Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium Member Libraries
http://www.archive.org/details/statisticalanalyOOkrue
Dewey
HB31
.M415
/U.1S-I5
working paper
department
of economics
A STATISTICAL ANALYSIS OF CRIME AGAINST
FOREIGNERS IN UNIFIED GERMANY
Alan B. Krueger
Jorn-Steffen Pischke
95-15
April 1995
massachusetts
institute of
technology
50 memorial drive
Cambridge, mass. 02139
A STATISTICAL ANALYSIS OF CRIME AGAINST
FOREIGNERS IN UNIFIED GERMANY
Alan B. Krueger
Jom-Steffen Pischke
95-15
April 1995
OF TECHNOLOGY
AUG 16
1995
LIBRARIES
95-15
A
Statistical Analysis of
Crime Against Foreigners
in Unified
Germany
Alan B. Krueger
U.S. Department of Labor and Princeton University
Jorn-Steffen Pischke
MIT
April 1995
We thank Dietmar Harhoff and Helmut Seitz for helpful discussions. We are grateful to Henning
Colsman-Freyberger for able research assistance and to Lisa Krueger for preparing the maps. We
Wortmann for data and the
support. The views expressed
thank Dietmar Harhoff, Manfred Kuechler and Martin
Policy Reform and the Sloan Foundation for financial
and do not
of Labor or the U.S. government
are solely those of the authors
reflect the position or opinion
Institute for
in this
paper
of the U.S. Department
Abstract
Germany has seen
a high and rising rate of anti-foreigner violence during the early 1990s.
analyze the determinants of crime against foreigners
and nature of such crime
at the
we assembled
a
new
county level based on newspaper reports.
dataset on the
We
number
find significant
differences in the patterns of violence in the eastern and western parts of the country.
incidence of anti-foreigner crime
is
is
The
higher in the east and rises with distance from the former
west German border. Economic variables
of crime once location in the east
To
like
unemployment and wages matter little for the
controlled for.
We
also find
no
relation with the
level
number
of foreigners in a county in the west. These results are hard to reconcile with simple economic
models of crime.
Introduction
1.
German
unification and the transition to a market
been accompanied by a high and
economy
former East Germany has
in
rising rate of violence against foreigners.
In 1992, the police
recorded 6,336 crimes against foreigners, 2,544 of which were violent crimes. The high number
of attacks that have occurred during the
need
to
last
few years and
their high
understand the determinants of crime against foreigners.
determine the correlates of violence against foreigners in Germany
Two main
few
years.
of youths.
The
strands of literature on ethnic violence in
first,
and
rate highlight the
In this paper,
we
seek
to
at a regional level.
Germany have
surfaced in the past
rooted in sociology and social psychology, consists of surveys primarily
These surveys
political leanings,
growth
try to evaluate the respondents' attitudes
towards foreigners,
their
The goal of
these
their propensities to participate in acts of violence.
studies is to isolate potential causes for individuals' participation in anti-foreigner crimes (see
Schnabel (1993) for a
large
number of such
critical
review of these studies).
surveys.
The second
strand of literature has
focus, and has tried to determine the degree to
to rising right-wing
extremism.
Much
Clear conclusions are rare despite the
more of a
which the current increase
of this literature
is
political science
in violence is related
rather conjectural.
In an intriguing
recent paper, Alber (1994) hypothesizes that, "anti- foreign violence should predominantly occur
in regions with a high proportion of immigrants, with high
political center that is
occupied by conservative and patriotic political
ostentatiously de-legitimate excessive nationalism." Alber
to test these hypotheses.
unemployment, and with a
elites
recommends using
local
who do
not
regional-level data
Our paper
is
the first that
we
are aware of to analyze regional-level data on the incidence
To
of crime against foreigners in Germany.
number and nature of
this
end we have assembled a new dataset on
anti-foreign crime at the county level.
police records are not publicly available,
we
the
Because disaggregated data from
constructed our data set from a comprehensive
review of newspaper reports on crime against foreigners.
Using regression-based analyses, we
probe which economic and demographic variables are associated with the incidence of violent
acts against foreigners.
draws on a large
Although
tradition in the
this
approach seems
American
to
be novel in the German context,
it
social science literature investigating the determinants
of lynchings at the turn of the century, and, for more recent periods, crime rates in general.
Some
clear results
emerge from our
analysis.
There are significant differences
incidence and pattern of the violence in the eastern and western parts of the country.
incidence of anti-foreign violence, either per foreigner or per capita,
states,
and many of the attacks
the Polish border.
Economic
strain, as
is
consistent with
much
The
greater in the eastern
in the east are concentrated in the rural coastal areas
and along
measured by high unemployment or low wages, seems
to contribute little to the incidence of violence
finding
is
in the
once location
in the east is held constant.
This
of the literature on crime for the U.S. and with the survey
evidence for Germany. In particular Heitmeyer (1992) stresses that even multiple indicators of
economic pressures do not help
We
also find
no relationship between
attacks for west
opportunities.
to predict likely
Germany.
involvement in violence
at the individual level.
the fraction of foreigners in a county
and the number of
This contradicts theories trying to relate crime
to
criminal
In general, our results are difficult to reconcile with simple
In particular, Alber's plausible hypotheses that crime should be positively
against foreigners.
associated with the level of immigrants and the rate of
supported.
A more complex
issues arising
unemployment
in an area are not
process seems to be at work, perhaps a repercussion of the difficult
from unification between east and west Germany.
The remainder of the paper
is
organized as follows. Section 2 provides some background
on Germany's migration history and the
some major events which seemed
acts.
economic theories of crime
to
rise
of violence against foreigners.
It
also describes
have played an important role as catalysts for future criminal
Section 3 describes our data set
Sections 4 and 5 present our main findings.
We
first
describe the broad regional patterns of the violence and of our covariates before presenting
regression results.
2.
The Rise
These findings are discussed
in section 6, section 7 offers our conclusions.
in Anti-Foreigner Violence
Post World
War
II
Germany
continually
has
had a large inflow of foreigners.
Nevertheless, the extent and composition of immigration has changed dramatically over time and
has differed considerably between the East and the
recruited as guest workers.
remained short-term
still
visitors in the East,
make up by
times, foreigners
While foreign workers from Communist countries
European workers who came
They
West At
to the
and
their
West during
far the largest
were actively
in the
Third World
numbers remained small, many of the southern
the
boom
group of foreigners
in the
in
1960s have settled permanently.
Germany.
Nefarious acts by right-wing extremists increased in both parts of
1980s.
According
during the
between 188 and 481 right-wing extremist acts were
to official reports,
counted East Germany for the two year period 1988 and 1989
period,
Germany
457 violent right-wing crimes took place
in the
(Kniitter, 1993).
West (BMI,
1994).
During the same
In the
mid 1980s,
right-wing extremist parties began to single out foreigners as targets in their propaganda.
number of
violent crimes against foreigners in the
West increased from 50
in
1985
to
The
152
in
1990 (Sippel, 1993).
In October 1990, East and
West Germany were
unified into a single nation.
union was accompanied by mass layoffs and plant closing in the
unemployment
The
rates.
