Appendix and Supplementary Materials:
These materials are organized into several parts. In the first part, we discuss the way in which the
dataset is constructed and how we measure each of the different concepts in our paper. In Part
Two, we include two tables in which we specify the model reported in our paper using two
alternative modeling methods: Multi-Level Modeling and Fixed Effects Regression. We report
these models by region and with all regions included. In Part Three, we furnish a series of tables
in which we include one additional control variable at a time in our model, using the ARFIMA
Technique. In Part Four, we display a correlation matrix between the variables that were utilized
in our model and the additional control variables that we did not report in our manuscript. In Part
Five, we run our models with both country and year binaries at the same time. This is the same
model that is reported in the paper, just with country binaries in addition to year-level binaries. In
Part Six, we run our models using the ARFIMA-MLM, Multi-Level Modeling, and Fixed Effects
Regression techniques with temperature as our only independent variable. In Part Seven, we
provide summary statistics for all of the variables used in our analysis.
Part One-Dataset Construction and Measurement:
Below readers may examine how we constructed our data set comprised of Countries' homicide
rates, temperature readings and various control factors. We discuss the key aspects and sources
of each type of data included in our final data set, which is available here (link).
Construction of Variables:
I. Dependent variable: Homicide vars. "homicide", "homiciderate", and "loghom"
Homicide data between 1995 and 2012 were collected from the UN as they are slightly more
complete for non-western nations than WHO data. Although one might go back further than
1995 data quickly become imbalanced before that era (leading to an overrepresentation of
industrialized nations). What is more earlier waves of UN data exist they do not appear fully
compatible as visual breaks in the data (large spikes or declines) were present.
The original data may be found here:
Set one is composed of most recent data (2000-2012), and can be found
athttp://www.unodc.org/gsh/en/data.html .
The second and third dataset were drawn from UN's 6th and 7th wave of law enforcement and
can both be found here:
http://www.unodc.org/pdf/crime/seventh_survey/7th%20all%20040331.xls (1998-2000)
http://www.unodc.org/pdf/crime/sixthsurvey/cs_2001_06_27.xls (1995-1997)
Website accessed 10-6-2014
Since collection of this data a new web interface has been posted at which all data from this time
period can be found (http://data.un.org/Data.aspx?d=UNODC&f=tableCode%3a1)
Homicide statistics from the UN are a blend of criminal justice (typically law enforcement) and
public health data sources, with CJ sources dominant. Although many nations had at least some
data available in our time period (1995-2012) we used the following selection criteria for
inclusion:
(1) Uninterrupted series of at least 8 years of data.
(2) A minimum of about 30 homicides per year, nations with 0 homicides in any given year were
excluded.
(3) Homicides should not display unreasonably large breaks between years (evidence of
reporting/definitional change). This is arguably a somewhat subjective rule, but it was used to
ensure that data were reported in a stable fashion.
(4) Avoiding nations with known political bias in reporting. Nations such as China, who -for at
least part of the time period under investigation- are known for underreporting or otherwise
distorting the true nature of crime were excluded.
Raw homicide data (var. "homicide") were combined with population figures (var "totalpop")
from the World Bank data (described below) to calculate rates per 100,000 (var. "homiciderate").
The natural log of var. homiciderate ("lnhomiciderate") was used in the analysis
II. Independent variable: Temperature, var. "tempcelcius"
The approach to collect temperature data follows that by Rotton and Cohn (2003) and Anderson
and DeLisi (2011) who selected the largest cities of each US state to develop a representative
sample for their study on US crime rates and climate change. In our case we select the largest
city (in some cases, cities) in a nation as homicides tend to cluster in large metropolitan areas. In
some cases we select multiple cities and combined the temperature readings based on population
weight for those cities (India and the US). The rationale behind selecting one city or a couple of
places rather than use the temperature average for the entire nation rest on the realization that
violence is highly concentrated in "socially disorganized communities" (Mares 2013b), most
typically larger urban places. In order to establish the most reasonable relationship between
climate and violence that relation has to be characterized as geographically accurate as possible.
Temperature data were drawn from NOAA's Global Historical Climatology Network (GHCN).
Monthly Mean Temperature data were aggregated to create average annual temperatures. Any
missing values were interpolated using linear interpolation.
Original temperature data may be found here:
http://gis.ncdc.noaa.gov/map/viewer/#app=cdo&cfg=cdo&theme=monthly&layers=1
Accessed 10-6-2014
In particular, our way of identifying places was by examining the sizes of the largest cities in a
nation. If a country has a dominant city (more than 2x population of next largest city) that one
city would suffice, unless the nation has a very large total population (100 million plus), or
extensive landmass with much climatic variation. In those cases we would seek out additional
large places and weigh temperature data according to population sizes. In some cases
measurement stations were not in or near our cities. Our preferences obviously was to keep
stations located within a few miles of the cities, in a few cases we had to use locations over 60
miles from the city in question, but we this would only be used if climatic conditions would be
reasonably identical (such as locations on the coast, for instance). In some cases nations had to
be dropped because no reasonable nearby locations could be identified, or too many missing data
were present in the data.
Nations for which approach was followed and city/cities for which temperature data were
collected (not all ended up being included as temperature data may have been incomplete, other
data may have been missing necessitating deletion of nation):
Kenya: Nairobi and Mombassa
Uganda: Kampala
Algeria: Algiers
Egypt: Elat, Israel
Morocco: Casablanca
South Africa: Cape Town, Johannesburg, Durban
Sierra Leone: Freetown
Bahamas: Nassau
Barbados: Barbados
Jamaica: Montego Bay
Belize: Felipe Carrillo Pue, Mexico
Costa Rica: Puerto Limon
El Salvador: San Salvador
Honduras: Teguciagalpa
Mexico: Mexico City and Monterrey
Argentina: Mercedes, Uruguay; Cordoba and Rosario
Bolivia: Tarija
Chile: Curico (located between Santiago, Conception and Valpraiso)
Colombia: Bogotá and Cali (Medellin not available)
Ecuador: Pichilingue (between Guyaquil and Quito, only available)
Guyana: Tumeremo, Venezuela
Paraguay: Concepcuion (few in centre)
Peru: Arequipa. Site reflects climatic conditions of Lima (near ocean), but slightly more south.
