussia had Wittenberg, the birthplace of the Refor-
运动, at its center, where Protestantism originated and was
conserved in its purest form. Prussia had uniform laws and
institutional frameworks, and ofﬁcial suicide ﬁgures were
collected as early as 1816. In contrast to cross-national anal-
yses, this makes county-level data within Prussia directly
comparable.

We have religion and suicide data for two points in time,
one early (1816 到 1821) and one late (1869 到 1871) 在里面
十九世纪. Our analyses mainly focus on the latter
period when suicide data are more reliable and background
data richer. But the ﬁrst time for which suicide statistics
were collected for the whole of Prussia is the 1816–1821
时期 (Mützell, 1825).5 The data average suicides over sev-
eral years, reducing noise due to random jumps in suicide
incidents. The data cover all 306 Prussian counties at the
时间. 这 1816 Population Census provides data on religion
and background controls (see online appendix C for details
on the different data sources).

We also digitized suicide statistics for 1869 到 1871,
again averaged over consecutive years. We combine these
data with a rich set of variables that the literature consid-
ers as determinants of suicide rates. Most prominent, 这
1871 Population Census contains shares of Protestants in
the county population, demographic characteristics, and lit-
eracy rates (see Becker et al., 2014). It also provides shares
of the population with different forms of physical and mental
disabilities. 这 1882 Occupation Census provides data on
the occupational structure, used as indicators of the stage
of industrial development. The data cover all 452 Prus-
sian counties (Kreise) 当时, divided into 11 provinces
(Provinzen) 和 35 districts (Regierungsbezirke).

There is a difference in the way suicide data were col-
lected at the beginning and end of the nineteenth century
(Hilse, 1871). 为了 1816 到 1821, data on suicides were drawn
from the local burial and death registers, which were often
run by the church. 在 1868, dedicated suicide statistics were
introduced for which the local state administration (the city
council or the local police) recorded every civilian suicide
on a separate data sheet. Background information on the per-
son committing suicide and the suicide circumstances were
collected with the explicit aim of understanding the factors
explaining suicides. The new data collection method was
used as the basis of very detailed suicide statistics from 1869
向前. The Prussian Statistical Ofﬁce exerted extensive
effort to ensure high data quality and dedicated eighty pages

5 Prussian statistics have published data on suicides as a death cause since

1777 (威尔克, 2004).

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

380

THE REVIEW OF ECONOMICS AND STATISTICS

Table 1.—Descriptive Statistics, Prussia, 1871

Suicide rate (每 100,000 inhabitants)
Suicide proportion (每 1,000 deaths)
Share of Protestants
Share of population < 15 years Share of population > 60 年
Average household size
Share of population living in towns
Share of labor force in manufacturing and services (1882)
Share of literate adults
Distance to Wittenberg (在 1,000 km)
Share of females
Share of Jews
Share of population born in municipality
Share of population of Prussian origin
Share blind (× 100)
Share deaf-mute (× 100)
Share insane (× 100)
Fatal accident rate (每 100,000 inhabitants)
Fatal accident proportion (每 1,000 deaths)
Latitude (in rad)
Longitude (in rad)
Year when annexed by Prussia

意思是
(1)

13.00
4.78
.64
.36
.07
4.79
.28
.34
.88
.33
.51
.01
.59
.99
.09
.10
.23
42.35
15.17
.91
.22
1,751.69

标清
(2)

8.33
3.17
.38
.03
.02
.34
.22
.15
.13
.15
.02
.01
.12
.02
.03
.05
.17
15.80
5.00
.03
.08
111.05

Minimum
(3)

Maximum
(4)

.00
.00
.003
.23
.03
3.83
.00
.08
.37
.00
.44
.00
.32
.74
.03
.02
.02
9.37
3.77
.84
.11
1,525

37.06
15.76
1.00
.43
.11
5.86
1.00
.82
.99
.73
.55
.13
.87
1.00
.24
.42
1.56
114.52
37.48
.97
.39
1,866

Suicide rates are average annual rates in 1869 到 1871. Data for 452 Prussian counties from the 1869–1871 Suicide Statistics, 这 1871 Population Census, 和 1882 Occupation census. See main text and online

appendix C for details.

in its quarterly journal to providing background information
and ﬁrst results on the new suicide statistics (Hilse, 1871).
The care given to data collection and the amount of detail
given in the suicide tables are impressive and reassuring
signs of data quality.6

Descriptive statistics for the 1869–1871 period in table 1
reveal that the average annual suicide rate across all Prussian
counties was 13.0 每 100,000 inhabitants, ranging between
0 in only one county (Adenau) 到 37.1 (Schönau). 上层
panel of ﬁgure 1 shows substantial geographic variation in
suicide rates across Prussia. Prussian suicide levels are some-
what higher than in Germany today, where the suicide rate
曾是 10.3 每 100,000 inhabitants in 2004 (经合组织, 2009).
The comparison of our historic data with modern data pro-
vides no indication of a systematic underreporting in the late
十九世纪, unless one believes that suicide rates had
a signiﬁcant downward trend over the twentieth century.7

Another check on whether there is systematic under-
reporting of suicides in some counties is to cross-check the
suicide data with other mortality data. 因为, 尤其
in Catholic parishes, a religious funeral ceremony was some-
times not granted for proven suicides, there may in principle
be an incentive to underreport suicides and classify them as
fatal accidents (Kollarits, 1919). If this were the case, 这
incidence of reported suicides and fatal accidents should be
negatively correlated. In our data set, suicide rates and fatal
accident rates are in fact uncorrelated; their raw correlation

6 例如, eleven different means of suicide are provided, hanging
and drowning being the two most widespread categories (see table A.1 in
the online appendix).

7 La Vecchia, Lucchini, and Levi (1994) do not ﬁnd substantial trends in
suicide rates in developed countries from 1955 到 1989, and Chen et al.
(2012) refer to substantial increases.

is −0.004 ( p-value 0.932) in the full sample and 0.110
( p-value 0.362) 在里面 71 counties with a Catholic share
higher than 90%. This indicates that systematic underreport-
ing of suicides is unlikely. According to Kollarits (1919),
the standard way to get a religious funeral ceremony was to
claim that the suicide was caused by aberration. 在这种情况下,
even the Catholic Church approved a religious ceremony,
and suicide rates and their denominational differences are
not misreported.

The average share of Protestants in a county was 64.2%
在 1871, 反对 34.5% Catholics (the remainder being 1.1%
Jews and 0.2% other Christian denominations). 因此, 两个都
Protestants and Catholics are not just a small minority but
constitute a sizable fraction of the Prussian population.
此外, there is substantial variation across counties,
ranging essentially from 0 到 100% Protestants or Catholics.
多于 75% of the counties have a share of at least 80%
of either Protestants or Catholics, 以及超过 60% 有
a share of at least 90% of one denominational group. 在
restricted analyses, we even focus on samples of countries
where the share of Protestants is smaller than 2% or larger
比 98% 甚至 0.1% 和 99.9%.

The bottom panel of ﬁgure 1 depicts the geographic vari-
ation of Protestant shares across Prussia. The close mapping
between the geographic distribution of Protestant shares and
suicide rates is directly evident. 实际上, the raw correlation
between the two across the 452 counties is 0.66 (statistically
signiﬁcant at the 1% 等级). When plotting the two against
彼此 (ﬁgure A.1 in the online appendix), 有一个
clear positive association between the share of Protestants
in a county and the suicide rate, and the average suicide
rate is notably higher in all-Protestant than in all-Catholic
counties.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

THE EFFECT OF PROTESTANTISM ON SUICIDE

381

Figure 1.—Suicides and Protestantism in Prussia, 1871

Suicide rate (average annual suicides per 100,000 inhabitants), 1869–1871, and share of Protes-
短裤, 1871. County-level depiction based on 1869–1871 Suicide Statistics and 1871 Population Census,
分别. See online appendix C for data details.

乙. Basic Evidence from 1869 到 1871

To probe the association between Protestantism and
suicide in a multivariate setting, we estimate a simple
least-squares model,8

SUICi = α + β PROTi + Xi γ(西德:2) + εi,

(1)

where SUICi is the suicide rate in county i, PROT is the
share of Protestants in the county, and X is a set of control
变量. Our most basic set of controls includes the shares
of the county population below 15 years of age and above 60
年龄, 分别, and average household size. 这样的

measures of age and family patterns are standard determi-
nants considered in suicide equations. In richer models, 我们
will also consider a host of additional possible correlates of
suicide as control variables (see Helliwell, 2007, and Chen
等人。, 2012, for extensive overviews of factors considered in
empirical suicide research).