The
eastern
Economic
east, leading to depression-level
economy was propped up by massive
subsidies from the west.
collapse of other economies in Eastern Europe, the opening of the borders of the former
Communist
Germany
countries,
and ethnic conflicts on the Balkan led
million immigrants.
made up
share grew to
At
just
1991,
Germany had
a gross annual inflow of
more than one
In addition, a growing share of the arrivals were asylum seekers.
numbers grew from 121,000
seekers
By
in the early 1990s.
to greater migration flows into
10%
30% by
in
of
1989
all
to
256,000
in
1991 and 438,000 in 1992. Although asylum
immigrants (including ethnic German
re-settlers) in 1989, their
1992.
the time of unification, foreigners
made up about
8 percent of the population in the
west. There were hardly any immigrants in the east, as most guest workers from the
period returned home.
There was also
little
for jobs in the east given the high rate of
government decided
Their
communist
reason for economic migrants in the west to look
unemployment
there.
Nevertheless, the federal
that eastern states should bear their fair share of the
burden of housing
newly arriving asylum seekers, so many former Soviet barracks were converted
homes
to
for
asylum seekers.
Figure
1,
based on aggregate monthly data released by the federal police, shows the
number crimes against
foreigners. There are various peaks in these data, associated with specific
events which were widely covered by the media.
On
Tuesday, September
a tragic assault against foreigners living in an asylum seeker
state of
home
in
17, 1991, there
Hoyerswerda
in the eastern
Saxony, about 20 miles from the Polish border. This incident incited a week of
closed.
demands and
throwing crowd.
number of
During the night of Tuesday, September 24, authorities gave
the police evacuated
Thirty-three people were injured during these riots.
The number of violent
in to these
some 230 asylum seekers through a clapping and
anti-foreigner attacks shot
rioting,
Local residents demanded that the asylum seeker
with more than 100 participants involved.
home be
was
up dramatically, from 104
in
At the same
August
to
961
stone-
time, the
in October.
anti-foreigner crimes also rose, from 152 in 1990 to 1257 in 1991 (Sippel,
1993).
There are three other incidents
in the
number of crimes
Station for
in the
in late
appear to have served as catalysts for similar bursts
against foreigners.
Asylum Seekers
Vorpommern
that
in
The
August of 1992. The pattern was similar
bombings of two residences
fire
are the riots against the Central Intake
Rostock-Lichtenhagen in the eastern costal
number of crimes occurred two month
with another
first
in
later
when
to
state
of Mecklenburg-
Hoyerswerda. Another spike
three Turkish residents died in fire
Molln, Schleswig-Holstein (west). The next spike
bombing of a Turkish home
is
associated
in Solingen, Northrhine-Westfalia (west) in
May
1993, killing five. In between these events, the
remained
above the level
far
Interestingly,
many
number of
attacks fell back to lower levels, but
in early 1991.
large-scale riots, such as the well
known
riots in
Rostock
—
the east.
These attacks were directed against asylum seekers homes. And
and
less well
known
incidents in
Wismar, Cottbus, and Quedlinburg
were supported by the neighboring German population.
The well know
different.
The
individuals.
It is
Germany
crimes.
local population
A
in perspective.
like a large
total
number.
was outraged by
total
is
rioters
rather
bombings by small groups of
fire
But many of these were minor crimes
Germany,
assaults,
rate against foreigners of
against foreigners in
By
contrast, the violent
like
propaganda and
there were 2,544 violent
and vandalism with major
about 3 per 100,000 residents.
of 150,000 violent crimes recorded in
188 per 100,000 residents.
each case the
picture in the west
number of crimes
These include homicides, arson and bombings,
were a
took place in
these crimes.
In 1992, the worst year of anti-foreign violence in
Overall, there
in
—
of 13,000 criminal acts were committed in 1992 and 1993,
damage. This implies a violent crime
rate of
The
Molln and Solingen were
also important to place the magnitude of the
which seems
threats.
attacks of
Hoyerswerda and
crime
Germany
in 1992, for a
rate for the U.S.
crime
was 758 per
100,000 in 1991. Other European countries have been experiencing crime against foreigners as
well during this period. While published statistics are not easily comparable across countries
seems clear
the
that the level of crime is
UK, which has
a similar
much lower
number of
racial
in
it
most other countries with the exception of
crimes as Germany.
In addition to the
foreigner sentiment
number of
1
It is
anti-foreigner crimes, Figure
questions on social and political attitudes.
They
One
of the questions asks respondents to state what
re-settlers as a
The
problem
The
results that
we show
economic conditions remain the most important concern
refer to
(i.e.
anti-
In addition, the poll asks a variety of
they feel are the two most important problems facing Germany.
throughout).
shows an index of
derived from the monthly survey "Politbarometer," a poll used to
determine political preferences of the voting population.
for the west only (since
1
are
in the east
any response naming foreigners, asylum seekers, or ethnic German
not the violence against foreigners
pattern is striking.
itself).
There was a surge of anti-foreigner sentiment
in early 1989.
Immigration into West Germany had been rising steadily during the past five years. Although
anti-foreigner crime had been rising since 1985, there
was no unusual increase
in
such incidents
Anti-foreigner sentiment subsided during the period of turmoil in Eastern Europe
in that year.
and the unification period.
It
rose during the Hoyerswerda riots, and remained
much
higher
during the next two years with peaks around times of major incidents such as Rostock, Molln,
and Solingen.
This raises a question as to whether public opinion
attacks
or
whether
publicity
about
Ohlemacher (1994) has attempted
anti-foreigner
is
a major driving force behind the
violence
influences
public
to address this issue using a different question
sentiment.
from the
"Politbarometer" survey. Ohlemacher's question refers to the perceived misuse of asylum laws.
His sample
is
correlates the
1
limited to monthly data between October 1991 and
December 1992. Ohlemacher
monthly number of incidents with the contemporaneous and lagged sentiment
This series was provided to us by Manfred Kuechler.
7
index.
He
finds the strongest correlation
he recognizes that
is
it
find
in Molln.
results are driven
The Politbarometer
directly preceding the riots in
the
in explaining the
temporal pattern of crime.
Ohlemacher's conclusion rather tenuous. His correlations are based on only 12
The
observations.
most variation the
data.
by a single data
point: the
May
bombing
in the
months
1,
was asked
in the poll
1994 period with two exceptions: June 1991 and July 1992. 2
available surrounding the
is
no indication
find
by Ohlemacher
alternative question, plotted in Figure
Hoyerswerda and Solingen, but not before the Rostock
we
the fire
Hoyerswerda and Rostock. These are obviously the months with
This means that the attitude index
the larger sample
month before
poll did not ask the question used
The
for the entire January 1991 to
riots.
period of the events of
Using the alternative question and
that public opinion leads the
number of
crimes.
The Newspaper Data Set
3
Official police data
level of states.
and subject
Unified
on the extent of crime against foreigners are not available below the
Germany has 16
to aggregation concerns.
data set on the
We
While
likely causality runs in both directions, he concludes that public opinion
is
an important driving factor
We
between crime and the lag of public opinion.
number
As
states,
a consequence,
3
The corresponding values
Zeitung
is
the
state-level analysis rather
we have assembled
a
new
remote
county-level
incidents of crime against foreigners derived from newspaper reports.
refer to this data set as the Zeitung
2
which makes a
in figure
German word
Data
1
Set.