Uruguay: Rocha, (nearest to Montevideo).
Venezuela: Maracay B.A. Sucr. (Between Caracas and Valencia)
Kazakhstan: Almaty
Tajikistan: Dushanbe
Hong Kong: Shanwei, China
Japan: Tokyo and Nara (near Kyoto, Kobe, Osaka)
Mongolia: Bulgan (closest to Ulan Batar)
Philippines: Manilla
Singapore: Sitiawan, Malaysia
Thailand: Bangkok
Bangladesh: Agartala, India (east of Dhaka)
India: New Delhi, Mumbai, Calcutta
Pakistan: Bhuj Rudramata, India (near Karachi): Lahore
Sri Lanka: Hambantota (Southeast of Colombo)
Armenia: Amasia (West of Yerevan)
Azerbaijan: Alat (Southwest of Baku)
Georgia: Bolnisi (Southwest of Tbilsi)
Belarus: Mariyna Gorka (Southeast of Minsk)
Czech Republic: Milesovka (Northeast of Prague)
Hungary: Hurbanovo (West of Budapest)
Poland: Siedlce (East of Warsaw -only available)
Moldova: Kisinev
Romania: Varfu Omul (Northwest of Bucharest)
Denmark: Copenhagen
Estonia: Tallinn
Finland: Helsinki, dominant
Ireland: Dublin
Lithuania: Vilnius and Kanaus
Norway: Oslo
Sweden: Linkoeping (Southwest of Stockholm)
United Kingdom: Cet Central (near Birmingham)
Greece: Hellinikon
Italy: Rome and Verona
Portugal: Porto
Serbia: Belgrade
Spain: Navacerrada (Madrid) and Barcelona
Macedonia: Lazaropole (Southwest of Skopje)
Austria: Vienna
Belgium: Brussels
France: Paris
Germany: Köln, Berlin, Munich
Netherlands: De Bilt (near Utrecht)
Switzerland: Zurich, and Geneva
Australia: Perth, Adelaide, Sydney
United States: New York, Los Angeles, Chicago
Canada: Toronto
New Zealand: Wellington
Nations ultimately included in our analysis:
Algeria, Australia, Austria, Bahamas, Bangladesh, Barbados, Belgium, Belize, Canada,
Colombia, Costa Rica, Czech Republic, Egypt, El Salvador, Estonia, Finland, France, Germany,
Greece, Honduras, Hungary, India, Ireland, Italy, Jamaica, Japan, Kazakhstan, Kenya, Lithuania,
Macedonia, Mexico, Moldova, Mongolia, Morocco, Netherlands, New Zealand, Norway,
Pakistan, Paraguay, Philippines, Poland, Portugal, Romania, Serbia, Singapore, South Africa,
Spain, Sri Lanka, Sweden, Switzerland, Tajikistan, Thailand, Uganda, United Kingdom,
Uruguay, United States, Venezuela
III. Control Variables
Multiple control variables were vetted, but many were not available for developing nations,
restricting our opportunities to utilize them for all nations in the sample
Used in manuscript:
1. Population: var "totalpop"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
Measure reflects total annual population as estimated or by census. Used to compute homicide
rates, not included separately in analysis.
2. Consumer Price Index: var. "cpi" and “cpidiff” (differenced)
Source: World Bank: http://data.worldbank.org/country, accessed 10-6-2014
From File Meta Data: "Consumer price index (2010 = 100). Consumer price index reflects
changes in the cost to the average consumer of acquiring a basket of goods and services that may
be fixed or changed at specified intervals, such as yearly. The Laspeyres formula is generally
used."
CPI was used to reflect the immediacy with which price hikes may affect violence levels
3. Infant mortality: var. "mortalityinfant"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
From File Meta Data : " Infant mortality rate is the number of infants dying before reaching one
year of age, per 1,000 live births in a given year."
4. Youth Unemployment: vars. "maleyouthunemployed" and "unemdiff” (differenced)
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
From File meta Data: "Unemployment, male youth total (% of total labor force ages 15-24)
(modeled ILO estimate). Youth unemployment refers to the share of the labor force ages 15-24
without work but available for and seeking employment."
Male youth unemployment rather than all unemployment is used because it affects the group at
highest risk for offending and victimization in homicides. We expect that variation in homicide
would be most affected by young males, the available time that youth have to spend on
delinquent activity, including violence, may be impacted by their unemployment rates.
5. Immigrant Population: var "migrants"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
From File Meta Data: "International migrant stock (% of population) International migrant stock
is the number of people born in a country other than that in which they live. It also includes
refugees. The data used to estimate the international migrant stock at a particular time are
obtained mainly from population censuses. The estimates are derived from the data on foreignborn population--people who have residence in one country but were born in another country.
When data on the foreign-born population are not available, data on foreign population--that is,
people who are citizens of a country other than the country in which they reside--are used as
estimates. After the breakup of the Soviet Union in 1991 people living in one of the newly
independent countries who were born in another were classified as international migrants.
Estimates of migrant stock in the newly independent states from 1990 on are based on the 1989
census of the Soviet Union. For countries with information on the international migrant stock for
at least two points in time, interpolation or extrapolation was used to estimate the international
migrant stock on July 1 of the reference years. For countries with only one observation, estimates
for the reference years were derived using rates of change in the migrant stock in the years
preceding or following the single observation available. A model was used to estimate migrants
for countries that had no data."
Although the literature on immigrants and violence is certainly somewhat mixed, a recent
emerging consensus suggests (Martinez 2014) that migrants may actually lower the incidence of
violence in some nations (the US, for instance). For this reason we included the proportion of
migrants as a longer term variation.
