IS CRIME SEASONAL?

January 1984

by Carolyn Rebecca Block

Statistical Analysis Center

ILLINOIS CRIMINAL JUSTICE INFORMATION AUTHORITY

William Gould, Chairman
J. David Coldren, Executive Director

U.S. Department of Justice 92936

National Institute of Justice
This document has been reproduced exactly as received from the
person or organization originating it. Points of view or opinions stated
in this document are those of the authors and do not necessarily
represent the official position or policies of the National Institute of
Justice.

Permission to reproduce this copyrighted material has been

granted by
Illinois Criminal Justice Information
Authority
to the National Criminal Justice Reference Service (NCJRS).

Further reproduction outside of the NCJRS system requires permis-

sion of Ihe copyright owner.
 • !." l

Printed by authority of the State of Illinois

January 1984
Number of copies: 300
Printing order number: 84-21
 .QQNTENTS

ACKNOWLEDGEMENTS ,.,.. t •••••••••••••••••••••• ~I •••••• iii

•••••••••• •

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . 1
ISSUES IN THE SEASONALITY OF CRIME • . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Length of Series ............................................ 3
Local versus State or National .............................. 6
Does Crime Occur Seasonally, or Is It Reported Seasonally? .. 9
IS CRIME SEASONAL? . . . . . . . . . . . . . . . . . . . . . . . . . . • • . . . . . . . . . . . . . . . . 15
Is Violent Crime Seasonal? , ................................ 16
Homic ide ............. , ................................... 16
Forcible Rape ........................................... 20
Robbery ................................................. 21
Aggravateu Assault ...................................... 23
Is Property Crime Seasonal? ................................ 25
Burglary ............ , ................................... 25
J"arceny /Thei't ........................................... 25
Motor' Vehicle Theft ..................................... 28
Sttnllllary .. , . . . . . . . . . . . . . . . . . . . . . . . , . . . . . . ,. . . . . . . . . . . . . . . . . . . . . . . . 29
REVIEW BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

i
ii
 ACKNOWLEDGEMENTS

The analysis in this report was supported, j n part, by a

grant from the Bureau of Justice Statistics, IITime Series Pattern
Description for Criminal Justice Decision Makers.II

Eal~ly versions were presented at a Workshop on Law and Jus-

tice Statistics sponsored by the American Statistical Association
(August, 1983) and at the Max Planck Institut, Freiburg, Germany
(October, 1983). Comments and questions by members of these
audiences were helpful and encouraging.
The report will be included in a volume of the proceedings
of the Workshop, published by the Bureau of Justice Statistics,
to appear in 1984.
Statistical Analysis Center publications require the joint
efforts of a number of staff members, and this report is no ex-
ception. James "Chip" Coldren, Larry V. Dykstra, Paul Fields,
Chris Devitt, and Scot.t Levin read and commented on early drafts.
'l1he substantive comments of Louise S. Miller and the formatting
and organizational suggestions of Kevin Morison were especially
helpful. Louise was responsible for the re-creation of the
Chicago assault data set from two partially complete sources.
This report summarizes the analysis of many data sets ob-
tained over the years through the help and cooperation of
numerous people. A eomplete list is found in the introduction to
the section, "Is Crime Seasonal?"
The collection of the Chicago homicide data has continu.ed
over many years, and would not have been possible without the
help and cooperation of the Chicago Police Department, especially
for'mer Superintendent Richard J. Brzeczek, Commander Rudolph E.
Nimocks, former Acting Superintendent (now retired) Michael
Spiotto, Deputy Superintendent Thomas J. Lyons, Chief William
Hanhardt, Lieutenant Michael Cacci tola, and Sergeants Timothy
Tidmarsh, Patrick Conway and Michael Provenzale. Collection of
1965 to 1978 data was supported in part by PHS Research Grant No.
lROlM27575, NIMH, to the Center for Studies in Criminal Justice;
Franklin E. Zimring, director of the Center, and Richard L.
Block, of Loyola University, directed data collection and
analysis at t.his stag(~. Ifhe 1979 to 1981 data were coded by
Keith Coopridel' of the Authority, under the direction of Carolyn
Hebecca and Richc:u'd Il. Bloele Time series files are maintained
by the Authority. Por a list of publications using these data,
contact the author of this report.

iii

Preceding page blank

 The time series files of Canadian homicide were created by
Craig McKie of Statistics Canada, with the assistance of Bill
0' Grady, who "'TaS on a work assignment as a Carleton University
graduate student. Canadian homicide data are collected centrally
by the Canadian Centre for Justice Statistics (and prior to 1981,
by its predecessor, the Justice Statistics Division of Statistics
Canada), and we are grateful for the help and advice of Doug
Brown and Geoff Kubrick of the Centre for Justice Statistics.