3
This methodology has a long history
are interpolated.
for newspaper.
in
sociology, as
much
of the literature on lynchings and urban racial conflicts in the U.S.
on newspaper reports (see
Our
Frankfurt,
data set
e.g.
The Archiv
Germany.
newspapers and magazines.
4
list
originally
collected this
The
compiled by the Archiv
list
is
much more
dailies are typically regional
list
coded for us
checked by
newspapers with an extensive
The German newspaper
5
contains short summaries of each reported event
in a consistent fashion
us.
The
Sozialpolitik in
concentrated than the U.S. market, so these publications provide fairly
comprehensive coverage of the entire nation.
The
fiir
from several German daily and weekly
national circulation and coverage; and the biggest papers are included.
market
based
Danzger, 1975, Inverarity, 1976, and Olzak, 1989).
derived from a
is
is
final
by "Research on Demand"
sample consists of a
January 1991 and June 1993.
6
total
in Berkeley,
this list
were
CA, and coding was
of 1,056 incidents which occurred between
For most of these events
of crime, and often a description of the victims.
The contents of
we know
the date, location, the type
In a subset of cases
information on the number of victims and the number of perpetrators.
we have more
detailed
The following excerpt
gives a sample of the incidents reported on a randomly selected date (October 11, 1991), and
describes the nature of the information available to us.
4
The number of newspapers
varied
somewhat over
time.
In late 1992, the newspapers
included were Arztezeitung, Frankfurter Allgemeine, Frankfurter Rundschau,
Zeitung,
Suddeutsche
Zeitung,
Tagesspiegel,
die
tageszeitung,
Bayernkurier,
Allgemeines Sonntagsblatt, Freitag, Hamburger Rundschau, Parlament,
im Bundestag, and
5
The Archiv
found that
6
Neue Zuricher
Deutsches
Profil, Spiegel,
Woche
Zeit.
fiir
Sozialpolitik experimented with including
this resulted in
more papers
at
some time
but
only a small increase in the number of reports.
.The original data set contained 1,218 events.
We
eliminated incidents in which
we
could
not determine the location or date, incidents that were double entered, and incidents that did not
involve anti-foreign violence.
Sample of Incidents
-
Two
window of
skin heads shatter a
for Friday,
October
11,
1991
a dormitory for foreign workers in Rangsdorf/Zossen
County.
-
Three young Germans throw rocks
-
A
-
Also
Nigerian
is
at an
beaten severely by a 19 year old
in Saarbriicken,
Blood
stains
amputated under the hip
We
own
according to his
getting off a bus, sedated, and thrown
the evening.
asylum seeker
in
show
on
German
in
downtown
account, a refugee from Sri
train tracks.
that he has
in Erlensee near
He
is
Hanau.
Saarbriicken.
Lanka
abducted after
is
found by police near the tracks
been run over by a
train.
His right leg has
to
in
be
an emergency medical procedure.
aggregated the events in our database to the level of counties.
of data on 543 counties. Every incident
is
The sample
consists
assigned the same weight, independent of the severity
of the crime and the number of perpetrators involved.
For example, the
riots in
Rostock are
counted as one incident, just as the case of two skinheads shattering the window in a dormitory
for foreign workers
problematic.
To
mentioned
in the
sample from October
11,
assess whether the severity of the crime matters
1991 above.
we
This
is
clearly
provide separate analyses
for our full sample and for serious crimes (murder and arson) only.
We
provide two checks on the accuracy of the newspaper-based database.
First,
we
aggregated our Zeitung data to the state level, and computed the number of incidents per person
in the state.
We
then compared the official state-level police anti-foreigner crime rate to this
state-level Zeitung
correlated.
The
regression of the
crime
rate.
We
would expect the two sources of data
correlation coefficient
number of
to
be positively
.54.
A
on the number of incidents from
the
between the police data and Zeitung data
incidents from the police data
10
is
Zeitung data and an intercept yields a coefficient of 1.55
of .65), which suggests that
(std. error
the 1.55 police crimes are registered for each incident reported in the newspaper.
dummy
variable to this regression indicating whether the state
is in
7
If
we add
the east or west, the
a
dummy
has statistically insignificant effect while the Zeitung incident rate continues to be significant
This suggests that the Zeitung data do not have an obvious east-west regional
would
bias.
Although we
prefer to find a higher correlation between the police and Zeitung data, these results
suggest that the Zeitung data set conveys
some
signal
on the frequency of crimes against
foreigners.
Second,
we formed
nationwide monthly totals of the Zeitung data to compare to the
national
numbers released by the
period.
Three conclusions can be drawn from the
of
all
the incidents per
police.
This comparison
month recorded by
figure.
the police.
is
shown
First,
This
is
we
in Figure
2 for our sample
count only about one tenth
many
not surprising, as
of the
crimes the police investigate will be rather minor violations that are not reported in the press.
Second, our data track the temporal pattern of the police data rather well.
In particular,
we do
not see over reporting in the press after the events of Hoyerswerda, Rostock, and Solingen.
Third, after the Rostock riots our data tend to record a
police data.
the
When we
same drop
off.
limit the Zeitung
sample
much
smaller fraction of crimes than the
to violent crimes,
however,
we do
not observe
Figure 3 shows a scatter plot of the monthly police data against our data.
This also demonstrates the strong temporal correspondence between the two sources of data. The
figure
7
shows the regression
The
line
between the two. The correlation between the two
intercept term in this regression has a coefficient of 6.
11
series is 0.75.
Finally, Figure
4 compares our data
to the police data
We have grouped the data into four major groups:
or attempted murder which
we
primarily include unlawful entry,
newspaper coverage
is
call
murder,
weapons
8
by
the type of crime committed.
violent crime resulting in the death of a victim
arson, assault and battery, and others
offenses, threats, and graffiti.
best for the most serious crimes while
we miss many
Since the coverage of more serious crimes by the Zeitung data
estimate models using
It is
is
which
clear that the
of the minor crimes.
more complete, below we
our data as well as the subset of more serious crimes (murder and
all
arson).
One might
question whether the county
There are 543 counties
population of 147,000.
in
Germany, 215
is
the right level of observation for a our study.
in the east
and 328
Small counties, however, can have a population of only 20,000.
may
Perpetrators in crimes against foreigners
not necessarily
crime took place. However, according to results from police
locals.
For example, Raisch (1994) reports for the eastern
live in the
town of
in the west, with an average
the crime or within
come from
files,
state
the locality where the
most of the perpetrators are
of Saxony, that
20 kilometers. He also contends
that
90%
of suspects
77 out of the 82
suspects investigated by the police for participation in the Hoyerswerda riots are from that county
or the neighboring areas. All the remaining five had family ties in Hoyerswerda. These results
suggest that the county unit
8
The
murders
9
is
a reasonable level of observation.
police definition of murder
in the
newspaper data than
Leenen (1992)
is
9
narrower, with the result that there are actually more
in the police data.
cites similar results for Northrhine-Westfalia.
12
Variables
Our
covariates include several county-level measures of socioeconomic conditions, such
as the fraction of foreigners and the
unemployment
zur Entwicklung der Stadte, Kreise und
of the Cities, Counties, and
Towns
rate.