6. Battle Casualties: vars. "battledeath" and "batdeathrate"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
From File metadata: "Battle-related deaths (number of people) Battle-related deaths are deaths in
battle-related conflicts between warring parties in the conflict dyad (two conflict units that are
parties to a conflict). Typically, battle-related deaths occur in warfare involving the armed forces
of the warring parties. This includes traditional battlefield fighting, guerrilla activities, and all
kinds of bombardments of military units, cities, and villages, etc. The targets are usually the
military itself and its installations or state institutions and state representatives, but there is often
substantial collateral damage in the form of civilians being killed in crossfire, in indiscriminate
bombings, etc. All deaths--military as well as civilian--incurred in such situations, are counted as
battle-related deaths."
Battle related deaths are used to account for any changes in homicide levels that may be linked to
group conflicts.
Not used in manuscript, but used in alternative specifications in appendix:
1. Income (not used): vars. "income", and "incomediff"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
From File Meta Data: "GNI per capita, Atlas method (current US$) GNI per capita (formerly
GNP per capita) is the gross national income, converted to U.S. dollars using the World Bank
Atlas method, divided by the midyear population. GNI is the sum of value added by all resident
producers plus any product taxes (less subsidies) not included in the valuation of output plus net
receipts of primary income (compensation of employees and property income) from abroad.
GNI, calculated in national currency, is usually converted to U.S. dollars at official exchange
rates for comparisons across economies, although an alternative rate is used when the official
exchange rate is judged to diverge by an exceptionally large margin from the rate actually
applied in international transactions. To smooth fluctuations in prices and exchange rates, a
special Atlas method of conversion is used by the World Bank. This applies a conversion factor
that averages the exchange rate for a given year and the two preceding years, adjusted for
differences in rates of inflation between the country, and through 2000, the G-5 countries
(France, Germany, Japan, the United Kingdom, and the United States). From 2001, these
countries include the Euro area, Japan, the United Kingdom, and the United States."
Income was dropped from the analysis due to high levels of correlation with multiple other
independent variable (especially CPI and Unemployment), thereby potentially distorting some of
the results (effects can be seen in alternative specifications below).
2. Prison Population (not used): vars. "prison" and "prisonrate” (rate per 100,000)
Sources: United Nations and the International Centre for Prison Studies
Since UN data is limited on incarceration we added in data from ICPS to fill gaps. Remaining
missing data is interpolated (linear). In general, the prison population is calculated by adding all
inmates of local, state and federal prisons/jails together and may include people awaiting trial,
and political prisoners.
UN Data is obtained through:
http://www.unodc.org/documents/data-andanalysis/statistics/crime/CTS2013_Persons_detained.xls (2000-2012)
http://www.unodc.org/pdf/crime/seventh_survey/7th%20all%20040331.xls (1998-2000)
http://www.unodc.org/pdf/crime/sixthsurvey/cs_2001_06_27.xls (1995-1997)
Website accessed 10-6-2014
ICPS data is accessed through: http://www.prisonstudies.org/world-prison-brief
Website accessed 10-6-2014
Prison population was dropped from the final analysis as it did not attain significance in all but
one model (Europe) and led to a reduction in the overall sample size.
3. Number of Police officers (not used): vars. "Police", “Policerate” (rate per 100,000)
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
While police levels is an interesting variable could not be used as a control variable as the data
were not complete enough for our final sample and would have reduced the number of nations in
our sample too much. Its effects appear largely negative (more police, fewer homicides) and
may thus be seen as one way to mitigate the predicted effects of rising temperatures, but it must
be stressed that the data available for this variable is largely limited to Western nations, where
climate effects on crime appear lower.
4. Life expectancy at birth (not used): var. "lifeexpectancy"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
Life expectancy is expressed in years.
Life expectancy and child mortality, unsurprisingly correlated very highly. We decided to retain
child mortality because it is measured as a rate offering slightly greater precision. We also
believe the care very young children receive better represents attitudes toward life and ability of
nations to preserve life, whereas life expectancy may be biased by the demographic composition
of a nation.
5. Urban population (not used): var. "urbanpop"
Source: World Bank http://data.worldbank.org/country, accessed 10-6-2014
Urban population is expressed as a percentage of total population. While we do not dispute that
urban population growth may be relevant when comparing nations, we do not think that urban
growth within nations manifests itself rapidly (in the span of under two decades) alters homicide
levels.
Nations used in final analysis:
The following nations had complete enough data to be utilized in the final analysis:
Algeria, Australia, Austria, Bahamas, Bangladesh, Barbados, Belgium, Belize, Canada,
Colombia, Costa Rica, Czech Republic, Egypt, El Salvador, Estonia, Finland, France, Germany,
Greece, Honduras, Hungary, India, Ireland, Italy, Jamaica, Japan, Kazakhstan, Kenya,
Lithuania, Macedonia, Mexico, Moldova, Mongolia, Morocco, Netherlands, New Zealand,
Norway, Pakistan, Paraguay, Philippines, Poland, Portugal, Romania, Serbia, Singapore, South
Africa, Spain, Sri Lanka, Sweden, Switzerland, Tajikistan, Thailand, Uganda, England and
Wales, Uruguay, United States, Venezuela.
Part Two-Alternative Modeling Methods:
In this section, we specify two alternative routes by which we can model our results: multi-level
modeling with fixed and random effects, and fixed effects regression.