The ﬁrst column of table 2 shows the strong positive
association between the Protestant share and the suicide
速度. 平均而言, all-Protestant counties have a suicide
rate that is 14.5 suicides per 100,000 inhabitants higher
than all-Catholic counties. Viewed against an average sui-
cide rate of 13.0, this is a substantial difference across
religious denominations. Column 2 adds the list of basic
demographic control variables. The signiﬁcant positive asso-
ciation between Protestantism and suicide remains largely
unchanged in the multivariate speciﬁcation. Suicide rates
are signiﬁcantly negatively related to larger shares of young
(以下 15 年) and old (超过 60 年) 人口. The fact
that suicide initially increases with age is a standard result in
suicide research. The inversely U-shaped pattern of suicide
rates declining again with larger shares of old people may
indicate a declining suicide inclination after reaching a cer-
tain age. As an indicator of longevity, it may also capture an
effect of the level of economic development, which may pro-
tect from suicide disposition. The negative relation of suicide
rates with average household size mirrors the importance of
the family generally found in the suicide literature.

Columns 3 到 5 add further control variables. 以前的
work has found urbanization, economic conditions, and edu-
cation to be factors related to suicide (Helliwell, 2007; 陈
等人。, 2012). We add the share of population living in towns,
the share of the labor force working in manufacturing and
服务 (as a measure of economic development), 和
share of literates to the basic model. None of these measures
enters the model signiﬁcantly, and the point estimate on the
share of Protestants is barely affected. Column 6 adds a set of
dummies for the 35 Prussian districts (Regierungsbezirke),
the administrative layer between counties and provinces, 到
该模型. This speciﬁcation excludes all the variation that
exists across districts and exploits only the within-district
variation. To the extent that there is unobserved regional
heterogeneity, district dummies should capture most of its
substance. While the estimated association between Protes-
tantism and suicide is somewhat reduced in magnitude, 它
remains highly robust.

Column 7 uses the suicide proportion—the number of sui-
cides divided by the ﬂow of deaths in the same period—as
an alternative dependent variable. This measure takes into
account that average mortality rates differ across counties.
再次, there is a signiﬁcant association of Protestantism with
suicides. The lower point estimate is in line with the smaller
value range of this variable (见表 1).

8 Our qualitative results are conﬁrmed in Poisson and negative binomial
regression models that use the number of suicides rather than suicide rates
as the dependent variable (not shown). 相似地, the signiﬁcance of our
main results holds in models that cluster standard errors at the level of 35
districts (not shown).

Identifying Exogenous Variation in Protestantism

A concern with the evidence so far is that religious afﬁlia-
tion may not be exogenous to the suicide model. Speciﬁcally,

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

382

THE REVIEW OF ECONOMICS AND STATISTICS

Table 2.—Protestantism and Suicide in Prussia, 1871

Dependent Variable

Suicide Rate
(每 100,000 inhabitants)

Suicide Proportion
(每 1,000 deaths)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Share of Protestants

Share of population < 15 years Share of population > 60 年

Average household size

Share of population living in towns

Share of labor force in manufacturing

and services (1882)
Share of literate adults

35 district dummies
持续的

观察结果
R2

14.496
(.669)∗∗∗

12.328
(.612)∗∗∗

−70.781
(10.911)∗∗∗
−30.120
(18.102)∗
−7.575

(.758)∗∗∗

12.306
(.611)∗∗∗

−66.868
(15.498)∗∗∗
−23.869
(23.812)
−7.529

(.755)∗∗∗
.754
(1.838)

12.411
(.634)∗∗∗

−66.580
(15.590)∗∗∗
−22.354
(24.118)
−7.364

(.771)∗∗∗
.091
(1.862)
1.807
(1.683)

12.528
(.705)∗∗∗

−67.552
(15.811)∗∗∗
−15.241
(27.741)
−7.317

(.761)∗∗∗
.089
(1.867)
2.437
(2.089)
−1.614
(3.054)

3.691
(.370)∗∗∗
452
.433

68.928
(5.626)∗∗∗
452
.627

66.655
(8.046)∗∗∗
452
.627

65.148
(8.464)∗∗∗
452
.628

65.868
(8.738)∗∗∗
452
.628

9.812
(.906)∗∗∗

−57.218
(17.233)∗∗∗
13.799
(34.247)
−1.727
(1.568)
.212
(1.826)
5.550
(2.646)∗∗
3.900
(4.040)
是的
23.571
(13.034)∗
452
.738

4.928
(.272)∗∗∗

−21.537

(6.394)∗∗∗
9.570
(11.133)
−2.077

(.285)∗∗∗
.548
(.686)
.118
(.790)
.020
(1.123)

18.352
(3.372)∗∗∗
452
.611

Ordinary least squares (OLS) estimation. Heteroskedasticity-robust standard errors in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Data for Prussian counties from the 1869–1871 Suicide Statistics, 这 1871

Population Census, 和 1882 Occupation Census. See main text and online appendix C for details.

whether a person adheres to a speciﬁc faith may to some
extent be a choice variable that is correlated with the error
term of equation (1). 例如, in the early twentieth
世纪, Kollarits (1919), a Hungarian publishing in a Ger-
man journal of neurology and psychiatry, hypothesized that
the higher incidence of suicide among Protestants may sim-
ply result from selection of suicide-prone people into the
Protestant denomination. 然而, direct conversion was
in fact minimal in the nineteenth century: 仅有的 0.01% 的
Catholics (或者 766 out of more than 7 million Catholics) 骗局-
verted to Protestantism per year over the period 1859 到
1867, mostly in the course of marriage to a Protestant partner
(Hilse, 1869).

But endogeneity may take another form of unobserved
heterogeneity, in that three centuries earlier, during the Ref-
ormation, regional conversion to the new Protestant faith
may not have been orthogonal to suicide proneness, 哪个
may exhibit strong intertemporal persistence. Most of the
denominational variation across Prussia in the nineteenth
century can be traced back to denominational choices of
local rulers in the roughly 300 political entities that made
up Germany during the Reformation in the sixteenth and
early seventeenth centuries, mostly motivated by religious
conviction and power politics vis-à-vis the pope and the Ger-
man emperor. While it seems unlikely that the adoption of
Protestantism was directly related to pre-Reformation pat-
terns in suicide, it might have been indirectly related to
correlates of suicide such as economic conditions, 城市的-
性, 教育, and mental disposition. 例如, 地区
where people are naturally inclined to try to change a bad
status quo rather than giving up may have been more will-
ing to adopt the new denomination that emerged from a
protest movement (“Protestantism”), and such people may
also be less prone to commit suicide when matters turn bad.

Such issues of causality pose a fundamental challenge for
empirical identiﬁcation that has not been directly addressed
在里面 (mostly sociological) literature so far.

To identify exogenous variation, we exploit the concentric
spread of the Reformation from Wittenberg, where Luther
initiated the new denomination. As is visible in the bot-
tom panel of ﬁgure 1, the Reformation spread in the areas
around Wittenberg but was less successful farther away from
Wittenberg. The geographically concentric dispersion of the
Reformation allows us to employ an instrumental variable
(IV) strategy that uses a county’s distance to Wittenberg as
an instrument for the share of Protestants in the county. 我们
thereby restrict the analysis to a speciﬁc part of the denomi-
national variation that is arguably exogenous to variation in
important drivers of suicide rates. Our identifying assump-
的, which we probe in greater detail below, is that the
concentric pattern is unrelated to suicide apart from its effect
through Protestantism.

桌子 3 reports results of the IV estimation of the effect
of Protestantism on suicide rates. Distance to Wittenberg is
a strong instrument for the share of Protestants in a county,
as is evident from an F-statistic of the instrument in the ﬁrst
stage of 23 到 47 (depending on the included controls). 每个
100 km distance to Wittenberg is associated with a Protestant
share that is 7 到 9 percentage points lower (columns 1 到
4). The second stage uses only that part of the variation in
Protestant shares that is due to distance to Wittenberg to
predict suicide rates.