These are taken from "Aktuelle Daten
Gemeinden 1989/90" (Current Data on
the
Development
1989/90), published by the Bundesforschungsanstalt fur
Landeskunde und Raumordnung (BfLR, 1992), as well as some updated variables which we
obtained from the same
Germany, and not
institute.
regressors pertain to the
all
Since counties are of unequal size
dependent variable.
persons (the crime
we used
This
rate).
sociology literature
incidents divided
First
(e.g.,
is
more
Unfortunately,
the
we
same
date.
tried
two
number of
limited data are available for East
alternative approaches to scaling the
incidents of anti-foreign crime per 100,000
the typical variable analyzed in regional studies of crime in the
Land, McCall, and Cohen, 1990).
by the number of foreigners residing
in the
variable could be interpreted as a "victimization rate."
Second,
we used
the
number of
county. In this case the dependent
It
turns out that the results are
qualitatively similar with either variable.
Our main
dummy
regressors are the
unemployment
rate, the
share of foreigners in the county, a
variable for being located in the east, and several variables describing the degree of
urbanization of the county.
registered with the local
discouraged workers,
actively searching for
The unemployment
unemployment insurance
who have dropped
rate is
based on the number of individuals
office in
September 1992.
It
does not
out of the labor force, or unemployed workers
work but do not qualify
for
13
UI
benefits.
reflect
who
are
The unemployment
rate alone
may
not be a good indicator for economic conditions in a
In particular, easterners living closer to the former western border can
county.
easily to jobs in the west.
closer to the west.
may be
Moreover, social attitudes and police enforcement
Therefore,
distance from the former west
we
commute more
different
created an additional variable for counties in the east
German
border. This
is
measured by the
we
west German counties
The number of
region as of
German mainland excluding West
constructed the distance to the west
foreigners in a county
December 1991. This includes
be registered within the
and
11
who
is
the
number of those
first
month of
Berlin. For
their arrival in
leave the country
may
not be counted.
Germany, but the time
Ethnic
In terras of their appearance towards natives they
Germany
Asylum seekers
Thus, the small number of asylum seekers whose application
immigrants from the same countries.
in
In
registered in a particular
foreign workers, unless they are in
is
may
for less
typically will
lags differ across
denied very rapidly
German immigrants
Russia, Romania, and Poland) are also not counted since they
arrival.
10
this variable is set to zero.
than three months (e.g. seasonal workers), and asylum seekers.
states.
the
straight line (Euclidean)
distance between the geographic center of a county and the closest county in the west.
each case,
—
(primarily from
become German
citizens
upon
be indistinguishable from other
In fact, there are incidents of attacks of ethnic
our database. Also excluded are foreign military personnel. This
may especially
Germans
be a problem
for Soviet soldiers in the east Nevertheless, the foreign share variable should give a reasonably
accurate measure of the fraction of foreigners in a county.
10
Dietraar Harhoff provided the geographic coordinates of the counties to us.
"Based on personal communication with
the Federal Ministry of the Interior.
14
We
include a variety of regressors to describe the degree of urbanization.
dummy
importantly, four
the county itself).
Most
variables for the degree of urbanization of the region (not necessarily
These are major metropolitan areas (urban), regions with some urbanization
(moderately urban), rural region not on the periphery (moderately rural) and rural region on the
periphery (rural).
12
We
also use a
city (this variable is called
kilometer.
Finally,
we
dummy
"KS," for the
variable for whether the county
German
is
an independent
"Kreisstadt") and the population per square
include the travel time by car (in minutes) to the nearest major
metropolitan area.
For the western counties more variables are available.
skilled
workers
(at least college)
We
young adults
in the population, as this
only variable
we have
of the average manufacturing
A
4.
is
also intended to include a
group represents most of the perpetrators. However, the
the fraction age 15 to 49.
wage (manufacturing
12
Finally,
payroll divided
we
include the logarithm
by employment).
patterns of crime against foreigners. Table
of the main variables in our dataset.
Most of
1
provides descriptive
the foreigners are concentrated in the
western part of Germany. The share of the population that
percent
measure of the proportion of
Descriptive Analysis
We first examine basic
statistics
use the fractions of high
and of low skilled workers (no qualification beyond secondary
school, such as an apprenticeship).
available
We
is
foreign in the east
is less
In both parts of the country this share has been increasing since unification.
These are the four "siedlungsstrukturelle Regionstypen" of the BfLR.
15
than one
The
total
number of
incidents of crime against foreigners in our dataset
crimes against foreigners occurred in the west, but
quarter as
many people
100,000 population
~
as the
that is
west This
more than
it
must be recalled
leads to an incident rate
—
the
three times as high in the east
of foreigners living in the east, the number of incidents per foreigner
almost
fifty
times higher in the east than in the west
twice as high in the east than in the west.
understates the true
1,056.
in the east
More
that the east has only a
number of
incidents per
With such a low share
is
even more skewed
The mean unemployment
unemployment
In addition, the eastern
employment conditions
is
~
rate is about
rate
probably
because of large scale government
programs.
The
Map
1
regional dispersion in these variables can be seen
shows
more
clearly in
1
through
4.
the rate of incidence of crimes against foreigners per 100,000 residents. There are
numerous white spots on the map, indicating counties without incidents
eastern
Maps
Germany,
the crime rates are not only higher
in our data.
Within
on average but the counties with high
rates
of crime tend to be clustered along the coast in the north and along the eastern border with
Poland.
Map
2 contains the number anti-foreign crimes per 1,000 foreigners, displays a similar
pattern.
The unemployment
rates in various regions are
outlines the eastern and western parts of the country.
differences in
shown
in
Map
Nevertheless,
3.
This
it is
map
essentially
clear that there are
economic conditions within the eastern and western regions: areas of Saxony,
Brandenburg, and Sachsen-Anhalt are doing better than other eastern counties, while in the west,
the northern
coast,
Ruhr, and Saar and Pfalz areas have relatively high unemployment.
Map 4 shows
the density of foreigners.
16
The
foreign population
is
clustered in the
southwestern state of Baden- Wiirttemberg (home of
much
of the machinetool and car industry),
along the Rhine valley and in the Ruhr area (where west Germany's heavy industry
This pattern
is
located).
when workers where needed
the result of guest worker recruitment in the 1960s,
primarily in the metal manufacturing sectors.
is
In the east, only Berlin has a high concentration
of immigrants.
The
east and the west differ in the national composition of immigrants, as well as in their
absolute numbers.
many of whom
The dominant groups
are guest workers
in the
west are Turks, Yugoslavs,
who have become
from abroad. Table 2 shows
the victims.
communist
13
In the east,
46%
who make up much
of the victims
come from
of the recent
Eastern Europe and about
the west, Turks are the largest group of victims (40%), slightly
foreign population.
The
that these patterns are also reflected in the nationality of
former communist) developing countries. Only
(or
and Greeks,
long-term or permanent residents.
foreigners in the east are predominandy eastern Europeans,
arrivals
Italians,
Other southern Europeans make up
8%
of the victims are Turks. In
more than
10%
22% from
their share
among
the
of the victims, and eastern
Europeans make up 15%. Africans and Asians, who make up a small fraction of foreigners
the population, are clearly over represented
among
the victims.