We would also like to acknowledge, with thanks, the efforts

of Kenneth W. Candell, Vicki Maj or and Chris Waskiewicz of the
Uniform Crime Reporting Program, United States Department of
Justice, who provided the 1973 Illinois supplementary homicide
report data and the 1970 through 1982 United states Index crime
data, and have been very patient and generous with their time
over the years in answering questions and giving help and advice.

Lily E. Christ of John Jay College provided the New York

City series. Susan Williams of the Bureau of Criminal Statistics
and Special Services, California Department of Justice, provided
the California data, and was very helpful in answering questions
of interpretation.

The National Crime Survey (NCS) data were provided by

Richard L. Block from a users' file cf NCS data created by Block
and Wesley Skogan.

iv
 Researchers and policy makers often take for granted that
seasonal fluctuation in crime is &n established fact. To suggest
otherwise goes against the grain of a long tradition in criminol-
ogy. Indeed, Brearley (1932:161-199) begins his review of the
literature on criminal seasonality with Hippocrates, and Wolfgang
(1966) cites scholarly works dating from 1825.
Quete1et, a Belgian statistician and one of the earliest in-
vestigators of seasonal fluctuation :\.n crime (18112: 90; also see
Sylvester,1982) , states,
The seasons, in their course, exercise a very marked
influence: thus, during summer, the greatest number
of crimes against persons are committed and the
fewest against property; the contrary takes place
during the winter.
The assumption that crime occurs sea::;onally continues to be
made today. In heralding the IIWar on Crime II of the Johnson ad-
ministration, the President's Commission on Law Enforcement and
the Administration of Justice (196(:27) stated that
Murder is a seasonal offense. Rates are generally
higher in the summer, except for December, which is
often the highest month and almost always 5 to 20
percent above the yearly average. In December, 1963,
following the assassination of President Kennedy,
murders were below the yearly average by 11 percent,
one of the few years in the history of the UCR that
this occurt'ed.
The more recent B...§.Qort to the Nation on Crime and Justice (BJS,
1983: 11) states that, although personal larceny under $50 and
robbet'y are exceptions,
. . . almost all types of personal and household
crimes are more likely to occur during the warmer
months of the year.
One of the most influential basic criminology textbooks in the
United States, Sutherland and Cressey (1978:82), states that
statistical studies show very uniformly that crimes
against property reach a maximum in winter months,
and crjmes against the person and against morals in
the summer months.
The ~mswer to the qnestion, Is crime seasonal? 1/ is not as
II

straightforward as these quotes suggest. Some types of crimes

fluctuate with the seasons, while others do not. The same crime

1
 may show seasonal fluctuation in one geographic area, but not in
another. In addition, the decision as to whether or not a
particular series is seasonal depends upon the conceptual and
operational definition of seasonality the decision-maker uses.

This report first reviews some analytical issues that must

be considered in decisions regarding the seasonal fluctuation of
crime. Second, using analyses that the Statistical Analysis Cen-
ter has conducted over the years, the report answers the ques-
tion, "Is crime seasonal? II for specific crime types, geographic
areas, and time periods.

The report ends with a review of analyses, published and

unpublished, that address the question of seasonal fluctuation in
crime. Studies were not selected for inclusion on the basis of
methodology or any other quality criterion, but simply to provide
a comprehensive review of the existing literature. The biblio-
graphic review also includGs a summary of the findings of a
previously unpublished survey of seasonality analyses done by the
states.

