Homicide Declines, 600-2060 AD:
A Generalising Framework
Manuel Eisner
Institute of Criminology
University of Cambridge
16 Theories of the Crime Decline
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Incarceration
Better policing (Sherman, Zimring)
Demographic trends (Blumstein)
Abortion
Lead/environmental neurotoxins (Nevin)
Unemployment, consumer sentiment,
Inflation (Rosenfeld)
Self-control (Eisner/Pinker)
Psycho-pharmaceuticals (Finkelhor)
End of crack/cocaine epidemic
State legitimacy (Roth/LaFree)
Divorce – family stability
More effective public health/crime prevention policy (Finkelhor)
Changing norms and values
Dissipation of effects of cultural revolution (Pinker)
Technical Crime Control (Clarke et al.)
Methodological Approaches
1. National Case Studies
– One period, one country:
– Typical answers: Compstat; drugs/crack epidemic, abortion,
incapacitation
2. International Trend Analyses
– Several places, on period
– Typical answers: anomie (Messner/Rosenfeld), loss of legitimacy
(LaFree), Postmodernity (Young).
3. Historical Cross-cultural Generalising Perspective
– All violence declines, at all times, in all places.
Is there a limited number of universal mechanisms behind
all major declines of interpersonal violence in all societies
at all times?
i.e.
How do societies become less conflict-ridden and homicidal
places?
A Homicide Scale
Violence as Politics
Murders per 100,000 pop
1000
> 200 per 100,000
Iraq after Invasion
South African Townships
Ciudad Juarez, Mexico
•
•
•
•
•
Unrelated men in public space
Organized and functional
Private justice & revenge
Violent enterpreneurs
Fights over goods and territories
> 100
Civil War
100
10-100 Violent
Societies
10
1-10
Semi-Pacified
Societies
1
< 0.3 per 100,000
Japan in 1970-90s
England, 1920s
Denmark, 1950s
•
•
•
•
Large % domestic, female vict.
Disorganized and disfunctional
Individual pathologies
Marginal groups
< 1
0.1
Violence as Pathology
Pacified
Societies
The Evolutionary Perspective
Buss, D. M. and T. K. Shackelford
(1997). "Human aggression in
evolutionary psychological
perspective." Clinical Psychology
Review 17: 605-619.
What purposes does violence amongst men serve?
-> adaptionist, strategic, functional view informed by evolutionary theory
(Archer, Pellegrini, Pinker, Buss/Shackelford, Tremblay)
• Proactive instrumental violence
Violence as a means of gaining access to scarce resources, particularly
goods, power/status, sex
– Examples: raids, robberies, private taxation, piracy, staged fights, attacks on
same-sex competitors.
• Reactive retaliatory violence
Violence as a means of defending against aggressors, producing
protection, maintaining norms of reciprocity
– Examples: vendetta, revenge killing, vigilantism.
Under what circumstances does killing (a group member)
becomes a strategically less attractive option?
Some hypotheses:
– A high availability of law reduces the chances that people perceive the
need to use violence as a means of self-help (retaliation and revenge)
– A high legitimacy of social order increases the chances that people
believe they can fulfil desires in cooperation with others
– A high level of protection from risks increases the chances of investing
into building up durable relationships
– A cultivation of self-restraint and manners reduces the chances for
friction resulting in violence
Decline 1
Elite Violence, 600-1800 AD
Eisner, Manuel (2011). The Decline of Regicide in an
Evolutionary Perspective, British Journal of Criminology, 2011.
Why?
• The involvement of elites in intra-societal violence is not a
constant.
• Theorists such as Elias, Cooney, Black expect a decline in
elite violence.
What?
• All monarchs of all major dynasties across all of Europe from
600 to 1800 AD (N = 1820).
• Coded for cause of death and further particulars.