The positive effect of Protestantism on suicide rates is
highly robust in the IV speciﬁcations (columns 5 到 8). 在
事实, the IV point estimates are signiﬁcantly higher than the
OLS estimates. Depending on the model, A 10 percentage
观点
increase in the share of Protestants in a county
increases the suicide rate by 2.0 到 2.4 suicides per 100,000

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

THE EFFECT OF PROTESTANTISM ON SUICIDE

383

Table 3.—Instrumental Variable Estimates Using Distance to Wittenberg

Dependent Variable

First Stage
Share of Protestants

Second Stage
Suicide Rate (每 100,000 inhabitants)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Share of Protestants

Distance to Wittenberg (在 1,000 km)

Share of population < 15 years Share of population > 60 年

Average household size

Share of population living in towns

Share of labor force in manufacturing

and services (1882)
Share of literate adults

持续的

观察结果
R2
F-statistic (instrument)

−.936
(.137)∗∗∗

−.863
(.145)∗∗∗
−.599
(.739)
−2.550
(1.587)
−.155
(.059)∗∗∗
-.048
(.101)

−.909
(.136)∗∗∗
−.747
(.742)
−3.177
(1.491)∗∗
−.195
(.059)∗∗∗
.149
(.104)
−.587
(.122)∗∗∗

.947
(.038)∗∗∗
452
.135

2.084
(.454)∗∗∗
452
.152

2.535
(.443)∗∗∗
452
.189

−.693
(.145)∗∗∗
.093
(.738)
−7.136
(1.620)∗∗∗
−.226
(.059)∗∗∗
.172
(.099)∗
−.940
(.130)∗∗∗
1.053
(.200)∗∗∗
1.797
(.451)∗∗∗
452
.235

28.019
(3.051)∗∗∗

20.485
(2.672)∗∗∗

19.969
(2.417)∗∗∗

24.016
(3.836)∗∗∗

−68.989
(17.176)∗∗∗
−23.759
(27.266)
−5.467
(1.159)∗∗∗
−.686
(1.968)

−67.873
(17.065)∗∗∗
−19.051
(26.607)
−5.163
(1.179)∗∗∗
−2.604
(2.084)
5.628
(2.509)∗∗

−4.988
(2.084)∗∗
452
.056
46.807

52.669
(11.255)∗∗∗
452
.498
35.547

49.416
(11.467)∗∗∗
452
.521
44.882

−79.150
(18.385)∗∗∗
61.177
(45.356)
−3.845
(1.590)∗∗
−3.605
(2.345)
14.021
(4.855)∗∗∗

−17.929

(6.076)∗∗∗
51.711
(13.031)∗∗∗
452
.406
22.793

Instrumental variable (IV) estimation. Heteroskedasticity-robust standard errors in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Data for Prussian counties from the 1869–1871 Suicide Statistics, 这 1871

Population Census, 和 1882 Occupation Census. See main text and online appendix C for details.

inhabitants. The pattern of IV and OLS results suggests
that without the Reformation, suicide rates would have been
lower in regions that turned Protestant due to their proximity
to Wittenberg (“compliers”) than in regions that remained
Catholic. This negative bias in the OLS estimates is con-
sistent with a Reformation pattern where regions with a less
suicide-prone population tended to select into Protestantism.

D. Validity of the Instrumental Variable Strategy

The validity of the IV model rests on the assumption
that the initial concentric spread of the Reformation led
to exogenous variation in the Protestant share (see online
appendix B.1 for details on the analyses summarized in
this section). Factors that likely contributed to the historical
diffusion process include costs of traveling and informa-
tion diffusion, increasing dissimilarity of dialects, Electoral
Saxony as an early role model, regional political alliances,
and historical randomness. Consistent data on suicide rates
before the Reformation are not available. But the lack of a
signiﬁcant association of the distance-to-Wittenberg instru-
ment with several proxies of suicide proneness observed
before the onset of the Reformation—including proxies
for economic conditions, urbanity, 教育, 和文化
predisposition—lends credibility to assuming exogeneity of
the instrument. There is also ample anecdotal evidence that
the initial spread of Protestantism was indeed associated with
increases in suicide incidences.

仍然, the fact that we have shown in Becker and Woess-
mann (2009) that Protestantism affected literacy and eco-
nomic development raises a question of interpretation: Was
it these other outcomes of the Reformation, 而不是
Protestantism per se, that led to the increase in suicides?

As shown in table 3, the estimated effect of the Protestant
share on suicide rates is largely unaffected by conditioning
on the share of the workforce that moved out of agriculture,
and it in fact increases (although not statistically signiﬁ-
cantly so) when conditioning on the share of literate adults,
which is signiﬁcantly negatively associated with suicides in
this speciﬁcation. This suggests that the estimated effect of
Protestantism on suicides is barely affected by any effect of
the Reformation on literacy and economic development. 这
conclusion is corroborated when adding controls for further
dimensions of educational and economic development and
for proxies for two additional potential nonreligious out-
comes of the Reformation, income inequality and the number
of social protests (see table A.2 in the online appendix). Nei-
ther measure enters signiﬁcantly or affects the estimate on
Protestantism. 全面的, the results seem most consistent with
a strong effect of Protestantism per se on suicides.

乙. Robustness Analyses

As shown in online appendix B, the empirical results prove
very robust to a large set of robustness tests. Among oth-
呃, we control for additional demographic variables such
as gender, the share of Jews, internal and external migra-
的, and the share of married people (see table A.3 in the
online appendix). To account for differences in suicides that
stem from aspects of general mental illness that are not well
depicted in a rational choice framework, we use information
on the share of people classiﬁed as having physical or mental
disabilities, including being “insane.” This measure does not
vary by denomination in our data, and holding the shares of
people with different disabilities constant does not affect our
结果. 此外, we can add geographic controls, 例如

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

384

THE REVIEW OF ECONOMICS AND STATISTICS

latitude, longitude, and their interaction; altitude; and a set
of dummies indicating the year in which the county became
part of Prussia. We also control for local weather conditions
such as rainfall and temperature (table A.4 in the online
appendix).

To rule out bias from denomination-speciﬁc reporting
bias, we analyze whether some suicides might be hidden
as fatal accidents or other alternative death causes. Neither
fatal accident rates nor total mortality rates are higher in
Protestant areas, and controlling for the fatal accident rate
does not alter our results (see table A.5 in the online appen-
dix). To account for denominational sorting into occupations
with different frequencies of fatal accidents, we also include
an extensive set of controls for employment shares in 32
部门. Results are also robust when using the suicide pro-
portion (suicides per death incidents) rather than the suicide
速度 (suicides per inhabitants) as an alternative outcome
措施.

We also look into effects of religious minorities and
religious concentration (table A.6 in the online appendix).
Cross-tabulated data conﬁrm that the county-level results do
not derive from ecological fallacy (table A.7 in the online
appendix). They also allow us to probe into denomination-
speciﬁc suicides by gender and effects of being a religious
minority on suicide, rejecting the existence of important
nonlinearities in the effect of Protestantism on suicide.

F. 证据来自 1816 到 1821

While the 1869–1871 data are in the ﬁrst statistical investi-
gation speciﬁcally devised to analyze suicides, ofﬁcial burial
and death registers provide data on suicides as early as 1816
到 1821. These are the earliest data covering all of Prus-
是, again available at the county level. Suicide rates are
reported separately by gender for each county. 平均而言,
male suicide rates are about four times higher than female
suicide rates (see table A.8 in the online appendix). The set
of control variables available in the 1816 Population Census
is not as rich as in the later data. 然而, the same types of
basic demographic control variables are available: the share
of the population younger than 15 years and the share older
比 60 年, as well as the share of the population living
in towns. 此外, the number of public buildings per
capita can serve as an indicator of economic development
and the enrollment rate in primary schools as a measure of
教育, and we again have information on fatal accident
费率.

在 6.5 suicides per 100,000 inhabitants, the average sui-
cide rate in the 1816–1821 data is only half the average
suicide rate reported in the 1869–1871 data. This raises the
concern of possible underreporting of suicides in the ofﬁcial
burial and death registers, where some of the suicides may be
classiﬁed as fatal accidents. This may be particularly the case
where priests denied a church burial ceremony for those who
committed suicide (a practice prohibited by Prussian law
only in 1845; see Hilse, 1871). 然而, while underreport-
ing of suicides might affect the size of the estimated effects, 它

would affect the qualitative results only to the extent that the
degree of underreporting varies by denomination. If we take
the 1869–1871 data as a benchmark, we can assess the rela-
tive difference in reported suicides over time for Protestant
and Catholic counties. Counties with a share of Protestants
higher than 90% have an average suicide rate of 9.3 suicides
每 100,000 inhabitants in 1816 到 1821, 相比 17.4 在
1869 到 1871. In Protestant counties, reported suicides from
1816 到 1821 are thus lower by a factor of 1.9. Counties with
a share of Catholics higher than 90% have an average sui-
cide rate of 2.8 从 1816 到 1821, 相比 4.7 为了 1869
到 1871. In Catholic counties, reported suicides from 1816
到 1821 are thus lower by a factor of 1.7. This is an indica-
tion that, if anything, Protestants underreport slightly more
in the 1816–1821 period compared to Catholics not only in
absolute terms but even in relative terms, putting the stakes
against ﬁnding an effect of Protestantism for 1816 到 1821.
此外, we can again control for fatal accident rates in
our regressions to guard against bias from misclassiﬁcation
of suicides as fatal accidents.