I,

2
 ISSUES IN 'rHE; SEA$ON~LITY OF CRIME

The question, IIIs crime seasonal?" is not easily answered

with a yes or a no. The answer depends upon the statistical
criteria used to make the decision (see Block,1983). Methods of
seasonal analysis are not completely objective. Their results
must be interpreted and researchers using the same method may
j

come to differing conclusions. There is even disagreement among

statisticians on the very definition of seasonality.1
For the purpose of this report, we will use Kallek's (1978:
15) simple and straightforward definition of seasonality.
Seasonality refers to regular periodic fluctuations which
recur every year with about the same timing and with the
same intensity and which, most importantly, can be mea--
sured and removed from the time series under review.
Not all analysts concur with this definition. A case in
point is Warren, et al. (1981), who fo..tnd homic ide to have a
seasonal pattern that changes from year to year. That is, a
"peak month" in some years is a IItrough month" in other years.
The authors conclude that homicide is seasonal, but II inconsis-
tent. II Since year-to-year consistency is implied in Kallek' s
defini tion, if we accept that definition, II inconsistent sea-
sonalityll is a contradiction in terms.
This section reviews several methodological and analytical
issues that affect an investigator's decision as to seasonality.
These issues may explain the seemingly contradictory results of
some studies of seasonal fluctuation in crime.

Length of Series

Generally, a number of years of data are necessary in order

to answer the question, "Is this crime seasonal? with confi-
II

dence. A series shorter than seven years is considered too short

for a definite decision about the presence of seasonality (see
Block, 1983). The reason for this becomes clear if you consider
that, in one year, you observe one instance of each month. In
six years, you would have only six observations of Januaries, six
observations of Februaries, and so on.
With an increasing number of observations (years), seasonal
fluctuation can be described more accurately. A few extreme
values will have less effect on the total analysis in a long

lPor a complete discussion of alternative conceptual and

operational definitions, see Block (1983).

3
series than they will in a short series. Also, with a very short
series, only strong seasonal fluctuation is likely to produce
statistical significance. In general, the longer the series, the
more likely that relatively weak seasonal fluctuation (that is,
however, consistent over time) will be significant.

Table 1 shows the effect of length of series on a seasonal

analysis, using the same method (Census X-11) and the same sta-
tistical criteria (F of stable seasonality and percent contribu-
tions of the seasonal and irregular components over a one-month
span). The P value is an indicator of significance, while the
percent contribution of the seasonal component is analogous to a
measure of association. 2 The series, seven Index crimes in
Illinois, begin in 1972, but extend either to 1977 (six years) or
to 1981 (10 years).

Table 1

Census X-1I Results, Illinois Index Crimes

Six-Year Series Ten-Year Series

(1972-1977) (1972-1981)
Index Crime Stable % Contribution Stablel% Contribution
P Season. Irreg. P Season. Irreg.

Murder 1.5 30·5% 69.7% 2.8 29.0% 70.4%

Forcible Rape 8.1 27.2 72.2 10.0 20.5 79.0

Robbery 7.1 34.4 60.5 10.9 32.3 65.0
Agg. Assault 1~9 .5 63.1 35.7 45.7 59.6 38.5
Burglary 10.8 52.8 43.2 19.3 55.9 41. 7
Larceny/Theft 10 11.6 81. 2 17.2 95.8 80.7 18.2

Motor Veh. Theft 16.2 51. 9 46.8 13.0 39.1 59.6

2Por a more complete discussion, see Block (1983) and the

studies listed in its annotated bibliography. Por a brief review
of the statistics in Table 1, see the introduction to the Review
Bibliography, page 32 below.

------------------------~---.----.-~-----
 Because we cannot assume that observations in a time series
are independent, the stable seasonality F should be interpreted
as one indicator of the degree of seasonality, not as an exact
measure of significance. The II Plewes l"ule-of-thumb II (Block, 1983)
uses the irregular contribution as a means of interpreting the F
value. The Plewes criteria are: if the irregular contributes 30
percent or more of the total month-to-month variation, the deci-
sion should be "no stable seasonality, II regardless of the F
value; if the percent contribution is 25 to 29, the F value
needs to be at least 15, and if the percent contribution is 15 to
24, the F value needs to be at least 2.41 for the series to be
considered seasonal. An F value less than 2.41 indicates no
stable seasonality, regardless of the irregular contribution.