Table 1. Climate Change and Homicide Rates, Multi-Level Modeling Results
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North America/
AUS/NZ
.144**
(.044)
.030 **
(.011)
-.006
(.008)
.014
(.081)
-.066
(.165)
.010
(.048)
-1.264
(1.591)
.069 *
(.041)
-.004
(.009)
.032 *
(.014)
.065 *
(.038)
.022
(.043)
.139 *
(.065)
.605
(1.065)
-.001
(.024)
.007
(.007)
.007
(.010)
-.000
(.036)
-.091***
(.016)
.004
(.003)
1.757**
(.758)
.002
(.012)
.001
(.002)
.003
(.004)
.005
(.017)
.005
(.028)
.126
(.108)
.719***
(.285)
.005
(.015)
.013 *
(.006)
.033 **
(.012)
-.128 *
(.055)
-.001
(.013)
-
.028 ***
(.003)
-.010
(.015)
.050 *
(.024)
1.113***
(.047)
-.022 **
(.007)
-.371
(1.165)
-3.133***
(.303)
1.78*10-17
(-)
.0003
(-)
2.06*10-22
(-)
.021
(-)
7.56*10-19
(-)
1.90*10-20
(-)
1.72*10-15
(-)
3.29*10-25
(-)
.001
(-)
.006
(-)
5.12*10-23
(-)
1.41*10-23
(-)
1.31*10-21
(-)
2.69*10-27
(-)
.0003
(-)
.006
(-)
8.75*10-30
(-)
2.59*10-27
(-)
2.83*10-22
(-)
2.40*10-26
(-)
6.22*10-26
(-)
.0002
(-)
.003
(-)
.019
(-)
.0004
(.0006)
.0002*
(.0001)
4.52*10-20
(-)
.007*
(.004)
.0003
(.003)
-4
Fixed Effects Parameters 1,2,3,4
Temperature
.041***
(.012)
Unemployment
.007 **
(differenced)
(.003)
Price Index
.017***
(differenced)
(.005)
Infant Mortality
.020
(.022)
Migrant
-.026
Percentage
(.017)
Battle Deaths
.005
(.004)
Constant
.488*
(.239)
Random-Effects Parameters
Temperature
Unemployment
(differenced)
Price Index
(differenced)
Infant Mortality
Migrant
Percentage
Battle Deaths
.003*
(.001)
.0001
(.0001)
.0004*
(.0002)
.013***
(.003)
.004*
(.002)
2.16 * 10-20
(-)
.061
(.266)
3.03*10-22
(-)
.0004
(-)
2.97*10-24
(-)
3.49*10-23
(-)
2.01*10-22
(-)
7.73*10-18
(-)
Table 1 (Continued). Climate Change and Homicide Rates (Multi-Level Modeling results)
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North
America/AUS/NZ
Variance
(Constant)
N
Chi-Squared
Prob>ChiSquared
Log-Likelihood
1.077*
(.584)
817
105.71
<.0001
4.852
(-)
65
51.63
.0001
1.377
168
56.33
.0001
2.017
(-)
117
152.82
<.0001
.557
(-)
160
271.18
<.0001
.070
(.060)
239
218.12
<.0001
7.29*10-21
(-)
68
4046.81
<.0001
95.718
9.018
-10.513
48.057
87.717
124.807
74.387
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors.
4. No battle casualties were recorded in included European countries during study period, hence no estimate could be created
Table 2. Climate Change and Homicide Rates, Fixed Effects Regression Results
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North America
/AUS/NZ
Temperature 1
.018** 2
(.007) 3
.003*
(.002)
.006**
(.003)
.040**
(.015)
-.031*
(.016)
.008*
(.004)
.800***
(.056)
760
5.91
<.0001
.100*
(.048)
.025**
(.009)
.002
(.008)
-.091*
(.042)
-.317*
(.175)
-.003
(.060)
1.695**
(.522)
59
3.48
.006
.072***
(.023)
.008
(.005)
.019**
(.006)
.010
(.042)
.067
(.147)
.018
(.048)
.886***
(.221)
156
3.40
.004
.016
(.016)
.001
(.005)
.006
(.006)
.022
(.016)
-.166***
(.035)
.011**
(.004)
1.303***
(.162)
108
6.92
<.0001
-.006
(.013)
-.0001
(.002)
.001
(.005)
.089***
(.014)
.078**
(.031)
.056
(.034)
-.096
(.119)
149
14.07
<.0001
.022*
(.012)
.001
(.004)
.006
(.009)
.134*
(.064)
-.012
(.016)
.001
(.026)
.019*
(.009)
.018
(.015)
.666***
(.108)
.017
(.017)
-.838
(1.248)
-2.066**
(.650)
64
9.79
<.0001
Unemployment
(differenced)
Price Index
(differenced)
Infant Mortality
Migrant
Percentage
Battle Deaths
Constant
N
F-Statistic
Prob>F-Statistic
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed test.
3. The values in parenthesis are standard errors.
4. No battle casualties were recorded in included European countries during study period, hence no estimate could be created
4
-.374**
(.127)
224
3.19
.009
Part Three-Including One Additional Control at a Time using ARFIMA
Table 1. Climate Change and Homicide Rates among All Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.056 ***
.056 ***
(.012)
(..011)
Male Youth
.003
.002
Unemployment
(.002)
(.002)
Price Index
.004 ***
.004 ***
(.001)
(.001)
Infant Mortality
.0004
.001
(.009)
(.008)
Migrant Percentage
-.004
-.003
(.009)
(.009)
Battle Deaths
.002
.002
(.006)
(.006)
Income
-.007
(.006)
Prison Population
-
Model Three
Model Four
Model Five
Model Six
.049 **
(.011)
.004 *
(.002)
.004 ***
(.001)
.009
(.008)
-.007
(.009)
.002
(.005)
-
.035 **
(.011)
.003
(.002)
.008 ***
(.001)
.020
(.011)
-.004
(.009)
.012
(.028)
-
054 ***
(.011)
.004 *
(.002)
.004 ***
(.001)
-.011
(.010)
-.001
(.009)
.001
(.006)
-
.058 ***
(.011)
.003
(.002)
.004 ***
(.001)
-.0002
(.009)
-.003
(.010)
.002
(.006)
-
-
-
-
-.0000005
(.0000003)
-
-
-.045 *
(.016)
-
Number of Police
Officers
-
.0002
(.0002)
-
Life Expectancy
-
-
-
Urban Population
-
-
-
Intercept
.049
.033
.022
(.342)
(.132)
(.419)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
1.529
1.517
1.458
1.501
(1.529)
(1.232)
(1.208)
(1.225)
Year intercept
.089
.00006
.034
.144
(.298)
(.008)
(.185)
(.379)
N
673 (57 Countries, 666 (57 Countries, 634 (55 Countries, 541 (50 Countries,
13 Years)
13 Years)
13 Years)