Suicide rates in all-Protestant counties are 7.2 larger
than in all-Catholic counties on average. This difference is
reduced to 4.7 but remains highly signiﬁcant in OLS regres-
sions that control for the age structure of the population,
urbanization, public buildings, and school enrollment (桌子
A.9 in the online appendix). Both male and female suicide
rates are signiﬁcantly higher in Protestant areas. 然而,
as a direct corollary of the substantially higher male suicide
费率, the point estimate on Protestantism is substantially
higher for men than for women. 实际上, the male effect
为了 1816 到 1821 is quantitatively in the same range as the
average effect for 1869 到 1871.

桌子 4 reports the IV results that use distance to Wit-
tenberg as an instrument for the share of Protestants. 这
IV estimates suggest that Protestantism raises male suicide
rates by 23.4 suicides per 100,000 inhabitants, female suicide
rates by 7.1, and average suicide rates by 15.0. To exclude
possible bias from underreporting of suicides as accidents,
columns 3 到 5 control for fatal accident rates. Fatal acci-
dent rates are not signiﬁcantly related to suicide rates in the
multivariate regressions, and the estimated effect of Protes-
tantism on suicide is barely affected. 再次, the positive
effect of Protestantism is also evident when measuring sui-
cides per deaths rather than per inhabitants (柱子 6). 这
1816–1821 analyses thus conﬁrm a strong positive effect of
Protestantism on suicide also for the early nineteenth century
and show it for both genders.

G. A First-Difference Model

With data from two points in time—1816 to 1821 和
1869 to 1871—it is also possible to estimate a ﬁrst-difference
模型. By testing whether any change in the Protestant share
over time is associated with a contemporaneous change in
the suicide rate, such a model effectively removes county
ﬁxed effects, disregarding any differences in the levels of
suicides and Protestantism across counties and focusing only

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

THE EFFECT OF PROTESTANTISM ON SUICIDE

385

Table 4.—Protestantism and Suicide in Prussia, 1816: IV Estimates

Dependent Variable:

First Stage

Second Stage

Share of Protestants

Distance to Wittenberg (在 1,000 km)

Share of population < 15 years Share of population > 60 年

Share of population living in towns

Public buildings per capita

School enrollment rate

Fatal accident rate (每 100,000 inhabitants)

持续的

观察结果
R2
F-statistic (instrument)

Share of
Protestants

全部
(1)

−.869
(.195)∗∗∗

−4.251

(.943)∗∗∗

−8.089
(1.845)∗∗∗
-.078
(.088)
.091
(.056)
.443
(.104)∗∗∗

2.647
(.421)∗∗∗
306
.360

Suicide Rate
(每 100,000 inhabitants)

Suicide Proportion
(每 1,000 deaths)

全部
(2)

全部
(3)

Males
(4)

Females
(5)

14.989
(3.020)∗∗∗

14.971
(2.873)∗∗∗

23.439
(4.701)∗∗∗

7.066
(1.621)∗∗∗

15.253
(21.538)
1.281
(39.500)
6.570
(1.431)∗∗∗
−.262
(1.165)
−2.074
(1.998)

−7.369
(10.830)
306
.024
19.827

15.117
(20.667)
.555
(36.464)
6.566
(1.428)∗∗∗
−.256
(1.129)
−2.045
(1.881)
-.001
(.021)
−7.225
(10.029)
306
.026
24.861

19.218
(33.926)
4.341
(59.446)
11.932
(2.536)∗∗∗
−1.900
(1.706)
−2.941
(3.160)
.010
(.034)
−9.985
(16.502)
306
.023
24.861

11.345
(11.620)
−.576
(21.749)
1.922
(.900)∗∗
1.278
(1.033)
−1.174
(.963)
-.014
(.011)
−4.881
(5.700)
306
.078
24.861

全部
(6)

5.240
(1.059)∗∗∗

4.804
(7.797)
7.929
(13.593)
1.809
(.493)∗∗∗
.087
(.394)
−.886
(.676)
.0001
(.008)
−2.761
(3.755)
306
.092
24.861

Instrumental variables (IV) estimation. Heteroskedasticity-robust standard errors in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Data for Prussian counties from the 1816 Census. See main text and online

appendix C for details.

on changes over time. The identifying assumption of this
ﬁrst-difference model is quite different from our IV model
(which assumes that the concentric spread of the Protes-
tant Reformation from Wittenberg is not otherwise related
to suicide proneness). It assumes that in the absence of any
differential change in the Protestant share, there would have
been no systematic difference in the change in suicide rates.
Changes in denominational shares within counties during
this time are likely to mostly reﬂect migration patterns such
as movements of Polish Catholics to the coal mines of the
Ruhr area, as well as differential fertility and mortality, 所以
that the analysis cannot exclude selection bias. 仍然, 第一个-
difference model provides an alternative test of our main
hypothesis that Protestantism affected suicide rates.

We can perform the ﬁrst-difference analysis for 272 县-
领带 (出于 306 counties that existed in 1816) that can
be linked over time because there were no or only negli-
gible changes in their county boundaries between 1816 和
1871. The observed range of county-level changes in Protes-
tant shares is in fact quite substantial, ranging from a 25
percentage point decrease to a 22 percentage point increase
across counties. As is evident from the ﬁrst column of table 5,
both the initial share of Protestants and its change over time
are signiﬁcantly positively related to the change in the sui-
cide rate over time.9 The larger the increase in the Protestant
分享, the larger the increase in the suicide rate. To test for
robustness to contemporaneous changes in other potentially

confounding factors, 柱子 2 includes controls for contem-
poraneous changes in urbanization and the age structure, 作为
well as their initial levels. Column 3 additionally controls for
the initial level of the dependent variable, the suicide rate in
1816. The effect of changes in Protestant shares on changes
in suicide rates is highly robust in these speciﬁcations.

Despite the substantial range of observed changes in
Protestant shares over time, the majority of counties expe-
rience relatively small changes in denominational shares.
To ensure that these are not affecting the ﬁrst-difference
结果, columns 4 和 5 restrict the analysis to those 114
(50) counties whose Protestant shares change by at least 2 (5)
百分点, 分别. Results are robust in these
smaller samples. The ﬁrst-difference analysis thus indicates
a robust positive association between increased Protestant
shares and increased suicide rates over time.

IV. Discriminating between Sociological

and Theological Explanations

The evidence so far conﬁrms a causal effect of Protes-
tantism on suicide, but it does not discriminate between the
sociological and the theological explanations for this effect.
在这个部分, we devise a series of empirical tests that try
to provide such discrimination.

A. Church Attendance and the Relevance

of Social Cohesion and Religious Beliefs

9 Mechanically, the change in the Protestant share is negatively correlated

with its initial level.