In four cases, increasing the length of the serieS increased

the F value. This happened for murder, forcible rape, robbery,
and burglary. However, the stable seasonality F values of ag-
gravated assault, larceny/theft, and motor vehicle theft were
lower in the longer series.

The percent contribution of the seasonal component over a

one~month span does not always increasE: when the length of the
series increases. For example, for forcible rape, the seasonal
contribution is 20.5 percent in the longer series, and 27.2 per-
cent in the shorter series, although the F is higher in the lon-
ger series. This is also true of murder and robbery. Why should
this be so?

It is, of course, possible that there is less seasonal fluc-

tuation in the 1978-1981 years than in the previous six years.
In that case, the addition of the 1978-1981 years to the series
would decrease the seasonal fluctuation overall. However, if
that were true, why would the F value increase? The answer to
this apparent contradiction is that a longer series allows a more
accurate description of seasonal activity. This more accurate
description tells us that the seasonal contribution is less.
Thus, it is possible that these violent crime series contain a
weak degree of seasonal fluctuation, and that this weak fluctua-
t:on might produce high enough F values to be considered seasonal
by the Plewes criteria in series that are even longer than these.

The literature of crime seasonality contains numerous ex-

amples of seasonal analyses based on very short series, sometimes
only one year (see the Review Bibliog't'aphy, below). It is not
surprising that different analysts, analyzing the same crime but
for varying time periods, would reach differing conclusions about
the presence of seasonal fluctuation in that crime.

5
 Local versus State or National

What is the appropriate level of geographic aggregation to

answer the question, II Is crime seasonal? II Should we look for an
answer at the local level, or at the state or national level? On
one hand, lIusing local data may lead to erroneous conclusions
about the seasonal pattern because of the small numbers involved
and the possibility of local intervention or prevention effortsll
(Michael and Zumpe, 1983). On the other hand, most administra-
tive and policy decisions in law enforcement are made at the lo-
cal level, and relate to local policy. If these decisions are
based on information about seasonality at the national level,
they may lead to erroneous local-level conclusions (Coldren,
1980).

Chart 1
SEASONAL FLUCTUATION OF AGGRAVATED ASSAULT
ILLINOIS. 1972 THROUGH 1961
CENSUS X-II ADDITIVt:. ADJUSTMENT

6
 The same phenomenon may vary seasonally in one place but not
in another. Some geographic areas, such as a college town, a
tourist mecca, or the home of the state fair, have an influx of
population during certain seasons of the year. In addition, if
the weather is an underlying cause of seasonal fluctuation, areas
with different climates may experience differing patterns of sea-
sonality. For examples, see the IISurvey of the states" (Review
Bibliography, below), especially Arizona, Delaware, and Maine.

The argument that the effect of local intervention or other

"errorll is lIequalled out II in nationally aggregated data is
spurious. Error never disappears, though it may be hidden. At
the local level, with knowledge of the local situation, there is
a better chance to discover and then control for local effects.

The case of aggravated assault demonstrates that conclusions

about seasonal fluctuation in a crime may be very different,
depending on the local area concerned. In Illinois, the F of
stable seasonality in Index assault from 1972 to 1981 varies by
jurisdiction: for example, Chicago, Cook County (excluding
Chicago), SMSA counties containing a central city, suburban SMSA
counties not containing a central city, and Quincy, a typical
small city of less than 50,000 population (see Chart 1). The
more urban the place, the more aggravated assault known to the
police appears to fluctuate seasonally.

Such a change in the F, a measure of significance (see page

4, above), might be argued to be caused by a change in the number
of observations. It is true that a longer series, or a series
containing more observations, is more likely to produce a high F,
other things being equal, and there are fewer reported assaults
in Quincy than in Chicago. However, the lack of seasonal fluc-
tuation in Quincy is evident not only in the F statistic, but
also upon inspection of the graph (see Charts 2 and 3). Despite
the difference in scale, there is a pattern of seasonal fluctua-
tion in Chicago, but no discernible seasonality in Quincy.

7
 Chart 2
CHICAGO INDEX AGGRAVATED RSSAUll, !~12 10 2981
sounCE, SAC EOITlON, ILLINOIS UNIFOnM CnlME REfOnT5. I NOEX AGGRAVATED ASSRULT • AGGRRVRTED RSSAUI. T. RGGRAVATED
OffENSE DATA • aAlTERl, ~TTE/1PT MURDER.