13 Years)
Log-Likelihood
-129.794
-131.715
-85.377
-65.034
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
.040
(.155)
.0000001
(.00000009)
.052
(.250)
1.299
(1.140)
.000000001
(.00003)
673 (57 Countries,
13 Years)
-129.564
1.558
(1.248)
.033
(.183)
673 (57 Countries,
13 Years)
-144.243
3. The values in parenthesis are standard errors
Table 2. Climate Change and Homicide Rates among African Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.165 *
.168 *
(.087)
(.092)
Male Youth
.016 *
.016 *
Unemployment
(.011)
(.011)
Price Index
.015 ***
.015 ***
(.005)
(.005)
Infant Mortality
-.130 *
-.131 *
(.075)
(.076)
Migrant Percentage
-.100
-.094
(.190)
(.197)
Battle Deaths
.177 *
.176 *
(.074)
(.075)
Income
-.021
(.195)
Prison Population
-
Model Three
Model Four
Model Five
Model Six
.202 **
(.082)
.014 *
(.008)
.012 *
(.004)
-.105 *
(.065)
-.080
(.203)
.042
(.054)
-
.015
(.138)
-.053 *
(.025)
.009
(.007)
-.027
(.112)
.054
(.056)
-.035
(.066)
-
.080
(.076)
.020 *
(.010)
.023 ***
(.004)
-.222 ***
(.067)
-.151
(.162)
.160 *
(.063)
-
.121 +
(.076)
-.004
(.011)
.012 ***
(.004)
-.073
(.071)
.154
(.166)
.111 *
(.067)
-
-
-
-
-.00003
(.00002)
-
-
-.170 ***
(.040)
-
-
Number of Police
Officers
-
-
-.003
(.002)
-
Life Expectancy
-
-
-
-
Urban Population
-
-
-
-
Intercept
.143
.143
.740
-.001
(.953)
(.955)
(.944)
(.846)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
5.121
5.137
3.943
6.031 * 10-16
(2.263)
(2.266)
(1.986)
(2.456 * 10-8)
Year intercept
.032
.034
.036
.013
(.178)
(.184)
(.191)
(.112)
N
62 (6 Countries,
62 (6 Countries,
48 (5 Countries,
28 (3 Countries,
13 Years)
13 Years)
13 Years)
12 Years)
Log-Likelihood
-38.046
-38.764
-22.868
-17.094
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors
.312
(.680)
-.161 ***
(.042)
.091
(1.490)
2.390
(1.546)
.046
(.215)
62 (6 Countries, 13
Years)
-32.846
12.975
(3.602)
.042
(.205)
62 (6 Countries,
13 Years)
-35.254
Table 3. Climate Change and Homicide Rates among Asian Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.018
.020
(.031)
(.030)
Male Youth
-.014
-.013
Unemployment
(.013)
(.013)
Price Index
.0004
-.0002
(.003)
(.003)
Infant Mortality
-.021 +
-.020 +
(.014)
(.013)
Migrant Percentage
-.088 ***
-.078 ***
(.021)
(.021)
Battle Deaths
.001
.002
(.005)
(.005)
Income
-.034 *
(.020)
Prison Population
-
Model Three
Model Four
Model Five
Model Six
.013
(.030)
-.020 +
(.013)
-.003
(.003)
-.026 *
(.015)
-.075 **
(.024)
.001
(.005)
-
.017
(.043)
-.028 *
(-.016)
.005
(.006)
-.007
(.027)
-.094 ***
(.021)
-.126
(.099)
-
.023
(.019)
-.006
(.012)
-.001
(.003)
-.061 ***
(.015)
-.040 *
(.017)
-.0002
(.005)
-
.010
(.032)
-.014
(.013)
-.00006
(.003)
-.023 *
(.015)
-.086 ***
(.021)
.001
(.005)
-
-
-
-
-.0000001
(.0000004)
-
-
-.222 ***
(.044)
-
-
Number of Police
Officers
-
-
.001
(.001)
-
Life Expectancy
-
-
-
-
Urban Population
-
-
-
-
Intercept
.053
.081
-.011
-.049
(.425)
(.447)
(.452)
(.539)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
1.342
1.264
1.302
1.790
(1.158)
(1.125)
(1.141)
(1.338)
Year intercept
.016
.044
.035
.019
(.125)
(.211)
(.187)
(.138)
N
99 (9 Countries,
99 (9 Countries,
94 (9 Countries,
64 (8 Countries,
13 Years)
13 Years)
13 Years)
13 Years)
Log-Likelihood
-23.522
-25.019
-28.838
-30.846
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors
-.028
(.427)
-.010
(.011)
-.019
(.442)
.317
(.563)
.138
(.372)
99 (9 Countries, 13
Years)
-18.691
1.332
(1.154)
.024
(.156)
99 (9 Countries,
13 Years)
-26.714
Table 4. Climate Change and Homicide Rates among European Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1
Temperature
.018
.018
(.014)
(.014)
Male Youth
.005 *
.004 *
Unemployment
(.002)
(.002)
Price Index
.016 *
.016 *
(.007)
(.007)
Infant Mortality
-.220 ***
-.220 ***
(.065)
(.065)
Migrant Percentage
-.029 **
-.029 **
(.010)
(.010)
Battle Deaths
n/a 3
n/a 3
Model Three
Model Four
Model Five
Model Six
.021
(.015)
.005 *
(.002)
.021 **
(.008)
-.188 ***
(.069)
-.028 **
(.010)
n/a 3
.022
(.015)
.005 *
(.003)
.018 **
(.007)
-.212 ***
(.066)
-.028 **
(.010)
n/a 3
.018
(.014)
.005 *
(.002)
.018 **
(.007)
-.220 ***
(.064)
-.026 **
(.010)
n/a 3
.023 *
(.014)
.004 *
(.002)
.017 **
(.007)
-.202 ***
(.065)
-.026 **
(.010)
n/a 3
-
-
-
-
-
-
-
-.000001
(.0000008)
-
-
-.068
(.046)
-
-
Income
-
Prison Population
-
-.004
(.007)
-
Number of Police
Officers
-
-
-.002 *
(.001)
-
Life Expectancy
-
-
-
-
Urban Population
-
-
-
-
Intercept
-.127
-.131
-.131
-.128
-.127
(.109)
(.142)
(.113)
(.157)
(.119)
ARFIMA, Random Effects, variance and standard deviation
Country intercept
.097
.096
.099
.105
.089
(.311)
(.309)
(.315)
(.324)
(.299)
Year intercept
.003
.012
.004
.016
.006
(.058)
(.108)
(.062)
(.125)
(.079)
N
183 (15 Countries, 183 (15 Countries, 181 (15 Countries, 181 (15 Countries, 183 (15 Countries,
13 Years)
13 Years)
13 Years)
13 Years)
13 Years)
Log-Likelihood
40.058
36.242
33.820
27.116
38.954
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-values.