A ﬁrst test to discriminate between the two types of expla-
nation builds on their differing predictions with respect to

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
/
r
e
s
t
/

我

A
r
t
我
C
e
–
p
d

F
/

1
0
0
3
3
7
7
1
9
7
4
8
1
3
/
r
e
s
t
_
A
_
0
0
7
0
8
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

386

THE REVIEW OF ECONOMICS AND STATISTICS

Dependent Variable:

Change in Suicide Rate (每 100,000 inhabitants), 1816–1871

Table 5.—First-Difference Model on Changes between 1816 和 1871

Excluding Counties Where

Change in share of Protestants (1816–1871)

Share of Protestants (1816)

Change in share of population < 15 years (1816–1871) Change in share of population > 60 年

(1816–1871)

Change in share of population living in towns

(1816–1871)

Share of population < 15 years (1816) Share of population > 60 年 (1816)

Share of population living in towns (1816)

Suicide rate per 100,000 inhabitants (1816)

持续的

观察结果
R2

All Panel Counties

(2)

(3)

(1)

.099
(.040)∗∗
.081
(.007)∗∗∗

.084
(.037)∗∗
.078
(.008)∗∗∗
-.048
(.163)
.869
(.288)∗∗∗
-.007
(.026)
−.269
(.191)
.864
(.327)∗∗∗
-.012
(.021)

.078
(.036)∗∗
.099
(.009)∗∗∗
−.144
(.164)
1.007
(.279)∗∗∗
.004
(.026)
−.450
(.197)∗∗
.695
(.336)∗∗
.005
(.020)
−.333
(.078)∗∗∗
13.039
(9.476)
272
.404