.........."... ':''''''''' ....... ,.........................'............. '··7···" ....... ,... ;." ................":'...... "...... ,,!' .............,' j··.... ·•·•..·· ....1· ....··,·· ......"T....·.... '"..·,"j· ........·.... ·..·l
. I , !
~ i ; if!
- ......···"'··.... 7" .... •··•·..•..··['···.. •• ..·· ..•.. ·1"········ ....····'1'....•·· .... ·.. •.. y..····•..•.. ··•.. '1 ....·•·· ....·..··'1' ......·.. ·· ..··· .. 1·'··.. ·..,,·......"[············ ..·.. 1··.... ··,,······ ..7'".. ··· .. ·· ......1·....•·..·•..·····,..··, .... _· .. ••..1···....•• .. •••• .. '1
: .
f. ! ~

....•.. ,.......... ;...... '''. ,.......~ ............. "... ;.... ,.......... ,...~.................. ;.................. ~ ...........,...... ;............. 6~"J.

••••••• , ••. , . . . . . . . . . . . . . . . : .• " .. ' . " .. n

: .
.... ~ ..................
.
!..................
~."
: :
............... :"' ............ ".'i' ...... l\ .. ~ ....... : " .... ~., ..

.-:" ........ ,.........:.................. ~

(
.... ..............:.............. :
: ~
...................., .......j

I
....... ,. ;··· ..··.... ·..··.. !· ..·.. ·.. ·....·'' t···. ····. · ·. ··~ .............
; : :
........c .................. , ................... ,........... ,.... ; "., ............. ;

Chart 3
QUINCY INDEX AGGRRVRTED ASSAULT, !97~ TO 19S!
SOURCE, SAC EDITtOtl. ILLlNO!S UNIFORM CRIME nEPORTS. I NOEX AGGRAVATED ASSAULT • AGGRAVRTED ASSAULT. AGGRP'(RTED
OFFENSE DATA. BATTERY. ATTEMPT HUADER,

"7·· .. ·•• .. ··· .... ··t...... ·••• .. ,.... ·{· .. ·,',··".,····,;······ .. •.. "· .. '·~· .. •.. ·' .. •• .. ·"·:···.. ··'··· .... ···i .. ·.. ·· .......... ·i· ..••...... ·......1
, ~ i : i
............ j .. '··.... r· ..........·....· :..........··.. ··.·: ..........·....··:··..·............ ·:·................ ·1 .............. •.. ·j·. ·. . . . ·. ·. ,. . . ·. . . ·. ·'1
, '
. . . . ! :
. .. ,···· ...... ·'1' .... ·" .. , .. ······;'"
··•• ...... •••.. ···,:.··· . ., ......... '1 ....... ' •.•••. , : .................. , •........ , ............ • .. • .... ' ....... 1........ • ...• ..... ·• .... ' ............. , .................. :. ........• ..••••• .. I· .......•• .... ···~t
~ , . . . 1 i
: i
: i; !t
.,................ +...,.............. ~.......... .......... ... ".. ,. ,....... ! •. , ••• '.'." ..... ." ........ '., ...
, .
·.-i '......... ,'.... ; ." •.••. '., ..... !·· .. ···, .... ····,·'!", .. •·• .. ··· .. ·..
:
·l···,···.. · ..·.... t······· .......... ··~ ..···....···· ..·.. t·· .. ·......··..···r··......,........i
: : : !

~
.,'. . . . .'. . . .~. . . . . . . . . ~,. . . . . . .,. '~· . . . ·. . ·. . ·;· . ·. · . ,·. ·. t·. ·. ·. · . . . '~. . . . ·-··. . ·~
. . . . . , . . . . . . . . . . . .;. . . . . . . . . . . .H . . . . . j'................,_,." .. ,....... ,•.. ,,'..... ' ........ ................. .
~ ·.... ·...... ·• .. ·,!··.. "··,,,···..··t.............. ·,· ,...... ,......... ,.~.,
, .................i, .......,......... i,..................i...................~ ~