Table 2 indicates that foreigners do not randomly
that this type of
crime
is
fall
disproportionate directed at groups
terms of color, race, or religion, or
who
are
among
in
the
victim to anti-foreign attacks, but
who
differ
most recent
more from Germans
in
Moreover, there
is
arrivals.
a higher incidence of attacks against asylum seekers than against foreign workers in general.
13
We
know
the ethnicity of victims in only 378 out of the 1,056 incidents.
17
Fully
were
54%
of
incidents reported in the east
all
were attacks against asylum seekers, and
43%
classified as such in the west.
Regression Analysis of Regional Patterns of Crime
5.
We
relate the incidence of
characteristics of the county
crime against foreigners to measures of the socioeconomic
where
the incident took place and other variables.
Several
specifications and estimation techniques are used.
Table 3 presents basic
OLS
regressions for the pooled sample of
using incidents per 100,000 residents or incidents per
variable.
14
The model
column
in
unemployment
(1) includes as explanatory variables: the
However,
country.
this coefficient
Once we
column
in
rate has a positive
(2),
is
is
associated with
the
raw difference
dummy
unemployment
no evidence
rate,
In these regressions
t-ratio
of about
5.
variable that equals one for counties located in the west
rate
has a statistically insignificant and negative effect.
that within the
more crime. Furthermore,
in the
unemployment
and highly significant coefficient with a
Furthermore, the coefficient on the foreign share rises substantially.
hypothesis, there
counties
merely reflects the higher level of crime in the eastern part of the
also include a
the
German
1,000 foreigners as the dependent
the foreign share and a variety of indicators of the degree of urbanization.
the
all
crime incidence
rate
Contrary to Alber's
two parts of the country higher unemployment
the coefficient
on the west
dummy
is
larger than
between the east and the west (3.9 versus
2.2).
%
14
Notice that there are two counties in the east with less than 0.1
of foreigners. Because
these appear as zeros in our data we could not form the ratio of incidents to the number of
foreigners.
These two counties are omitted from the
foreigners.
18
set of regressions using incidents per
1
,000
Column
(3)
adds two distance variables as regressors: the distance from the western
The
border for the eastern counties and the travel time by car to the next metropolitan area.
distance to the west has a highly significant and large coefficient
km
implies that for every 100
of distance from the western border the number of incidents increases by 2.5 per 100,000
This
residents.
the
It
R
2
is
much
larger than the
increases substantially.
However,
this effect is
The
last three
mean of
15
Note also
Distance from a metropolitan area also increases the crime
that
rate.
hard to interpret given the other urbanization variables in the regression.
columns of Table 3 use the number of incidents per 1,000 foreigners as
dependent variable. The basic pattern of
the results tend to be
the dependent variable (1.3).
somewhat
association between the
less
results is the
same
as in
columns
pronounced. Nevertheless, there
unemployment
is
(1)
through
(3),
the
but
no evidence of a positive
We
rate victimization rate within the regions.
also
entered the foreign share of the population as an explanatory variable in these regressions. The
positive coefficient on this variable implies that, within regions, the victimization rate tends to
be higher in areas with relatively more foreigners.
In spite of our earlier comparison of the Zeitung and police data, one
whether the Zeitung data are biased
in
First, all
to
be
15
We
still
question
any particular direction. For example, the newspapers and
wire services tend to be located in major
than other areas.
may
cities,
and these
cities
may be
covered more thoroughly
have two additional pieces of evidence which make
this
seem
unlikely.
our newspapers are located in the west whereas almost half the incidents are reported
in the
east Furthermore, within the east most events took place
Such a pattern
away from
the west
in rural areas,
which we
is unlikely to result from reporting bias. More incidents are reported
where most of the newspapers are located.
19
further
would expect
to
be covered less comprehensively.
dummy
included a
variable for the presence of a
county in our regressions.
this
dummy was
In
16
(dpa
Germany's
is
of foreigners, and urbanization,
in
Table
we
in
Columns
4.
free
(1)
for the east and the west; the other
Agency
news
service.)
(dpa) office in each
The
coefficient on
crime and victimization rates in the east and west and
rate,
share
up the structure and run separate regressions for the east
and
and
(2),
(5)
and
(6) report
columns add regressors
Pronounced differences
the country.
largest wire
Press
account for these differences with variables such as the unemployment
to
and the west
German
we
small and insignificant.
view of the large differences
our inability
Secondly, following Danzger (1975),
comparable specifications
that are only available for parts of
exists in the correlates of violence in the
two
parts of the
country.
The
coefficient
on
marginally significant in
more
all specifications.
measures,
results for the
is
from below
in Frankfurt,
it
17
2%
The
that a
in
coefficient
1
even more negative and
on percent foreigners
residents.
17
No
regressor, apart
from the urbanization
Despite the fact that the percentage of foreigners varies
North Sea coast
has an insignificant influence on the number of crimes per resident
columns
(1)
and
rises to 2 or
percentage point increase in the foreign share
list
to
25%
Column
(4)
of dpa offices to us.
(2) yields an F-statistic
level.
20
An
Chow-test for the equality of all
of 13.1 which is significant far below the
differences are statistically highly significant.
coefficients in
the east is
in rural areas of eastern Bavaria or near the
Martin Wortmann provided the
The
rate
west are quite different.
individually significant
substantially
16
meaning
two more attacks per 100,000
The
%
unemployment
in the crime rate regressions
result in
1
the
adds additional variables, including qualifications, the fraction of adults between 15 and 49, and
None of
the manufacturing wage.
Urbanization
is
these variables
the only strong predictor of anti-foreign violence in the west: crime rates
are highest in the major metropolitan areas.
directly since the population density,
all
is statistically significant.
dummies
capture effects related to urbanization.
for degree of urbanization
we
residents than a rural area,
The
If
coefficients in table
for urbanization
we
is
only use the four exclusive
crime against foreigners
major
cities.
is
more
is
dummy
variables
incidents per 100,000
a large difference given an overall
dependent variable. Interestingly, the pattern by urbanization
there,
difficult to interpret
and city-counties, and travel time
find that a major center has about 0.5
which
4 are
mean of
0.9 for the
exactly the reverse in the east:
concentrated in rural areas which have 4.7
more
incidents than
Table 5 reports corresponding results using only serious crimes (homicide and arson) as
the dependent variable.
total crimes.
The
The
coefficients tend to be uniformly lower
In all other respects, the qualitative results differ very
due
to the
little
lower number of
from those
regressions in the previous three tables neglect the fact that there
number of counties
that
had no incidents of crime against foreigners
censoring of the data
may
bias the
to address this
problem.
First,
OLS
results.
is
in
Table
4.
a substantial
in our dataset.
This
We tried several alternative estimation strategies
and most directly comparable
Tobit models allowing for a censored dependent variable
at 0.
to the
OLS
models,
we
estimated
This yielded very similar results
for both the incidence rate and victimization rate.
An
alternative approach is to use the
rather than scaling
it
number of crimes
directly as the
by the population or the number of foreigners. In
21
dependent variable,
this case, the
model
is
most sensibly
treated as a count data
therefore nests the Poisson model.
we
is
influence on the crime rate.
expected
if
model.