1750-1799
1700-1749
1650-1699
1600-1649
1550-1599
1500-1549
1450-1499
1400-1449
1350-1399
1300-1349
1250-1299
1200-1249
1150-1199
1100-1149
1050-1099
1000-1049
950-999
900-949
850-899
800-849
750-749
700-749
650-699
600-649
Rate per 100,000
Trend in Regicide Rates, 600-1800 AD
50-Year Periods
3000
Note:
2500
2000
Average Regicide Rate
1060 per 100,000
(ruler-years)
1500
General Population Today
1 per 100,000
(person years)
1000
R² = 0.4598
500
0
Regicide Rates by Region, 600-1800 AD
4500
East:
Byzantium,
Ottoman Empire,
Russia, Bulgaria
4000
Rate per 100,000 Monarch Years
3500
3000
2500
All
North
Scandinavian
Monarchies, England,
Scotland
East
2000
North
West&South
1500
1000
500
0
7th
8th
9th 10th 11th 12th 13th 14th 15th 16th 17th 18th
West&South
All others
Why did Regicide Decline?
• Fewer internal rivals with sufficient military power to challenge royal
authority.
• Decreasing probability of successful acquisition of power through
usurpation/murder.
• Power transition becomes more law-bound (e.g. inheritance laws).
• Change in elites’ conduct of life (“from warriors to courtiers”).
• Higher stability and legitimacy of state institutions.
Decline 2
Homicide in Europe, 1300-2000
• Eisner, M. (2003) Long-Term Historical Trends in Violent Crime, Crime
and Justice.
• Johnson, E and E. Monkkonen (eds.) (1996). The Civilization of Crime;
Violence in Town and Country since the Middle Ages. Urbana: Illinois
University Press.
• Spierenburg, P. (2008) A History of Murder. Cambridge: Polity Press.
Homicide Rates across Europe, 1300-2000
80
Homicide Rate per 100,000
70
60
50
40
30
20
10
England
Netherlands
Scandinavia
Germany
Italy
0
1300 1400 1500 1600 1700 1800 1900 2000
See Eisner (2003). Long-Term Historical Trends in Violent Crime, Crime and Justice
Why did Homicide Decline?
• Increasingly stable state monopoly of power
– Elias (1978 [1939]) The civilizing process.
• Growing protection decreases necessity of violence as self-help
– Black (1983). Crime as Social Control.
• Interdependence and social complexity increase benefits of self-control
– Elias (1978 [1939]) The civilizing process.
• Disciplining institutions promote impulse control
– Krieken (1989). Violence, Self-Discipline and Modernity: Beyond the Civilizing Process
• Growing legitimacy and integration of state
Regicide and Homicide
250
Rate per 100000
200
England
150
Netherlands
Scandinavia
Germany
100
Italy
Regicide
50
0
600
800
1000
1200
1400
1600
1800
2000
Any way to test a theory empirically?
How about this hypothesis:
The decline was a result of regular frontal-lobe training in selfcontrol by means of reading books?
Book Production in Europe
600
Books per 1000 population
500
Great Britain
400
Ireland
Belgium
Netherlands
300
Germany
Switzerland
Italy
200
Spain
Sweden
100
0
1454 - 1500 1501 - 1550 1551 - 1600 1601 - 1650 1651 - 1700 1701 - 1750 1751 - 1800
Source: Baten, J., & Van Zanden, J. L. (2008). Book production and the onset of modern economic growth.
Journal of Economic Growth, 13(3), 217-235.
The Books-Murder Connection, 1500-1800
45
Homicides per 100,000 population
40
35
30
25
20
Murder
15
10
5
0
0
100
200
300
400
500
600
Books per 1000 Population
Note: All available pairs of 50-year estimates of book production and homicide rates, 1500-1800
N = 46.
… this looks like an elasticity
The Books-Murder Connection, 1500-1800
Homicides per 100,000 population
100
y = 28.804x-0.363
R² = 0.4464
10
Murder
Power (Murder)
1
0.1
1
10
100
1000
Books per 1000 Population
Note: All available pairs of 50-year estimates of book production and homicide rates, 1500-1800
N = 46.
Decline 3
Homicide in the USA, 1600-1800
Roth, Roland (2009). American Homicide. Cambridge (Mass):
Belknap.