. 1 !
... :.' ..... '.' ......... j.................. '?, ..................•!... '...........' '1!
. <
· ....... ·., ....... ,~ .......... ' ....... I· ......... ·••• .... ,;· ...... ",.· ..... '1 ........ ,., .... ~ .... , ............ u·· ... ' .................... , •• ,,,., ................... ~ .......... .. ~

. . .
ii i : • : ! : ;
1
'.! .......... ,....... ........................,............... ,........... ; .......... " •.
, ................ ~ .. " .. h ........ ~ ,' ..... ! ............... , ... ~ ... " ............. !.................. "i' ... j . . . . . . . . . . . . ,,!.· ......... ~ ...... ~ ...... toll· ...... ·!· .. •.. •••...... '··i•
~ ,

.. ..... " ........ . . . .

,..~ ,:.................. ~ ................... !..................~

8
 Does Crime Occur Seasonally, or Is It Reported Seasonally?

Most homicides are i in their characteristics, a subset of

another crime -- aggravated assault. They begin as a fight,
brawl, or argument. Although some homicides are precipitated by
a robbery, a rape, a gangland "hit, II or another non-assault cir-
cumstance, the great majority begin as a fight, brawl, or argu-
ment, and escalate to murder (see Block, 1977; Zimring, 1979;
Parker and Smith, 1979; Pierce and Bowers, 1981). In Chicago,
for example, 68 percent of the homicides known to the police from
1965 to 1981 began as an assault, 17 percent began as a robbery,
2 percent began as a rape or burglary, and the rest involved
other or unknown circumstances.3

Homicides that begin as fights or brawls differ from

homicides that begin as robberies in a number of ways, including
the weapon, the place and time of occurrence, the victim-offender
relationship, and many other characteristics. In fact, homicides
that begin as assaults are more similar in their characteristics
to aggravated assaults than they are to homicides that begin as
robberies (Block, 1977). They can almost be considered to be
separate types of crime. A homicide that begins as a fight or
assault can be thought of as a type of aggravated assault, one in
which the victim was injured so seriously that death resulted.

Because assault homicides are similar in most of their char-

acteristics to all aggravated assaults, we would expect them to
be similar in another characteristic, seasonal fluctuation. If
more assaults occur in the summer months, for example, we would
expect more assault homicides to occur in the summer. However,
Chicago data (Table 2) indicate that this is not necessarily
true. Aggravated assaults knom1 to the police fluctuate season-
ally. Assault homicides do not. Why does this occur?

In searching for an explanation I we hypothesized that, in

fact, assault does not occur more often in the summer than in the
winter months. Rather, it becomes known to the police more of-
ten in the summer months. More serious assaults, involving
serious injury and possibly hospitalization, become known to the
police with greater likelihood than less serious assaults. They
are reported by medical personnel and hospital staff, and are
more likely to be reported by the victims themselves (LEAA, 1972;

3S out'ce: Data collected by the Authority, the Center for

Studies 1n Criminal Justice of the Universi.ty of Chicago Law
School, and Loyola University from files maintained by the
Chicago Police Department. Time series files maintained by the
Authority.

!'l
 Table 2

Assault Homicide and Aggravated Assault *

Chicago 1.-1..2.6 7 to 1981

Additive Multiplicati¥¥
Assumption ** Assumption

Stable % Contribution Stable % Contribution

F Season. Irreg. F Season. Irreg.

All WeaQon T:tQes

Agg. assault 109.1 70.0% 27.3% 10li.5 70.3% 27.li%

Assault homicide 6.2 25.1 74.3 6.2 25.7 73.7

Firearm

Agg. assault 60.9 60.0% 35.4% 5li.2 57.4% 38.7%

Assault homicide 4.3 2li.2 75.2 li.2 2li.1 75.3

Other Weapons

Agg. assault 89.2 65.6% 31.5% 90.6 65.0% 32.4%

Assault homicide 3.8 29.3 70.3 3.8 2li.2 75.5

*For a discussion of these statistics, see page 5, above.

Data sources: Assault data set reconstructed by the Authority
from two sources I each containing partial data: Chicago Police
Department records, and the City of Chicago Municipal Reference
Library. Time series data are in 13 "police periods /I per year.
To obtain estimates of months for the present analysis, we used a
moving average. This probably decreased the amount of seasonal
fluc~~ation. Assault homicide: see note 3, above.
'Under an additive assumption, the seasonal and irregular
components are independent; under a multiplicative assumption,
they are not. For more detail, see Block (1983).