1.028
(.344)∗∗∗
272
.284

4.938
(9.150)
272
.365

Δ %Prot
< 2 p.p. (4) .091 (.040)∗∗ .116 (.016)∗∗∗ −.187 (.205) .761 (.307)∗∗ −.018 (.028) −.537 (.281)∗ .167 (.491) −.005 (.027) −.419 (.117)∗∗∗ 20.580 (13.606) 114 .464 Δ %Prot < 5 p.p. (4) .117 (.040)∗∗∗ .143 (.022)∗∗∗ −.329 (.239) 1.014 (.432)∗∗ −.013 (.029) −.531 (.301)∗ .422 (.669) −.018 (.033) −.604 (.163)∗∗∗ 18.748 (15.085) 50 .653 Ordinary least squares (OLS) regression on long difference in suicide rates. Heteroskedasticity-robust standard errors in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Data for Prussian counties from the 1816 Census, the 1869–1871 Suicide Statistics, and the 1871 Population Census. See main text and online appendix C for details. how the extent of church attendance in a community affects suicide rates and their dependence on Protestantism. From the sociological perspective, higher church attendance can be viewed as a sign of greater social cohesion. Thus, even if the integration into the Protestant community might be lower than in the Catholic community on average, in Protestant areas, Protestants should be relatively more closely embed- ded in their community when more people attend church. The sociological explanation would thus predict that higher church attendance should dampen the effect of Protestantism on suicide. By contrast, from the theological perspective, higher church attendance can be viewed as a sign of more devout church members. Higher church attendance would thus sig- nal stronger belief in Protestant doctrine, which should go along with an even stronger effect of Protestantism on sui- cide. There is ample evidence to support the assumption that theological beliefs are more strictly believed in areas with higher church attendance. For example, McCleary and Barro (2006) show that the correlations of monthly atten- dance at religious services with religious beliefs such as belief in afterlife, God, heaven, and hell, as well as with self-identiﬁcation as a religious person, are as high as 0.86 to 0.91 across countries (see their table 9). For the United States, Glaeser and Sacerdote (2008) show a strong sig- niﬁcant effect of religious beliefs (an index that combines belief in heaven, miracles, the devil, and the Bible as lit- eral truth) on monthly attendance at both the individual and the metropolitan-area levels (see their tables 1, 2, and 9). For Great Britain, Sawkins, Seaman, and Williams (1997) show that attendance at religious services is signiﬁcantly positively associated with the intensity of religious belief for both women and men (see their tables 1 and 2). These opposite predictions of the two explanations on how suicide rates change with church attendance rates provide us with a way to test the two channels against each other. To do so, we make use of the unique database of church atten- dance provided by the statistical surveys of the Protestant regional churches of Germany on the expressions of churchly life (see Hölscher, 2001, and Becker & Woessmann, 2013, for additional detail). First in 1862 and then more regularly on an annual basis starting in 1881, parish priests were to count the number of participations in Holy Communion on a preprinted form following uniform surveying directives. The data are available from regional archives at the level of church districts (Kirchenkreise), usually comprising 10 to 20 adjacent parishes. Our measure of church attendance is the number of participations in Holy Communion divided by the number of Protestants in a church district. To match our 1869–1871 suicide statistics, we take a simple average of church attendance in 1862 and 1881, the closest years with available data for most church districts.10 We map the church district data into our administrative county data by using 10 Results based on just the 1862 data are very similar. To ensure that the averaging is not affected by overall trends in church attendance, before taking the average of the two years, we ﬁrst regress the 1862 and 1881 data on each other and predict any missing value for a county in one year by the predicted value from these regressions. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 THE EFFECT OF PROTESTANTISM ON SUICIDE 387 Dependent Variable: Suicide Rate Table 6.—Discriminating between Sociological and Theological Explanations, 1871 Church attendance Share of Protestants Bottom quartile of church attendance Share Protestants × Bottom quartile of church attendance Top quartile of urbanization Share Protestants × Top quartile of urbanization Further controls (as in table 2) Observations Number of clusters R2 Counties with More Than 98% Protestants (1) −11.418 (3.542)∗∗∗ Yes 90 66 .442 All Counties (2) (3) (4) 10.934 (1.231)∗∗∗ −2.131 (1.409) 4.386 (1.903)∗∗ Yes 396 258 .642 11.224 (1.241)∗∗∗ −.952 (1.274) 3.269 (1.838)∗ Yes 396 258 .641 10.480 (1.382)∗∗∗ −1.948 (1.411) 3.891 (1.915)∗∗ −.558 (1.711) 2.691 (1.993) Yes 396 258 .647 Ordinary least squares (OLS) estimation. Heteroskedasticity-robust standard errors, clustered at the level of church districts, in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Sample of counties with church attendance data (396 counties). Data for Prussian counties from the 1869–1871 Suicide Statistics, the 1871 Population Census, the 1882 Occupation Census, and the 1862–1881 Church Attendance Data. See main text and online appendix C for details. GIS technology to compute the surface-weighted average of the available church district data for each county.11 In cases where more than one county falls within the same church dis- trict, we cluster our regression analyses at the church district level. We perform two types of analyses. First, we restrict the analyses to the ninety counties (with available church atten- dance data) that are virtually all Protestant (Protestant share larger than 98%) to test whether Protestant church atten- dance is signiﬁcantly related to suicide rates. As is evident from the ﬁrst column of table 6, suicide rates decline signif- icantly with higher church attendance. If church attendance increases from the 10th to the 90th percentile in this sample (0.402 to 0.706), suicide rates are 3.5 suicides per 100,000 inhabitants lower, equivalent to a ﬁfth of the sample mean. Declining suicides with increasing church attendance are in line with the sociological channel but not the theological channel. Second, in the full sample, we test whether the effect of Protestantism on suicide differs with church attendance rates. To allow for functional ﬂexibility, we interacted the Protestantism effect with indicator variables for four quar- tiles of Protestant church attendance. Given that effects for the upper three-quartiles do not signiﬁcantly differ from one another, the speciﬁcation in column 2 includes an interac- tion only with the bottom quartile of church attendance. As is evident, the effect of Protestantism on suicide is signiﬁ- cantly larger in counties where Protestant church attendance is low. At 15.3 compared to 10.9 suicides per 100,000 inhabitants, the difference is again substantial. With higher church attendance dampening rather than heightening the 11 We treat a county as missing data if church attendance information is missing for more than half of its surface, but results are robust in larger samples that include all counties with any church attendance information. effect of Protestantism on suicide, this ﬁnding again speaks in favor of the sociological channel and against the theo- logical channel. Note that given the lack of evidence of a causal effect of income on church attendance in nineteenth- century Prussia (Becker & Woessmann, 2013), church attendance is unlikely to capture other factors like economic development. An additional indication that church attendance is indeed related to higher social cohesion among Protestants comes from modern microdata from the German Socio-Economic Panel (SOEP), where we can observe frequency of church attendance and several measures of social cohesion at the individual level. The data show that Protestants who attend church regularly are more likely to regularly meet with family and relatives, meet with neighbors or friends, and vol- unteer in associations or social services than Protestants who do not attend church regularly.12 This pattern is in line with the interpretation that higher church attendance indicates greater social cohesion. B. Further Evidence on Channels from Historical Patterns It can also be argued that the sociological aspect of a less tightly integrated Protestant community is particularly relevant in the anonymous environment of urban areas. By contrast, rural communities may exhibit more social cohe- sion irrespective of denomination, thereby dampening the 12 For example, 83.1% of Protestants who attend church at least monthly mutually visit family members or relatives at least once per month, com- pared to 78.1% of Protestants who attend church less regularly. The equivalent comparisons are 83.2% versus 77.5% for mutual visits of neigh- bors, friends, or acquaintances and 36.7% versus 14.3% for volunteering in associations or social services. While the pattern is generally similar among Catholics, there is an interesting contrast in that churchgoing Catholics are less likely to meet friends on a weekly basis than non-churchgoing Catholics (42.6% versus 47.3%; Protestants: 49.8% versus 46.7%). l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 388 THE REVIEW OF ECONOMICS AND STATISTICS effect of Protestantism on suicide if the sociological channel is indeed at work. The theological channel does not predict the Protestantism effect to differ with urbanization. Allowing for a ﬂexible functional form, we interacted the Protestantism effect with indicator variables for four quar- tiles of the share of the county population living in towns. With the lower three quartiles not signiﬁcantly differing from one another, column 3 of table 6 includes only the inter- action with the top quartile of urbanization. The effect of Protestantism on suicide is indeed signiﬁcantly larger in urban areas, supporting the relevance of the sociological channel. However, as is evident from column 4, the inter- action of Protestantism with church attendance dominates its interaction with urbanity. Another way to gauge the relevance of the sociological community channel is to look at the suicide rates of Catholics by whether they live in a predominantly Catholic or Protes- tant area. While the results reported in the bottom panel of table A.7 in the online appendix do not provide evidence for systematic minority effects, suicide rates of Catholics are in fact substantially larger in areas where Protestants have a majority. They are below 5% in districts with less than 40% Protestants but above 14% in districts with more than 60% Protestants. This pattern may again cast doubt on the impor- tance of religious beliefs in generating the observed results and indicate instead the relevance of social structure, here in terms of social spillovers. C. Evidence from Modern Data A ﬁnal piece of evidence to differentiate between relevant channels builds on modern data. In modern times even more than before, Protestant doctrine is more accommodating with suicides. Leading Protestant theologians have turned against condemning suicide, arguing that an individuals can com- mit suicide in a state of peace with God. At the same time, the Catechism of the Catholic Church continues to be very explicit against suicide, noting in point 2325, “Suicide is seri- ously contrary to justice, hope, and charity. It is forbidden by the ﬁfth commandment.” Based on the theological chan- nel, one might thus expect the difference in suicide rates between true believers in Protestant and Catholic doctrine to sharpen in modern times, especially when large num- bers have left both churches, presumably leaving behind members more committed to the doctrine of their church. In contrast, if the sociological channel dominates, a smaller ﬂock of (ﬁrm) Protestant and Catholic believers should both ﬁnd consolation in their congregations. Data on suicides in modern Germany come from the mortality statistics