Empirical Observation:
Parallel to the development in Europe, Homicide rates among unrelated
adults in New England, New Netherlands, Virginia and Maryland fell
from over 100 per 100,000 to about 1-5 per 100,000 between the early
1600s and around 1790.
The Core Roth Hypothesis
• Both in the USA and in Europe there are four correlates of
declining homicide amongst unrelated male adults.
1.
2.
3.
4.
The belief that government is stable and that its legal and judicial
institutions are unbiased and will redress wrongs and protect lives
and property.
A feeling of trust in government and the officials who run it, and a
belief in their legitimacy.
Patriotism, empathy, and fellow feeling arising from racial, religious,
or political solidarity.
The belief that the social hierarchy is legitimate, that one’s position
in society is satisfactory and that one can command the respect of
others without resorting to violence.
Decline 4
The Victorian Success Story
• National series available in many countries
• Victorian decline of homicide much studied by early
generation of criminologists (Boschi, Durkheim, Ferri).
The bad drunk husband
The Victorian ideal of the
family
Homicide Trends, 1840-2005
12
10
Austria
Spain
Italy
8
Switzerland
Belgium
France
6
Germany
Netherlands
Denmark
Norway
4
Sweden
Ireland
England and Wales
2
Scotland
0
1840
1860
1880
1900
1920
1940
1960
1980
2000
Explaining the Victorian Success Story
• Cultivating character
– Wiener, Martin (2004,) Men of Blood. Cambridge: Cambridge University
Press
• Social/national/political Integration
• Rise of control bureaucracies
– Emsley, Clive (2007) Crime, Police and Penal Policy: European
Experiences, 1750-1940.
• Temperance, self-control movements
– Carter-Wood, J. (2004), Shadow of our Refinement.
• Elimination of collectivist honour (i.e. self-help) cultures.
– Durkheim, Emile (1957 [1899] Professional Ethics and Civic Morals
– Gallant,
Decline 5
The Homicide Decline Since the Early 1990s
• Are some of the mechanisms associated with declining
homicide also present during the past 20 years?
– Is it broad rather than a locally limited phenomenon?
– Is it significant?
– Does is share characteristics of larger and longer declines?
• Homicide data for 22 Western countries from 1950 to 2008,
– Based on WHO and LaFree Dataset
Mean (1950-2006) = 100 for each series;
Scoret = homicide ratet*100/mean
250
Austria
Belgium
Denmark
Finland
France
Germany
Hungary
Ireland
Netherlands
Norway
Poland
Portugal
Spain
New Zealand
USA
Canada
Australia
Sweden
Italy
Switzerland
England and Wales
Scotland
Mean-1SDEV
Mean
Mean+1SDEV
200
150
100
50
0
1940
1950
1960
1970
1980
1990
2000
2010
2020
Mean (1950-2006) = 100 for each series;
Scoret = homicide ratet*100/mean
250
Austria
Belgium
Denmark
Finland
France
Germany
Hungary
Ireland
Netherlands
Norway
Poland
Portugal
Spain
New Zealand
USA
Canada
Australia
Sweden
Italy
Switzerland
England and Wales
Scotland
Mean-1SDEV
Mean
Mean+1SDEV
200
150
100
50
0
1940
1950
1960
1970
1980
1990
2000
2010
2020
Mean (1950-2006) = 100 for each series;
Scoret = homicide ratet*100/mean
250
Austria
Belgium
Denmark
Finland
France
Germany
Hungary
Ireland
Netherlands
Norway
Poland
Portugal
Spain
New Zealand
USA
Canada
Australia
Sweden
Italy
Switzerland
England and Wales
Scotland
Mean-1SDEV
Mean
Mean+1SDEV
200
150
100
50
0
1940
1950
1960
1970
1980
1990
2000
2010
2020
Mean (1950-2006) = 100 for each series;
Scoret = homicide ratet*100/mean
250
Austria
Belgium
Denmark
Finland
France
Germany