10
Block and Block,1984). Less serious assaults, on the other hand,
tend not to come to the attention of the police unless they are
public. They at'e more likely to be public in the summertime. In
the warm months, an assault is more likely to occur outside, and
if it occurs inside, the windows are more likely to be open.
Thus, we hypothesized that the explanation for the seasonality of
assault and lack of seasonality of assault homicide is that
neither, in fact, occurs seasonally, but that more assaults be-
come known to the police in the summer months.

To test this hypothesis, we compared seasonal fluctuation in

the number of aggravated assaults known to the police, and
seasonal fluctuation in the number of aggravated assault vic-
timizations, for the same place and time period: the United
States from 1973 through 1979. The police data are Index ag-
gravated assaults reported by police jurisdictions to the FBI
through the Uniform Crime Reporting program. 4 The victimization
data are aggravated assault victimizations (defined the same as
Index aggravated assaults) estimated by the National Crime
Survey. 5 We expected to find much more seasonal fluctuation in
reported assaults than in assault victimizations. That is, in
fact, what we found, as Table 3 shows.

4These unweighted UCR data are the total offenses reported

by those agencies that reported to the FBI in every month of the
year in question. The number of reporting agencies increased
during the 1973-1979 period from 7,106 (representing a population
of 174,249,026) to 11,782 (representing 196,836,371). {Note that
the added agencies were mostly small jurisdictions, serving
cities and counties with smaller populations.) Thus, these data
cannot be used to examine trends over time. However, because
Census X-II results indicate the behavior of the seasonal com-
ponent of the series as opposed to the behavior of any "trend,1I
whether real or artificial, the data can be used as one indicator
of the presence of seasonal fluctuation. For another analysis of
FBI/UCR data, see FBI (1981) in the Review Bibliography, below.
5These v1ctimization estimates from the National Crime
Survey were provided by Richard L. Block and Wesley Skogan. They
represent the number of victimizations occurring in the NCS
sample, corrected for underrepresentation of various population
groups. The sample includes only noncommercial victims aged 12
and older. Victimizations in which the month of OCCUY'rence is
unknown, such as IIseries ll victimizations, are not included.
These missing data account for fewer than 1 percent of the rob-
beries or assaults in any month. Dodge and Lentzner conducted an
analysis of NCS data for a shorter (five-year) period, but did
not attempt to analyze crime categories comparable to police
categories. See US/BJS in the Review Bibliography, below. As
this report went to press, 1980 and 1981 NCS data became avail-
able. We plan to continue this analysis with more recent data,
as they are available.

11
 Table 3
Seasonality of Victimization and Reported Crime
United States, 1973 to 1979
Additive Multiplicative
Assumption AssumI;!tion
Stable % Contribution Stable % Contribution
P Season. Irreg. F Season. Irreg.

Aggravated Assault
,
Known to Police 94.2 83.7% 15.0% 131.3 85.0% 13.7%
Victim Survey 7.5 31. 0 68.4 7.4 29.9 69.5
Robber~

Known to Police 100.1.1 86.9% 10.6% 107.6 86.9% 11.0%

Victim Survey 4.6 34.0 63.7 4.5 27.3 70.4

*Por a discussion of these statistics, see page 5, above.

Having found that reported assault fluctuates with the

seasons much more than assault victimization, at least for the
United States as a whole, we looked at the same relationship for
another violent crime -- robbery (Table 3). Robbery victimiza-
tions, lil{e assault victimizations, do not fluctuate with the
seasons, but robberies known to the police do. This suggests
that less serious and less public robberies do not occur more
often in the summer, but are more likely to become known to the
police in the summer months.
Because the hypothesis is grounded on the issue of IIpublic ll
crime, and the degree to which a crime is public may be related
to whether it occurs in an urban area or not, a better test of
the hypothesis would compare seasonal fluctuation in victim and
police data in urban areas only. If this were done, we would ex-
pect that the difference in Census X-II results would be. even
higher. However, data are not currently available to do this. 6

6Victim survey data are not available for seven years in any
city, and there are some problems in utilizing the NCS to examine
metropolitan areas.