18
is
an
is
a
Gamma-mixture of Poissons and
1%
this variable
has a coefficient insignificantly
population only measures scale but has no other
Otherwise, the results for both the Poisson and negative binomial
specifications are similar to those obtained
we used
each county
also add the log of the population in the county as a regressor to
from one, which
Finally,
in
In each case, the Poisson specification is rejected at the
account for scale. In the negative binomial results
different
number of crimes
The negative binomial
of our preferred specifications.
In these models
the
Table 6 presents both a Poisson and a negative binomial regression
integer and typically small.
level.
model since
the exact date
The Weibull hazard has
from the
when an
OLS
models
in
Table
4.
incident occurred to estimate a Weibull duration
form
the
= ata exp(X'B),
'1
h(t)
where
h(t) is the probability
variables,
and
a
of an incident occurring
and B are parameters
event in a county are treated as
treated as right censored.
parameter
a
is less
The
left
to
be estimated.
at date
t,
X
a vector of explanatory
Spells from January 1991 until the
first
censored; spells from the last event until June 1993 are
results are quite similar to the
than one indicating that the hazard rate
is
count data models.
decreasing.
tend to be clustered in time, corresponding to the aggregate pattern
18
is
shown
The shape
Attacks in a county
in Figure 2.
54 cases we only know the month but not the exact date of an event These events
were allocated to a random date in that month. Spells between two events on the same day were
In
set to 0.1.
22
Discussion and Relation to the Literature
6.
Our
to differ in
analysis supports several tentative conclusions. First, anti-foreigner violence appears
important respects between east and west Germany.
foreign crime
is
much
greater in the east
In the east, anti-foreign crime
In the west, anti-foreign crime
rural areas.
The incidence
is
more common
in
is
rate of anti-
more common
in the
urban areas. Within the east
we
find that the incidence of crime against foreigners per capita rises with the fraction of foreigners
in the county, but there is
Second, contrary
of anti-foreign crime
is
no relationship between these variables
to
in the
our expectations and Alber's hypothesis,
unrelated to the
unemployment
rate in
we
west
find that the incidence
an area. This finding
however, with police records on suspects and survey studies of youths.
(1994), for example,
is
not
much
fraction of
1994).
1
1.3% of suspects investigated from 1991
greater than the
unemployment
young, more than
may
60%
are age
20 or below, so
Surveys of youths also yield a comparable
not a predictor for violent tendencies
unrelated to the
be
much
to
1993 were unemployed. This
Furthermore, the
higher in the east (see
e.g.
Raisch,
that a significant fraction are
result, that difficult
among
still
in school.
personal economic conditions are
those surveyed (see, e.g., Schnabel, 1993).
and time-series analyses of crime in the U.S. also find crime
unemployment
BMI
be explained by the fact that the perpetrators are typically very
partly
Interestingly, regional
to
consistent,
According
rate for the country as a whole.
unemployed perpetrators does not seem
This finding
to
is
rate.
to be
For example, Land, McCall, and Cohen, (1990) report
typical cross-sectional studies at the state,
SMSA,
or city level for the U.S. have found either no
or a weak, negative relationship between crime rates and the local unemployment
23
that
rate.
Third, the level of violence in the east
the western border, other things being equal.
this finding.
First,
because
it is
easier to
is
higher in counties that are located further from
There are
commute
economic opportunities are greater for commuters,
Second, the far eastern states
law and order. The variable
The
may
may
to the
at least
west
if
this variable
may
reflect
be more insulated and less exposed
was
significant after
we
is
near the west, and
economic conditions.
to foreigners
and western
controlled for travel time to a major
metropolitan area, or distance to the nearest county with under a
7.
the county
for
thus reflect differences in societal values and legal institutions.
fact that the distance variable
some weak support
two plausible explanations
5% unemployment rate, provides
for the latter hypothesis.
Conclusion
A wave of hostility and violence against foreigners has shaken Germany since unification
While the number of such crimes had increased already
in 1990.
tenfold increase from 1990 to 1991 alone.
attacks, that
make
this a
It is this
this rise in violence,
uncovered a number of important patterns. The
part of the country despite the fact that very
in particular
them
to
We
the poor
many
While social
rates of violence
scientists
have been higher
few foreigners actually
live there.
target of right
have yet
to
in the eastern
Foreigners, and
wing propaganda, alleging
of Germany's social problems.
feel relatively confident in ruling out a direct connection
economic conditions
a
our analysis, as well as previous work, has
asylum seekers, had been singled out as the
be responsible for
was
increase, and the political nature of the
problem of particular concern.
advance a coherent theory for
in the late 1980s, there
in the
east
In particular,
24
we
find
between the violence and
no connection between the
incidence of crime and the unemployment rate or the
indicating that personal
is
economic
also consistent with findings
wage
rate.
This confirms survey results
strain alone is not a strong predictor of violent tendencies.
on crime more generally. This does not mean
conditions in the east have no impact on crime against foreigners at
likely to be
more
Lastly,
the west.
that
conflicts.
is
in the east feel
But not
more
west
inclined towards violence as a
just the levels differ, the patterns differ as well.
west Once
Interestingly,
the rising
we do
wave of
it is
means of solving
Crime
show
social
against foreigners in
mostly concentrated
in metropolitan
not find any other important covariates explaining crime
violence and the surrounding political climate seemed to
generate enough legitimacy for this type of behavior,
specific local conditions.
between the east and
in the east is consistent with survey findings that
the east is mostly concentrated in rural areas, whereas
in the
but the connection
clear differences in anti-foreign crime
The higher incidence of crime
areas in the
economic
subtle.
we document some
many youth
all,
that the
It
it
broke out uniformly, independent of
This makes preventing and fighting such crime a difficult task, and
continues to provide a challenge to social scientists to identify the causes more clearly.
25
References
"Towards Explaining Anti-Foreigner Violence
Center for European Studies, Harvard University
Alber, Jens (1994)
in
Germany."
WP No.
53,
fur Landeskunde und Raumordnung (BfLR) (1992)
Laufende
Raumbeobachtung: Aktuelle Paten zur Entwicklung der Stadte, Kreise und Gemeinden
1989/90, Materialen zur Raumentwicklung Heft 47, Bonn: BfLR
Bundesforschungsanstalt
Bundesministerium des Inneren (BMI) (1994) Verfassungsschutzbericht 1993 Bonn:
,
BMI
Danzger, M. Herbert (1975) "Validating conflict data," American Sociological Review 40, 570-
584
Heitmeyer,
W.
(1992)
Die Bielefelder Rechtsextremismus Studie: Erste Langzeituntersuchung
zur Sozialisation mannlicher Jugendlicher,
Inverarity,
James M. (1976)
Weinheim/Munchen
"Populism and lynching
in
Lousiana,
1889-1896:
A
test
of
Erickson's theory of the relationship between boundary crises and repressive justice,"
American Sociological Review 41, 262-280
Hans-Helmuth "Die Entwicklung des Rechtsextremismus in Deutschland. Historische,
und psychologische Bedingungen seines Entstehens," in: BMI (ed.)