Hungary
Ireland
Netherlands
Norway
Poland
Portugal
Spain
New Zealand
USA
Canada
Australia
Sweden
Italy
Switzerland
England and Wales
Scotland
Mean-1SDEV
Mean
Mean+1SDEV
200
150
100
50
0
1940
1950
1960
1970
1980
1990
2000
2010
2020
Mean (1950-2006) = 100 for each series;
Scoret = homicide ratet*100/mean
250
Austria
Belgium
Denmark
Finland
200
France
Germany
Hungary
Ireland
Netherlands
150
Norway
Poland
Portugal
Spain
100
New Zealand
USA
Canada
Australia
Sweden
50
Italy
Switzerland
England and Wales
Scotland
0
1940
1950
1960
1970
1980
1990
2000
2010
2020
Timing of Peaks and Troughs , 1950-2008
Austria
Belgium
Denmark
Finland
France
Hungary
Netherlands
Norway
Poland
Spain
New Zealand
USA
Australia
Sweden
Italy
Switzerland
England and Wales
Trough
1963
1963
1960
1965
Peak
1983
1999
1992
1993
Trough
2008
2007
2005
2007
Trough
0.94
0.46
0.46
1.95
Peak
1.71
1.31
1.31
3.22
Trough
0.58
0.77
0.77
2.10
%
Increase T-P
83
186
186
65
%
Decrease P-T
-66
-41
-41
-35
1952
1964
1955
1954
1965
1963
1957
1957
1950
1961
1969
1969
1959
1984
1992
1996
1987
1992
1984
1991
1992
1989
1990
1990
1992
2002
2007
2006
2008
2008
2008
2008
2001
2008
2006
2007
2007
2007
2009
0.52
1.67
0.28
0.32
0.91
0.10
0.55
4.70
1.00
0.58
1.03
0.68
0.59
1.29
3.95
1.27
1.41
3.04
1.12
2.07
10.20
2.07
1.42
2.47
1.47
1.75
0.66
1.81
0.87
0.73
1.32
0.75
1.42
5.33
1.27
0.93
0.89
0.70
1.38
146
137
348
335
234
1055
276
117
107
145
139
117
199
-49
-54
-31
-48
-57
-33
-31
-48
-39
-34
-64
-53
-22
The Homicide Decline Since the Early 1990s
• Similar periodicity across Western countries
Mean Trough: 1959
Mean Peak:
1992
Mean Trough: 2007
• TSFA (time-series factor analysis) from 1950 to 2006:
– About 50 % of the variation across all countries shared by one
underlying factor.
• Similarity holds across countries with very different national
trajectories in
–
–
–
–
–
Policing
History of substance abuse
Unemployment
Imprisonment
Left/right political tradition
Decline 6 Looking into the future
•
•
•
Understanding = controlling
mechanisms = predicting
outcomes.
We know that liberal democratic
capitalist countries have
homicide rates of about .3-.4 per
100,000.
How do we get there in the next
30 years?
USA
14.00
12.00
10.00
8.00
USA
6.00
4.00
2.00
0.00
1940
1960
1980
2000
2020
2040
2060
To conclude: some problems
• How many cases are we looking at?
– 22 cases with 22 explanations
or
- 1 trajectory with 22 (or more) variations
• At what level of disaggregation do we need to look for
explanatory variables?
• What explains the commonalities in
– Timing of trough, peak, and decline?
• What are those commonalities and are they similar to previous
declines?
– Increasing political stability (the ‘victory’ of the capitalist/democratic
model of society)
– End of ‘self-fulfilment’, laissez faire parenting ideals
Some final thoughts
• Necessity of theory
– A theory that doesn’t contradict consolidated developmental,
situational, biological, etc. knowledge is not a disadvantage.
• Contextual Meaning of Predictive Indicators
– All macro-level indicators have multiple meanings. Moreover their
meaning can change over time and between contexts.
• Limitations of Quantitative Models
– In most cases the conditions for rigorous causal tests (e.g. TS
analyses of Granger causality) are not given
•
•
•
•
Lack of precision of outcome and predictor measures, lots of noise
Multiple meanings of indicators
Unknown ‘transfer function’ between cause and effect.
No data to test micro-level ‘thread needle’ of macro-level hypotheses