3. No battle casualties were recorded in included European Countries during study period, hence no estimate could be created
.010 +
(.007)
-.130
(.194)
.086
(.293)
.030
(.173)
183 (15 Countries,
13 Years)
36.932
Table 5. Climate Change and Homicide Rates among Former Communist Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
-.003
-.001
(.014)
(.014)
Male Youth
-.003 +
-.002
Unemployment
(.002)
(.002)
Price Index
.003 +
.002
(.002)
(.002)
Infant Mortality
.026 +
.027 +
(.019)
(.019)
Migrant Percentage
.047 **
.048 **
(.019)
(.018)
Battle Deaths
-.001
.002
(.038)
(.038)
Income
.051 *
(.031)
Prison Population
-
Model Three
Model Four
Model Five
-.005
(.012)
-.004 *
(.002)
.005 **
(.002)
.035 +
(.023)
.040 **
(.016)
-.038
(.033)
-
-.005
(.013)
-.004 *
(.002)
.005 **
(.002)
.020
(.021)
.046 **
(.017)
-.029
(.033)
-
-.004
(.014)
-.001
(.002)
.001
(.002)
.004
(.021)
.031 *
(.020)
.009
(.038)
-
-
-
-.000001
(.000004)
-
.097
(.217)
-.065 **
(.028)
.020
(.198)
.434
(.659)
.000004
(.002)
114 (10 Countries,
13 Years)
14.235
.396
(.629)
.00003
(.005)
131 (11 Countries,
13 Years)
14.109
Number of Police
Officers
-
-
-.0004
(.0003)
-
Life Expectancy
-
-
-
Intercept
.028
.036
.145
(.210)
(.210)
(.247)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
.445
.446
.477
(.667)
(.668)
(.691)
Year intercept
.0004
.0000004
.008
(.002)
(.001)
(.091)
N
131 (11 Countries, 131 (11 Countries, 117 (10 Countries,
13 Years)
13 Years)
13 Years)
Log-Likelihood
14.228
13.002
21.373
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors
Table 6. Climate Change and Homicide Rates among Latin American Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.043
.046
(.044)
(.044)
Male Youth
.010
.013
Unemployment
(.009)
(.010)
Price Index
-.005 *
-.004
(.003)
(.003)
Infant Mortality
.040
.030
(.033)
(.034)
Migrant Percentage
.012
.017
(.041)
(.041)
Battle Deaths
.264 ***
.259 ***
(.068)
(.068)
Income
-.033
(.052)
Prison Population
-
Model Three
Model Four
Model Five
Model Six
.051
(.044)
.009
(.009)
-.007 *
(.003)
.029
(.034)
.032
(.043)
.260 ***
(.067)
-
-.009
(.050)
-.007
(.010)
-.005
(.004)
.078 *
(.044)
.052
(.056)
.177
(.672)
-
.048
(.042)
.011
(.009)
-.005 *
(.003)
.033
(.035)
.005
(.038)
.258 ***
(.069)
-
.021
(.048)
.013
(.009)
-.005 *
(.003)
.027
(.034)
.026
(.044)
.270 ***
(.067)
-
-
-
-
-.000006 ***
(.000002)
-
-
-.073
(.078)
-
-
Number of Police
Officers
-
-
-.0003
(.0003)
-
Life Expectancy
-
-
-
-
Urban Population
-
-
-
-
Intercept
-.064
-.064
-.178
-.242
(.375)
(.383)
(.420)
(.452)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
.638
.667
.712
1.708
(.799)
(.817)
(.844)
(1.307)
Year intercept
.083
.086
.111
.023
(.287)
(.294)
(.332)
(.152)
N
146 (12 Countries, 139 (12 Countries, 142 (12 Countries, 102 (10 Countries,
13 Years)
13 Years)
13 Years)
13 Years)
Log-Likelihood
-56.336
-56.527
-60.573
-47.946
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors
-.064
(.481)
-.017
(.013)
-.064
(.311)
.482
(.694)
.186
(.432)
146 (12 Countries,
13 Years)
-57.678
.817
(.904)
.024
(.154)
146 (12 Countries,
13 Years)
-59.038
Table 7. Climate Change and Homicide Rates among North American Countries
Independent
Variable
Model One
(Original)
Model Two
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.028 **
.028 ***
(.009)
(.007)
Male Youth
.009
.007
Unemployment
(.010)
(.011)
Price Index
-.014
-.013
(.015)
(.016)
Infant Mortality
1.078 ***
1.057 ***
(.131)
(.115)
Migrant Percentage
-.022
-.026
(.019)
(.017)
Battle Deaths
.249
.353
(1.609)
(1.629)
Income
-.007
(.015)
Prison Population
-
Model Three
Model Four
Model Five
.024 ***
(.006)
.007
(.009)
-.007
(.014)
1.043 ***
(.096)
.007
(.026)
-.186
(1.604)
-
.032 ***
(.007)
.007
(.010)
-.001
(.018)
1.087 ***
(.105)
.002
(.028)
.033
(1.607)
-
.028
(.030)
.009
(.012)
-.013
(.016)
-1.093 ***
(.275)
-.021
(.023)
.243
(1.666)
-
-
-
-.068
(.050)
-
-
Life Expectancy
-
-
.001
(.0004)
-
Urban Population
-
-
-
Intercept
-.156
-.156
-.156
(.121)
(.140)
(.114)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
.001
.0004
.00001
(.036)
(.021)
(.00001)
Year intercept
.012
.017
.010
(.107)
(.130)
(.102)
N
52 (4 Countries,
52 (4 Countries,
52 (4 Countries,
13 Years)
13 Years)
13 Years)
Log-Likelihood
9.870
6.700
4.300
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-test.