accessible via controlled remote access (Todesursachenstatistik, EVAS 23211), covering all deaths at an individual level from 1992 to 2009. These statistics are based on the death certiﬁcate issued by the doctor declaring the death, in combination with the death registry certiﬁ- cate issued by the registrar of the municipality of residence. The death certiﬁcate contains information about diseases and signiﬁcant other health issues that have contributed to death. The classiﬁcation of causes of death follows the World Health Organization’s International Classiﬁcation of Dis- eases (ICD-9 until 1997 and ICD-10 since). In addition to the primary cause of death, the mortality statistics include demographic features such as gender, age, German citizen- ship, marital status, place of residence, date of death, and, importantly, religion. We use the years 1992 and 2009, the earliest and lat- est years available. In 1992, the mortality statistics report 885,374 deaths, out of which 13,459, or 1.5%, are suicides. In 2009, the number of deaths is 854,544, out of which 9,622, or 1.1%, are suicides, indicating a considerable decline in the suicide proportion over the seventeen years. Between 1992 and 2009, both the Catholic and Protes- tant churches lost many members due to secessions. But the share of Protestants leaving the church is nearly 50% higher. In 1991, Protestant churches had 29.2 million members and the Catholic Church 27.7 million (Eicken & Schmitz- Veltin, 2010). The number of members actively seceding from Protestant churches from 1992 to 2009 (not counting deaths or other movements) was 3.6 million: 12.2% of the initial stock left over the course of less than two decades. In the Catholic Church, the number of members seceding over the same period was 2.3 million, or 8.2%, of the initial stock.13 Consequently, former Protestants make up a larger share of the nonafﬁliated, an important factor when assessing suicide rates by denomination and nonafﬁliation.14 Table 7 reports OLS regressions of a suicide indicator on religious afﬁliation, controlling for basic demographic characteristics: a quadratic in age, gender, German citizen- ship, and marital status.15 In 1992, the regression reveals that suicide proportions are 0.18 percentage points higher for Protestants than for Catholics. This estimate is equivalent to 13.7% of the raw Catholic suicide proportion. The suicide proportion of citizens without religious afﬁliation is 0.44 percentage points higher than that of Catholics. It thus far exceeds that of any afﬁliated deceased, including Protestants. When county ﬁxed effects are included in the analysis, the estimates increase to 0.252 percentage points for Protestants and 0.597 percentage points for the nonreligiously afﬁli- ated.16 While the higher suicide proneness of Protestants 13 Data source: http://www.kirchenaustritt.de/statistik (accessed July 25, 2014). 14 The pattern of Protestant seceding from their church in higher numbers is consistent with the religious afﬁliation data in the mortality statistics. Among the people who died in 1992, 46.1% were Protestants, 33.9% were Catholics, and 16.0% were not religiously afﬁliated. By 2009, the Protestant share had decreased to 39.9%, the Catholic share had barely changed at 33.2%, and the nonreligiously afﬁliated had increased to 21.4%. 15 As not only the choice to secede from the church, but also strong postwar migration waves and increased regional mobility probably undermine the instrument characteristics of the historical spread of the Reformation for Protestantism in our modern data, the contemporary analysis stays purely descriptive. 16 Qualitative results are robust in probit and logit models, although the estimated marginal effects are somewhat smaller than the OLS estimates (e.g., 0.167 in the probit and 0.124 in the logit model rather than the 0.252 OLS estimate with county ﬁxed effects, not shown). l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 THE EFFECT OF PROTESTANTISM ON SUICIDE 389 Dependent Variable: Suicidea Table 7.—Protestantism and Suicide in Germany, 1992 and 2009 1992 2009 West Germany, 1992 (1) (2) .180 (.029)∗∗∗ .252 (.034)∗∗∗ (3) .027 (.026) .143 (.030)∗∗∗ .215 (.037)∗∗∗ .217 (.034)∗∗∗ (4) (5) (6) (7) (8) Protestant Share Protestants in county No religious afﬁliation Other religious afﬁliations Age Age squared Male German citizenship Family status controls County ﬁxed effects County-level controls Observations R2 .443 (.039)∗∗∗ Yes −.314 (.005)∗∗∗ .140 (.004)∗∗∗ .720 (.029)∗∗∗ .423 (.120)∗∗∗ Yes No No 885,374 .032 .597 (.049)∗∗∗ Yes −.312 (.005)∗∗∗ .138 (.004)∗∗∗ .715 (.029)∗∗∗ .417 (.122)∗∗∗ Yes Yes No 885,374 .033 .161 (.032)∗∗∗ Yes −.265 (.004)∗∗∗ .113 (.003)∗∗∗ .738 (.025)∗∗∗ .112 (.080) Yes No No 854,544 .029 .381 (.040)∗∗∗ Yes −.263 (.004)∗∗∗ .111 (.003)∗∗∗ .731 (.025)∗∗∗ .163 (.081)∗∗ Yes Yes No 854,544 .030 1.098 (.125)∗∗∗ Yes −.338 (.014)∗∗∗ .155 (.009)∗∗∗ .702 (.036)∗∗∗ .162 (.206) Yes Yes No 666,261 .038 1.065 (.121)∗∗∗ Yes −.338 (.014)∗∗∗ .155 (.009)∗∗∗ .702 (.036)∗∗∗ .155 (.206) Yes No Yes 666,261 .037 .213 (.037)∗∗∗ .046 (.130) 1.062 (.122)∗∗∗ Yes −.338 (.014)∗∗∗ .155 (.009)∗∗∗ .702 (.036)∗∗∗ .155 (.206) Yes No Yes 666,261 .037 .244 (.120)∗∗ .927 (.129)∗∗∗ Yes −.339 (.014)∗∗∗ .156 (.009)∗∗∗ .703 (.036)∗∗∗ .199 (.205) Yes No Yes 666,261 .037 Ordinary least squares (OLS) estimation. a Dependent variable multiplied by 100. Heteroskedasticity-robust standard errors in parentheses: Signiﬁcant at ∗10%, ∗∗5%, and ∗∗∗1%. Standard errors in speciﬁcations with county-level controls are clustered at the level of 327 counties. Other religious afﬁliations: Jewish; Muslim; other Christian; other religion; religion unknown (reference category: Catholic). Family status controls: single; widowed; divorced; unknown (reference category: married). County-level controls: log(county population); share of county population in twenty separate age groups; share of county population who are single, widowed, married, divorced; share of foreigners; share of workforce in four separate sectors; share of population receiving social beneﬁts; share of population receiving ﬁnancial support from relatives; share of population with ﬁve separate educational degrees. Data: Mortality Statistics (Todesursachenstatistik), 1992 and 2009; county-level variables: Population Census (Volkszählung), 1987. See main text and online appendix C for details. compared to Catholics is still visible in 1992, the fact that suicide proneness of those not afﬁliated with a religion— which are disproportionately formerly Protestant—is even higher raises doubts that Protestant religious beliefs are the predominant factor in the denominational pattern of suicides. In 2009, the estimate on Protestantism is reduced to an insigniﬁcant 0.03 percentage points, or just 2.8% of the raw Catholic suicide proportion. With county ﬁxed effects, the estimate regains statistical signiﬁcance, but at 0.14 per- centage points, it is 43% lower than the respective estimate in 1992. The suicide proportion of those without religious afﬁliation remains 0.38 percentage points higher than that of Catholics. Under the assumption that the most devout believers in the respective religious doctrines are more likely to stay in their church, the declining difference in suicides between Protestants and Catholics again speaks against a paramount role for the theological explanation. The continuing fact that religiously nonafﬁliated people, which disproportionately draw on people leaving Protes- tant churches, have the highest suicides also speaks for a dominant role of socialization rather than religious beliefs. The individual-level modern data also allow us to probe some more into the issue of ecological fallacy. Reli- gious afﬁliation at a regional level is available only in the most recent Population Census (Volkszählung) of 1987, available for West Germany only. As column 5 indicates, at 0.215, the point estimate on Protestantism in the 1992 speciﬁcation with county ﬁxed effects is slightly (although not statistically signiﬁcantly) smaller in West Germany than the equivalent estimate of 0.252 for the whole of Germany. The estimate barely changes when we replace the county ﬁxed effects by county-level controls from the 1987 Popu- lation Census (the shares of the county population in twenty separate age groups, in four family-status groups, in ﬁve education groups, receiving social beneﬁts, receiving ﬁnan- cial support from relatives, the share of foreigners, the share of the workforce in four sectors, and log county popula- tion). When we replace the indicator of individual Protestant afﬁliation with the share of Protestants among Protestants and Catholics in the county, the estimate is very similar at 0.244. But when we include individual and county-level Protestant afﬁliation together, only the individual Protestant afﬁliation remains signiﬁcant, with a point estimate that is barely affected, while the point estimate on the Protestant share in the county is reduced to virtually 0. This pattern suggests that the actual effect stems from the individual afﬁl- iation, but that estimates based on the county share provide unbiased estimates of the individual effect. That is, county- level estimates do not seem to suffer from ecological fallacy, which is reassuring for our historical analyses. Finally, modern data also allow us to test whether Protes- tantism indeed features a more individualistic and less community-oriented nature than Catholicism, as the basic hypothesis of the sociological channel suggests. For this, we use the individual-level data from the German SOEP in 2003. Catholics are in fact much more likely to go to church (see table A.10 in the online appendix). Among Catholics, 20.1% attend church at least weekly and 35.3% at least monthly, whereas these shares are only 4.9% and 16.3%, respectively, among Protestants. That is, Protestants are indeed substantially less likely to interact with their religious community on a regular basis than Catholics, which l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 390 THE REVIEW OF ECONOMICS AND STATISTICS may have detrimental consequences when they are in a suici- dal state of mind. The lack of cohesion within the religious community is not compensated by higher social cohesion outside the religious community. Indeed, Catholics are 2.8 percentage points more likely to visit family or relatives on a weekly basis and only slightly (1.6 percentage points) less likely to visit neighbors and friends weekly.17 As another indicator of social interaction beyond the religious commu- nity, Catholics are more than 3 percentage points more likely to go out in cafés or restaurants on a weekly or monthly basis than Protestants. Overall, these indicators suggest that the higher suicide proneness of Protestants is indeed related to a lower incidence of social cohesion, in particular within the religious community, as suggested by the sociological channel.18 Strikingly, religiously nonafﬁliated individuals show a lower incidence of community orientation than either Catholics or Protestants. While they obviously rarely attend religious events, they are also substantially less likely to reg- ularly visit family or visit friends than either Catholics or Protestants. Only with respect to going out for food or drink are they on a level roughly similar to Protestants. Again, this pattern is consistent with a sociological explanation for the greater incidence of suicides among those without a religious afﬁliation. V. Conclusion This paper studies the effect of Protestantism on suicide both theoretically and empirically. Theoretically, we model sociological and theological mechanisms through which Protestants are predicted to have higher suicide rates than Catholics. In the framework of an economic model of sui- cide, individuals who are in a suicidal mental state compare the expected utility from living with that from death. Accord- ing to the sociological aspect of religion based on denomi- national differences in group structure, as Durkheim (1897) argued, Protestant doctrine emphasizes religious individu- alism, whereas Catholics have a more