Extremismus und Gewalt, vol. 1, Bonn: BMI, 7-34
Kniitter,
gesellschaftliche
C,
and Lawrence E. Cohen (1990) "Structural covariates of
homicide rates: Are there any invariances across time and social space?" American
Journal of Sociology 95, 922-963
Land, Kenneth
Patricia L. McCall,
Leenen, Wolf Rainer (1992) "Auslanderfeindlichkeit in Deutschland.
oder Politikversagen?" Deutschland Archiv 25, 1039-1054
Politischer Rechtsruck
Ohlemacher, Thomas (1994) "Public opinion and violence against foreigners
Germany," Zeitschrift fur Soziologie 23, 222-236
in the reunified
Olzak, Susan (1989) "Labor unrest, immigration, and ethnic conflict in urban America, 18801914," American Journal of Sociology 94, 1303-1333
Raisch, Peter (1994)
"Bekampfung
rechtsorientierter Gewalttaten
im
der Sonderkommission Rechtsextremismus (Soko 'Rex') des
Hoffmann
(ed.)
Fremdenfeindlichkeit in Deutschland
,
Padagogik, Universitat der Bundeswehr, Miinchen, 61-75
26
Freistaat Sachsen
LKA
Sachsen,"
Symposium der
-
in:
Einsatz
Alfred
Fakultat fur
Sippel, Heinrich (1993)
(ed.)
"Rechtsextreraismus in Deutschland.
Extremismus und Gewalt,
Schnabel, Kai
Uwe
(1993)
vol.
1,
Eine Lagedarstellung,"
in:
"Auslanderfeindlichkeit bei Jugendlichen in Deutschland.
Synopse empirischer Befunde
BMI
Bonn: BMI, 35-52
seit 1990," Zeitschrift fur
27
Padagogik 39, 799-822
Eine
Map
Number
1
of Incidents of Crime Against Foreigners
per
1
00,000 Residents, County Level
Legend
3.75 to 34
1.75 to 3.75
1
1.75
to
0.15 to 1
0.0
(81)
(78)
(76)
(92)
(216)
Map
Number
2
of Incidents of Crime Against Foreigners
per
1
,000 Foreigners, County Level
Legend
6
0.5
0.15
0.001
0.0
to
to
to
to
to
339
6
0.5
0.15
0.001
(82)
(79)
(80)
(86)
(216)
Map
3
Unemployment Rate
by County (Sept. 1992)
Legend
14.2 to 19
8.2 to 14.2
5.2 to 8.2
(139)
5.2
(134)
2.5 to
(133)
(137)
Map 4
Foreigners as a Percent
of
County Population
Legend
D
7.1 to 25.7
(136)
3.2 to
0.5 to
7.1
(136)
(141)
0.1 to
3.2
0.5
0.0 to
0.1
(2)
(128)
in
1991
oieiqoJd
Jofe|/\|
sjeu6|ejoj
6uj>iu!lji 'aied
CD
X
CD
"O
_C
CD
"D
13
I
CO
CD
E
O
JD
E
L|juo|/\|
J9d S9OIU0 jo jaqoinN
bjbq 6uni!9Z
-co
(2>
CT>
1
1
1.
03
05
Q
1
c
-
"CD
N
-CM
o>
03
05
Q
-
-
bibq eo||Od
O
i
en
c_
CD
c:
en on
i—
en
CD
C
O
CD
cn
cn
nq
m
CD
a
CD
c_
Z3
CJ
cn
o
en
CD
-rH
CJ
CD
CD
CJ
C_
Z3
O
cn
CD
_0
B}BQ 3UJIJ3 3DIt0d
Figure 4
Type
Crime Against Foreigners
Police Data and Zeitung Data
of
Police Data
Mu
n
61
^
er
'
Arson
8.9%
Assault
9.6%
Other
81.1%
Zeitung Data
Murder
8.7%
Other
37.3%
Arson
34.6%
Assault
19.4%
Table
1
Descriptive Statistics*
West
East
Percent of Population That
All
is
Foreign
1989
-
8.2%
1991
0.68%
8.8%
7.3%
1992
0.87%
9.5%
7.9%
Number
1991
-
Number
1991
-
of Serious Zeitung Incidents (Arson
236
1993:6
Number
1991
-
1993:6
-
&
Murder)
415
3.13
1993:6
1,056
651
of Zeitung Incidents/100,000 Population
Number
1991
594
462
1993:6
of Zeitung Incidents
0.91
1.32
of Zeitung Incidents/1,000 Foreigners'"
0.10
4.63
0.18
Unemployment Rate
14.3%
1992
Notes:
a.
Number
of incidents
variables are from
b.
Foreigners
is
is
6.7%
derived from newspaper clippings.
BfLR.
as of 1991.
8.1%
All other
Table 2
Number
of Crimes by Nationality of Victim and Foreign Shares
Crimes
(column percent
Origin Shares
in parentheses)
East
West
Total
East
West
59 (46)
37 (15)
96 (25)
57.7
8.6
Polish
16 (12)
15 (6)
31 (8)
21.3
4.3
Romanian
21(16)
11(4)
32(8)
8.5
1.5
former Soviet Union
20(16)
6(2)
26(7)
11.6
0.7
(0)
9.3
0.8
Nationality
Eastern European
Hungarian
1(1)
(0)
other
2 (2)
5 (2)
6 (2)
7.0
1.4
Turkish
10(8)
100(40)
110(29)
0.8
30.8
Southern European
3 (2)
26 (10)
(8)
2.8
32.9
former Yugoslav
3 (2)
14 (6)
17 (5)
2.0
13.4
(0)
12 (5)
12 (3)
0.7
19.5
28 (22)
36 (14)
64 (17)
23.7
9.2
20 (16)
8 (3)
8(6)
28(11)
36(10)
19(15)
33 (13)
52 (14)
7.9
4.0
Latin American
6(5)
3(1)
9(2)
2.8
1.0
Western Countries
0(0)
(0)
3.1
12.4
German
4(3)
13(5)
17(5)
—
—
129 (100)
249 (100)
378 (100)
100.0
100.0
other
Asian
Vietnamese
others
African
Total
1
(0)
1
29
28
1
(7)
Table 3
Basic Relative Anti-foreign Crime Regressions for
all
German Counties
(standard errors in parentheses)
Incidents per 100,000 Residents
Incidents per 1,000
Foreigners
Unemployment Rate
West
Km.
to
Western Border
(1)
(2)
(3)
(4)
(5)
(6)
.221
-.096
-.102
1.068
-.276
-.220
(.043)
(.071)
(.069)
(.273)
(.463)
(.466)
-3.861
-.944
(.702)
(.778)
—
.025
—
—
—
—
-16.320
-8.988
(4.564)
(5.285)
—
.062
(.003)
Percent Foreign
Log Population Density
Moderately Urban
Moderately Rural
Rural
KS
Travel
Time
to
Metro
(.023)
.013
.085
.091
.482
.787
.768
(.054)
(.054)
(.051)
(.343)
(.350)
(.350)
-.819
-.664
-.502
-8.479
-7.845
-7.599
(.315)
(.308)
(.293)
(2.015)
(2.001)
(1.996)
-.135
.004
-.251
-2.965
-2.367
-2.034
(.374)
(.365)
(.373)
(2.387)
(2.367)
(2.540)
.658
.597
.132
4.258
3.981
3.843
(.541)
(-527)
(.525)
(3.458)
(3.421)
(3.569)
.623
1.062
.380
.928
2.741
2.380
(.532)
(.524)
(.538)
(3.404)
(3.405)
(3.657)
2.384
2.094
1.737
12.661
12.481
11.755
(.716)
(.699)
(.666)
(4.581)
(4.543)
(4.529)
—
—
.011
—
—
Area (minutes)
.001
(.027)
(.004)
R2
.146
.191
.272
.121
.142
.153
N
543
543
543
541
541
541
"This variable equals the distance from center of the county to the western border ignoring
West
Berlin,
and
for western counties.