3. The values in parenthesis are standard errors
-.156
(.120)
.002
(.065)
-.156
(.129)
.00001
(.00001)
.012
(.108)
52 (4 Countries, 13
Years)
8.687
.002
(.040)
.013
(.116)
52 (4 Countries,
13 Years)
8.048
Part Four-Correlation Matrix between Variables
Variable
Homicide
Rates
Temperature
Unemployment
Price
Index
Infant
Morality
Migrant
Percentage
Battle
Deaths
Income
Prison
Population
Number
of Police
Officers
Life
Expectancy
Homicide
Rates
1.000
Temperature
.362
1.000
Unemployment
-.008
-.246
1.000
Price Index
.317
-.018
.036
1.000
Infant
Mortality
.467
.394
-.099
-.391
1.000
Migrant
Percentage
.326
.128
.029
-.173
.291
1.000
Battle Deaths
-.354
-.096
-.133
.139
-.394
-.206
1.000
Income
-.258
-.152
-.155
.181
-.234
-.085
.221
1.000
Prison
Population
.462
.135
-.047
-.118
.030
-.092
.223
-.098
1.000
Number of
Police Officers
.050
.200
-.140
-.025
.429
.017
-.200
-.057
.009
1.000
Life
Expectancy
-.621
-.271
.034
.448
-.856
-.358
.383
.285
-.223
-.144
1.000
Urban
Population
-.027
-.339
.063
-.223
.090
-.034
-.175
-.042
.013
-.050
-.102
Urban
Population
1.000
Part Five-ARFIMA-MLM Models with Year and Country Binaries
Independent
Variable
All Countries
Africa
Latin America
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.050 **
.177 *
.050
(.014)
(.090)
(.055)
Male Youth
.002
.014
.017 *
Unemployment
(.002)
(.011)
(.009)
Price Index
.004**
.015 ***
-.007 *
(.001)
(.004)
(.002)
Infant Mortality
-.010
-.171 *
-.018
(.010)
(.082)
(.038)
Migrant
.010
-.055
.186 **
Percentage
(.011)
(.199)
(.072)
Battle Deaths
.002
.200 **
.258 ***
(.006)
(.076)
(.068)
Intercept
-1.240
1.841
-2.089 *
(1.310)
(1.371)
(1.024)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
1.640
1.230
.838
(1.281)
(1.109)
(.915)
Year intercept
.040
.189
.024
(.201)
(.435)
(.153)
N
673 (57
62 (6 Countries,
146 (12
Countries, 13
13 Years)
Countries, 13
Years)
Years)
Log-Likelihood
-34.980
-25.788
-38.554
Asia
Former
Communist
Europe
North America
and Australia
.018
(.046)
-.018
(.014)
-.0003
(.003)
-.030 *
(.017)
-.088 ***
(.028)
.001
(.005)
.247
(1.456)
-.004
(.015)
-.003
(.002)
.003
(.002)
.032
(.024)
.028
(.021)
-.001
(.038)
-.260
(.771)
.018
(.017)
.004
(.003)
.014 *
(.007)
-.252 ***
(.070)
-.026 *
(.013)
n/a 4
.035
(.046)
.010
(.010)
-.011
(.018)
.561
(.413)
.033
(.036)
-.437
(1.632)
-.590
(.850)
1.189
(1.090)
.027
(.165)
99 (9 Countries,
13 Years)
.572
(.756)
.012
(.109)
131 (11
Countries, 13
Years)
26.074
.202
(.449)
.006
(.075)
183 (15
Countries, 13
Years)
45.271
-9.494
-.037
(.490)
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-values.
3. The values in parenthesis are standard errors.
4. No battle casualties were recorded in included European Countries during study period, hence no estimate could be created.
.538
(.734)
.010
(.098)
52 (4 Countries,
13 Years)
10.948
Part Six- Models with Temperature as the Sole Independent Variable
Table 1: ARFIMA-MLM Models with Temperature as the Sole Independent Variable
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North America
and Australia
.030
(.040)
.153
(.495)
-.001
(.013)
.038
(.292)
.008
(.016)
.0002
(.036)
-.113
(.124)
-.156
(.378)
ARFIMA-MLM, Fixed-Effects, coefficients and standard error 1,2,3
Temperature
.054 ***
.211 *
.077 *
(.011)
(.098)
(.041)
Intercept
-.924 *
.025
-.064
(.329)
(.798)
(.309)
ARFIMA-MLM, Random Effects, variance and standard deviation
Country intercept
Year intercept
N
Log-Likelihood
1.556
(1.248)
.043
(.209)
3.206
(1.790)
.070
(.264)
.691
(.831)
.032
(.180)
1.837
(1.356)
.043
(.207)
.687
(.829)
.020
(.140)
.119
(.344)
.005
(.071)
.507
(.712)
.013
(.113)
694 (58
Countries, 13
Years)
-132.601
62 (6 Countries,
13 Years)
154 (12
Countries, 13
Years)
-51.622
112 (10
Countries, 13
Years)
-32.231
131 (11
Countries, 13
Years)
28.830
183 (15
Countries, 13
Years)
33.779
52 (4 Countries,
13 Years)
-36.655
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed t-values.