integrated religious community. As a consequence, Protestants will have a lower utility from keeping on living and a lower cost of committing suicide relative to Catholics. To this sociological channel, we add two mechanisms based on denominational differ- ences in theological doctrine that derive from a consideration of afterlife in individuals’ utility maximization. Protestant doctrine tends to stress that humans cannot affect God’s decisions by their deeds but depend on God’s grace (sola gratia), whereas Catholic doctrine grants that human access to heaven is affected by individual deeds. For Catholics, committing the deadly sin of suicide reduces the probability 17 There is no noteworthy difference in volunteering in associations or social services between Protestants and Catholics in these data. 18 Relatedly, using data from 32 countries in the International Social Sur- vey Programme (ISSP) 1998–1999, Arrunada (2010) shows that Catholics have a more personalized and Protestants a more impersonal social ethic. For example, Catholics are signiﬁcantly more likely to cover up for friends and to value the importance of family but less likely to trust strangers. of reaching heaven, thereby lowering the optimality of the suicide threshold relative to Protestants. Furthermore, since Catholic doctrine views confession as a holy sacrament but Protestant doctrine does not, the impossibility of confessing the sin of suicide creates a substitution effect away from sui- cide to other possible actions considered by Catholics who are in a suicidal mental state, again reducing the optimal- ity of suicide relative to Protestants in that state. Thus, both sociological and theological differences between Protestants and Catholics make suicide more likely among Protestants. When testing the model prediction that Protestantism increases suicides, our empirical model places particular emphasis on excluding biases from self-selection of suicide- prone individuals into religious denominations and from other forms of endogeneity and unobserved heterogeneity. For this, we construct a unique database from suicide sta- tistics and censuses that cover all Prussian counties in the early and late nineteenth century. In this setting, we exploit the concentric dispersion of Protestantism in Prussia in an IV model that instruments the share of Protestants in a county by its distance to Wittenberg. We ﬁnd that Protestantism increases the average annual suicide rate over the 1869–1871 period by about ﬁfteen to twenty suicides per 100,000 inhab- itants, a large effect compared to the mean suicide rate of thirteen suicides per 100,000 inhabitants. The result is robust to a rich set of controls for demographic, economic, educa- tional, and geographic background factors. Controls for the share of insane people in the population and for fatal accident rates address concerns of bias from denominational differ- ences in nonrational suicide causes and in underreporting of suicides. Likewise, we exclude that the higher Protestant shares identiﬁed by our instrument are related to unpleas- ant weather conditions and that our results are driven by religious concentration or ecological fallacy. We ﬁnd a posi- tive effect of Protestantism on suicide also in the 1816–1821 period, where the effect is larger for men than for women. Results are further conﬁrmed in a ﬁrst-difference model of changes between 1816 and 1871. Finally, we devise several tests to differentiate between the sociological and the theological channels. Most impor- tant, the effect of Protestantism on suicide tends to decrease rather than increase with church attendance, suggesting that the sociological role of a more integrated community dominates the theological aspect of a stronger devotion to religious beliefs. In addition, the difference in suicides between Protestants and Catholics recedes in modern data, whereas nonafﬁliated people (predominantly former Protes- tants) have signiﬁcantly more suicides, again suggesting a dominant role for socialization rather than theological belief. Of course, both sociological and theological mechanisms can be working at the same time, and conﬁrmative evidence for one channel does not rule out that the other channel also plays a role. The modern results also suggest that afﬁliation with a Christian church in general may hedge against suicide risk, possibly because of a higher degree of social integration in l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 THE EFFECT OF PROTESTANTISM ON SUICIDE 391 Christian communities in general compared to people not afﬁliated with any church (but possibly also because athe- ists do not belief in punishment with loss of afterlife utility, as depicted in our model). In this sense, our main results may also be interpreted as a positive effect of Catholicism on reduced suicide, both relative to Protestantism but in par- ticular relative to individuals with no religious afﬁliation, who may suffer from particularly low levels of perceived social cohesion. This aspect of the modern data opens an interesting direction for future research. In terms of the effect of Protestantism on overall well- being, our result that Protestantism increases suicide rates relative to Catholics contrasts with the ﬁnding that Protes- tantism furthers educational and economic development (Becker & Woessmann, 2009). Thus, the effect of Protes- tantism on well-being seems to be neither uniformly positive nor uniformly negative and may affect the average popu- lation differently from the very select subgroup of people who are in a suicidal state of mind. In fact, the two aspects may be related in a dark-contrasts paradox where suicide behavior is subject to a relative comparison of utility (Daly et al., 2011). Still, our results hold conditional on proxies for economic development and inequality, suggesting that religious denomination in the form of Protestantism is a main independent driver of regional differences in suicide rates. REFERENCES Arrunada, Benito, “Protestants and Catholics: Similar Work Ethic, Differ- ent Social Ethic,” Economic Journal 120:547 (2010), 890–918. Bankston, William B., H. David Allen, and Daniel S. Cunningham, “Reli- gion and Suicide: A Research Note on Sociology’s ‘One Law,’” Social Forces 62:2 (1983), 521–528. Barrett, David B., George T. Kurian, and Todd M. Johnson, World Christian Encyclopedia, 2nd ed. (Oxford: Oxford University Press, 2001). Becker, Gary S., and Richard A. Posner, “Suicide: An Economic Approach,” University of Chicago mimeograph (2004). Becker, Sascha O., Francesco Cinnirella, Erik Hornung, and Ludger Woess- mann, “iPEHD—the Ifo Prussian Economic History Database,” Historical Methods 47:2 (2014), 57–66. Becker, Sascha O., and Ludger Woessmann, “Was Weber Wrong? A Human Capital Theory of Protestant Economic History,” Quarterly Journal of Economics 124:2 (2009), 531–596. ——— “Not the Opium of the People: Income and Secularization in a Panel of Prussian Counties,” American Economic Review, Papers and Proceedings 103:3 (2013), 539–544. Chen, Joe, Yun Jeong Choi, Kohta Mori, Yasuyuki Sawada, and Saki Sug- ano, “Socio-Economic Studies on Suicide: A Survey,” Journal of Economic Surveys 26:2 (2012), 271–306. Cutler, David M., Edward L. Glaeser, and Karen E. Norberg, “Explaining the Rise in Youth Suicide,” in Jonathan Gruber, ed., Risky Behavior among Youths (Chicago: University of Chicago Press, 2001). Daly, Mary C., Andrew J. Oswald, Daniel J. Wilson, and Stephen Wu, “Dark Contrasts: The Paradox of High Rates of Suicide in Happy Places,” Journal of Economic Behavior and Organization 80:3 (2011), 435–442. Daly, Mary C., and Daniel J. Wilson, “Happiness, Unhappiness, and Sui- cide: An Empirical Assessment,” Journal of the European Economic Association 7:2–3 (2009), 539–549. Daly, Mary C., Daniel J. Wilson, and Norman J. Johnson, “Relative Status and Well-Being: Evidence from U.S. Suicide Deaths,” this review 95:5 (2013), 1480–1500. Durkheim, Émile, Le suicide: étude de sociologie (Paris: Félix Alcan, 1897). (Suicide: A Study in Sociology, trans. John A. Spaulding and George Simpson. Glencoe, The Free Press, 1951.) Eicken, Joachim, and Ansgar Schmitz-Veltin, “Die Entwicklung der Kirchenmitglieder in Deutschland: Statistische Anmerkungen zu Umfang und Ursachen des Mitgliederrückgangs in den beiden christlichen Volkskirchen,” Wirtschaft und Statistik 6 (2010), 576– 589. Glaeser, Edward L., and Bruce I. Sacerdote, “Education and Religion,” Journal of Human Capital 2:2 (2008), 188–215. Guiso, Luigi, Paola Sapienza, and Luigi Zingales, “Does Culture Affect Economic Outcomes?” Journal of Economic Perspectives 20:2 (2006), 23–48. Hamermesh, Daniel S., and Neal M. Soss, “An Economic Theory of Suicide,” Journal of Political Economy 82:1 (1974), 83–98. Helliwell, John F., “Well-Being and Social Capital: Does Suicide Pose a Puzzle?” Social Indicators Research 81:3 (2007), 455–496. Hilse, Carl, “Beiträge zur Kenntnis der Bewegung der Bevölkerung inner- halb der evangelischen und der römisch-katholischen Landeskirche des preussischen Staats in den Jahren 1859 bis 1867,” Zeitschrift des Königlich Preussischen Statistischen Bureaus 9 (1869), 305–318. ——— “Die Selbstmorde in Preussen im IV. Quartal 1868 und im Jahre 1869,” Zeitschrift des Königlich Preussischen Statistischen Bureaus 11:1–2 (1871), 41–120. Hölscher, Lucian, Datenatlas zur religiösen Geographie im protest- antischen Deutschland: Von der Mitte des 19. Jahrhunderts bis zum Zweiten Weltkrieg, 4 vols. (Berlin: Walter de Gruyter, 2001). Huang, Wei-Chiao, “Religion, Culture, Economic and Sociological Cor- relates of Suicide Rates: A Cross-National Analysis,” Applied Economics Letters 3:12 (1996), 779–782. Iannaccone, Laurence R., “Introduction to the Economics of Religion,” Journal of Economic Literature 36:3 (1998), 1465–1496. Iyer, Sriya, “The New Economics of Religion,” Journal of Economic Literature 54:2 (2016), 395–441. Kollarits, Jenö, “Ein Erklärungsversuch für die Selbstmordhäuﬁgkeit der Protestanten,” Zeitschrift für die gesamte Neurologie und Psychiatrie 49 (1919), 357–372. La Vecchia, C., F. Lucchini, and F. Levi, “Worldwide Trends in Sui- cide Mortality, 1955–1989,” Acta Psychiatrica Scandinavica 90:1 (1994), 53–64. Layard, Richard, Happiness: Lessons from a New Science (New York: Penguin Press, 2005). McCleary, Rachel M., and Robert J. Barro, “Religion and Political Econ- omy in an International Panel,” Journal for the Scientiﬁc Study of Religion 45:2 (2006), 149–175. Mützell, Alexander A., Neues Topographisch-statistisch-geographisches Wörterbuch des Preussischen Staats (Halle: Karl August Kümmel, 1825). OECD, Society at a Glance 2009: OECD Social Indicators (Paris: Organ- isation for Economic Co-operation and Development, 2009). Oswald, Andrew J., “Happiness and Economic Performance,” Economic Journal 107:445 (1997), 1815–1831. Pope, Whitney, and Nick Danigelis, “Sociology’s ‘One Law,”’ Social Forces 60:2 (1981), 495–516. Sawkins, John W., Paul T. Seaman, and Hector C. S. Williams, “Church Attendance in Great Britain: An Ordered Logit Approach,” Applied Economics 29:2 (1997), 125–134. Simpson, Miles, “Suicide and Religion: Did Durkheim Commit the Eco- logical Fallacy, or Did Van Poppel and Day Combine Apples and Oranges?” American Sociological Review 63:6 (1998), 895–896. Stark, Rodney, Daniel P. Doyle, and Jesse Lynn Rushing, “Beyond Durkheim: Religion and Suicide,” Journal for the Scientiﬁc Study of Religion 22:2 (1983), 120–131. van Poppel, Frans, and Lincoln H. Day, “A Test of Durkheim’s Theory of Suicide—Without Committing the ‘Ecological Fallacy,”’ American Sociological Review 61:3 (1996), 500–507. Wilke, Jürgen, “From Parish Register to the ‘Historical Table’: The Prussian Population Statistics in the 17th and 18th Centuries,” History of the Family 9:1 (2004), 63–79. World Health Organization, Preventing Suicide: A Global Imperative (Geneva: World Health Organization, 2014). l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / r e s t / l a r t i c e - p d f / / / / 1 0 0 3 3 7 7 1 9 7 4 8 1 3 / r e s t _ a _ 0 0 7 0 8 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 The Review of Economics and Statistics image

下载pdf