Table 4
Basic Relative Anti-foreign Crime Regressions for East and West
Germany
(standard errors in parentheses)
Incidents per 100,000 Residents
Unemployment Rate
Km.
to
Western Border
Incidents per 1,000 Foreigilers
East
West
East
West
East
West
East
West
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-.394
-.052
-.394
-.032
-3.277
.0002
-3.318
.003
(.171)
(.039)
(.165)
(0.42)
(1.194)
(.008)
(1.186)
(.008)
—
—
.020
—
—
—
.071
—
(.037)
(.005)
Percent Foreign
Log Population Density
Moderately Urban
Moderately Rural
Rural
KS
Travel time to Metro
Area
Fraction
Low
Qualifications
Fraction High
Qualifications
Fraction
Young Adults
Log Mfg. Wage
2.933
.027
1.982
-.030
-13.532
-.013
-17.114
-.014
(.989)
(.027)
(.987)
(.031)
(6.893)
(.005)
(7.101)
(.006)
-1.894
.054
-1.447
-.047
-10.095
-.035
-8.665
-.050
(.766)
(.172)
(.749)
(.179)
(5.330)
(.035)
(5.348)
(.036)
-.172
-.595
-.049
-.617
-4.813
-.115
-4.428
-.118
(.891)
(.184)
(.862)
(.186)
(6.212)
(.037)
(6.176)
(.038)
2.684
-.783
2.417
-.720
19.926
-.160
18.731
-.150
(1.445)
(.245)
(1.399)
(-247)
(10.061)
(.049)
(10.016)
(.050)
1.788
-.858
1.448
-.825
5.905
-.115
4.450
-.109
(1.292)
(.272)
(1.252)
(2.75)
(8.993)
(.055)
(8.968)
(.056)
4.652
.545
3.710
.589
21.938
.127
19.009
.130
(2.190)
(.308)
(2.131)
(.316)
(15.240)
(.062)
(15.220)
(.064)
.015
.005
.019
.005
.042
.001
.060
.001
(.008)
(.002)
(.008)
(.002)
(.060)
(.0004)
(.060)
(.0004)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
-.037
(.022)
-.004
(.054)
.020
(.042)
.444
-.006
(.004)
.00001
(.011)
.0001
(.009)
.089
(.138)
(.681)
R2
.199
.088
.254
.104
.173
.087
.188
0.099
N
215
328
215
328
213
328
213
328
Table 5
OLS
Regressions for Serious (Murder and Arson) Crime Rates
(standard errors in parentheses)
Incidents per 100,000 Residents
Unemployment Rate
Km.
to
Western Border
Incidents per 1,000 Foreigi lers
East
West
East
West
East
West
East
West
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-.188
-.044
-.188
-.036
-1.704
-.001
-1.727
-.001
(.093)
(.028)
(.089)
(.030)
(.577)
(.006)
(.572)
(.006)
—
—
.011
—
—
—
.040
—
(.003)
Percent Foreign
Log Population Density
Moderately Urban
Moderately Rural
Rural
KS
Travel time to Metro
Area
Fraction
Low
Qualifications
Fraction High
Qualifications
Fraction
Young Adults
Log Mfg. Wage
(.018)
1.268
-.023
.735
-.023
-6.575
-.009
-8.607
-.010
(.536)
(.018)
(.533)
(.023)
(3.335)
(.004)
(3.424)
(.005)
-.570
.109
-.319
.062
-3.378
-.016
-2.567
-.019
(.415)
(.123)
(.405)
(.129)
(2.578)
(.025)
(2.579)
(.026)
-.042
-.401
.028
-.397
-1.118
-.082
-.900
-.081
(.482)
(.132)
(.466)
(.134)
(3.005)
(.027)
(2.978)
(.027)
1.699
-.467
1.549
-.436
9.839
-.102
9.161
-.100
(.783)
(.175)
(.756)
(.178)
(4.867)
(.035)
(4.829)
(.036)
.958
-.576
.768
-.564
4.698
-.072
3.873
-.071
(.700)
(.195)
(.676)
(.198)
(4.351)
(.039)
(4.324)
(.040)
1.983
.295
1.454
.300
7.257
.069
5.594
.064
(1.186)
(.220)
(1.151)
(.227)
(7.373)
(.045)
(7.339)
(.046)
0.008
.004
0.010
.004
.023
.0006
.034
.0006
(0.005)
(.002)
(0.004)
(.002)
(.029)
(.0003)
(.029)
(.0003)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
-.011
(.016)
.026
(.039)
.004
(.030)
.036
N
(.003)
.006
(.008)
-.003
(.006)
-.001
(.099)
(.490)
R2
-.001
.162
.090
.225
.096
.155
.075
.176
.078
2.5
328
215
328
213
328
213
328
MIT LIBRARIES
3 9080 00940 0158
Table 6
Alternative
Number
Maximum
Likelihood Models for the Total
of Incidents of
Crime Against Foreigners
(standard errors in parentheses)
Poisson
Negative
Binomia
Unemployment Rate
Km.
to
Border
Log Population
Log Population Density
Moderately Urban
Moderately Rural
Rural
KS
Travel
Time
to
Metro
Area (minutes)
a
(Weibull shape
Model
I
East
West
East
West
East
West
(1)
(2)
(3)
(4)
(5)
(6)
-.117
-.038
-.121
-.052
-.132
-.051
(.028)
(.027)
(.041)
(.043)
(.034)
(.033)
.004
—
.004
—
.004
—
(.001)
Percent Foreign
Duration
i
(.001)
(.001)
.320
-.019
.346
.019
.280
-.006
(.132)
(.016)
(.222)
(.025)
(.153)
(.018)
.772
1.175
1.103
1.164
.708
.969
(.122)
(-071)
(.216)
(.144)
(.145)
(.087)
-.355
-.007
-.532
.009
-.344
.003
(.136)
(.120)
(.213)
(.187)
(.164)
(.152)
.222
-.517
.138
-.592
.295
-.708
(.149)
(.130)
(.231)
(.181)
(.188)
(.166)
.440
-.763
.386
-.846
.521
-1.368
(.229)
(.212)
(.349)
(.276)
(.283)
(.324)
.685
-1.002
.566
-1.009
.848
-1.761
(.206)
(.250)
(.307)
(.317)
(.251)
(.418)
1.037
.543
1.295
.594
.924
.291
(.343)
(.207)
(-549)
(.349)
(-421)
(.257)
.003
.007
.003
.007
.003
.008
(.001)
(.001)
(.002)
(.002)
(.002)
(.002)
—
—
—
—
.437
.473
(.019)
(.017)
parameter)
Pseudo
N
R2
0.225
.379
.097
.143
.105
.232
215
328
215
328
677
922
7966
Date Due
Lib-26-67