3. The values in parenthesis are standard errors.
8.457
Table 2. Multi-Level Models with Temperature as the Sole Independent Variable
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North America
and Australia
.210 **
(.081)
-2.790 +
(1.782)
.044
(.040)
2.107 *
(1.002)
.032
(.034)
.082
(.875)
.003
(.022)
1.001 ***
(.310)
-.011
(.015)
.063
(.189)
.028
(.036)
.329
(.292)
.002
(.001)
1.006
(.285)
884 (58
Countries, 18
Years)
196.64
<.0001
2.43 * 10-25
(4.99 * 10-24)
2.591
(1.579)
65 (6 Countries,
18 Years)
4.10 * 10-27
(1.35 * 10-22)
1.516
(.727)
136 (10
Countries, 18
Years)
79.68
<.0001
.001
(.002)
.607
(.287)
173 (11
Countries, 18
Years)
345.20
<.0001
2.00 * 10-11
(2.12 * 10-10)
.100
(.037)
251 (15
Countries, 18
Years)
186.62
<.0001
.003
(.003)
.076
(.188)
72 (4 Countries,
18 Years)
23.60
.072
3.89 * 10-21
(5.10 * 10-20)
.626
(.259)
187 (12
Countries, 18
Years)
15.90
.599
-156.78
-23.33
-59.20
-13.25
23.85
87.39
48.87
Fixed Effects Parameters 1,2,3
Temperature
.031 **
(.013)
Constant
.573 **
(.229)
Random-Effects Parameters
Temperature
Variance
(Constant)
N
Chi-Squared
Prob>ChiSquared
Log-Likelihood
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed test.
3. The values in parenthesis are standard errors.
93.36
<.0001
Table 3. Fixed Effects Regression Models with Temperature as the Sole Independent Variable
Independent
Variable
All Countries
Africa
Latin America
Asia
Former
Communist
Europe
North America
and Australia
Temperature 1
.016 *2
(.007) 3
.946 ***
(.026)
826 (58
Countries, 13
Years)
4.78
.029
.072 +
(.050)
.083
(.264)
59 (6 Countries,
13 Years)
.051 **
(.020)
1.935 ***
(.102)
175 (12
Countries, 13
Years)
6.18
.014
.039 *
(.020)
-.171
(.135)
126 (10
Countries, 13
Years)
3.75
.055
-.007
(.012)
1.066 ***
(.018)
162 (11
Countries, 13
Years)
.41
.524
.013
(.011)
-.028
(.037)
236 (15
Countries, 13
Years)
1.37
.243
-.025
(.024)
.967 ***
(.093)
68 (4 Countries,
13 Years)
Constant
N
F-Statistic
Prob>F-Statistic
2.09
.154
1. Fixed effects for each year in the dataset are omitted.
2. +=p<.1, *=p<.05, **=p<.01, and ***=p<.001. All p significance levels based on single-tailed test.
3. The values in parenthesis are standard errors.
1.13
.293
Part Seven- Summary Statistics for Variables in Analyses
Table 1: All Countries
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
9.960
16.292
17.200
1.705
8.140
10.340
.207
-2.875
2.500
138.778
30.192
70.200
78.371
10.760
7.895
.399
.672
169.07
73.130
196,573
23.030
10.390
8.340
2.510
1.610
121.290
6.850
1,486,029
2.582
1.10
.150
0
-5.740
27.35
45.970
12
152.656
52.80
40.231
41.614
9.880
880.31
83.480
12,197,105
Mean
Standard Deviation
Minimum
Maximum
14.360
20.180
22.630
19.670
2.790
13.630
.460
17.240
3.200
68.70
27.460
52.000
72.400
22.840
1.450
212.710
.120
171.630
61.650
42.570
24.600
7.130
1.110
494.720
.220
97.990
8.050
19.480
25.310
11.600
.150
0
-.390
55.630
45.970
11.660
135.330
37.400
4.080
3024.000
1.210
399.070
70.910
73.710
Table 2: African Countries
SS
Variable
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
Table 3: Asian Countries
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
4.450
23.160
11.400
3.500
8.530
4.840
.210
-1.730
2.500
15.750
30.190
28.000
76.690
19.710
8.380
531.710
.340
150.360
71.110
25.060
16.340
14.510
1274.170
1.260
115.600
6.810
16.450
1.100
.300
0
-5.740
27.350
60.190
125.890
52.800
40.230
8413.000
7.710
435.070
83.480
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
1.290
11.010
16.400
.500
3.220
8.550
.510
4.610
4.200
3.090
19.520
55.500
88.890
3.010
10.140
0
1.360
90.040
79.230
52.960
10.140
.840
4.710
0
2.510
24.220
1.610
30.230
57.000
1.600
2.020
0
-5.020
40.370
85.260
15.040
107.430
7.000
22.610
0
9.880
164.090
82.940
100.00
Table 4: European Countries
Table 5: Former Communist Countries
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
5.310
8.680
24.350
4.470
4.470
13.840
.820
-2.880
3.800
17.120
17.210
70.200
71.510
9.860
7.670
23.080
.380
222.240
71.200
74.250
27.540
7.440
5.900
149.780
.540
113.160
3.650
10.510
2.580
1.600
.600
0
-1.090
58.150
62.390
51.110
119.280
35.100
21.490
1382.000
2.070
595.090
78.080
97.510
Table 6: Latin American Countries
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
31.59
24.040
14.880
24.330
3.730
6.140
5.290
14.820
2.900
138.780
28.010
33.900
72.700
10.890
4.690
61.670
.280
221.590
72.670
1,036,202
25.380
3.280
4.890
231.170
.580
139.200
2.820
26.43
19.580
3.900
.230
0
-2.280
57.640
65.420
12,197,100
152.660
19.800
18.580
1389.000
3.220
880.310
79.710
3,288,948
Table 7: North American Countries
Variable
Mean
Standard Deviation
Minimum
Maximum
Homicide Rates
Temperature
Male Youth
Unemployment
Price Index
Infant Mortality
Migrant Percentage
Battle Deaths
Income Difference
Prison Population
Life Expectancy
Urban Population
2.570
11.160
13.830
1.920
5.160
2.740
.890
3.050
9.000
8.110
18.720
21.000
86.490
3.680
18.200
3.240
1.680
222.010
79.110
64.93
11.110
.590
3.540
27.460
1.960
162.620
1.640
17.040
67.640
2.800
10.710
0
-1.590
96.440
75.620
35.480
105.360
5.100
22.540
233.000
9.660
530.100
82.100
93.700