12
 Partial support is found in the above analysis of assault
data (see "Local versus State or National"). Aggravated assault
series are more likely to be seasonal the more urban the place.
It is also supported by the seasonality of Chicago assaults by
type of weapon, in Table 2 above. The F of stable seasonality
and the percent contribution of the seasonal component are higher
for assaults without a gun than for assaults with a gun. Perhaps
assaults without a gun are less serious, result in lesser or no
injuries, and are thus more likely to become known to the police
in public situations, esped.ally in the summer months.

13
14
 IS CRIME SEASONAL?

In this section, we summarize the fj.ndings of analyses of

seasonality in crime that the Statistical Analysis Center of the
Authority has conducted in the last few years, either in response
to requests from users or in conjunction with a research project.
The reader's evaluation of the findings presented here, and
indeed the evaluation of any analysis of seasonality, should take
into account the method and criteria used, the place and time
period, and the definition of the crime.
To facilitate comparison, all the analyses summarized here
use the same method, the Census X-II, and the same criteria, F of
stable seasonality and percent contributions of the seasonal and
irregular component s over a one·-month span. However, we have
repeated many of these analyses with a stochastic (ARIMA) model,
and results are available on request from the author.
Crime definitions used also are consistent throughout the
section: Index crimes, as defined by the Uniform Crime Reports.
Even when we analyze victimization estimates from the National
Crime Survey, we attempt to make NCS crime categories comparable
to Index crime categories (see note 5, above, and Block and
Block,1984). The section summarizes, in turn, findings for each
of the four violent Index crimes and three property Index crimes.
The eighth Index crime, arson, is not included, because, in gen-
eral, consistently defined data have not been collected for the
minimum seven-year period.
Although all other aspects are uniform, place and time
change from analysis to analysis, and are noted for each.
Sources of the time series analyzed in this section are the
following:
• United States 19'70 to 1982: Uniform Crime Reports, FBI.
See note 4, above.
• National Crime Survey: See note 5, above.
• Illinois, parts of Illinois: Authority edition of the
Illinois Uniform Crime Reports offense data and supplementary
homicide data. See Miller and Block (1983).
• Chicago aggravated assault: Data set reconstructed by the
Authority from two sources, each contdlning partial data: Chicago
Police Department r'3cords, and the City of Chicago Municipal
Reference Library.
• Chicago homicide: Data collected from police homicide
files, with the cooperation and assistance of the Chicago Police

15
Preceding page blank
Department over a number of years and changes in administration.
See note 3, above .
• California, Los Angeles: Authority time series version of
data obtained from the Bureau of Criminal Statistics and Special
Services, state of California.

• New York City: Lily E. Christ, John Jay College of

Criminal Justice, City University of New York.

o Boston: Deutsch and Alt (1977); see Review Bibliography.

$ Canada, Ontario: Time series files created by Craig McKie,

Statistics Canada, from homicide data collected by the Law
Enforcement Section, Canadian Centre for Justice Statistics,
Statistics Canada. See Block, et al. 1983.

Is Violent Crime Seasonal?

Homicide

The preponderance of the empirical literature finds that

homicide does not vary seasonally.7 In the Review Bibliography
below, see Munford, et al. (1976:220), Deutsch and Alt (1977),
Hay and McCleary (1979), Lamp (1982), and Michael and Zurnpe
(1983) . However, the prevailing point of view in criminology
seems to be that homicide is seasonal. See the quotes from the
President's Commission and Sutherland and Cressey in the
Introduction, above, and FBI (1981), Warren, et al. (1981).

As parts of various research projects and in response to

user requests, the Authority has analyzed hundreds of homicide
series for the presence of seasonal fluctuation. Most of these
series were types of Chicago homicide or types of Canadian
homicide; for example, homicide committed with a gun, homicide
committed with a knife, and homicide committed with a blunt in-
strument. In addition to the Chicago and Canadian series, we
have analyzed Boston homicide, California and types of California
homicide, Index homicide in the United States, Index homicide in
New York City, and homicide and types of homicide in Illinois
jurisdictions outside of Chicago. Table 4 summarizes results for
the major Chicago homicide component series.

7Although suicide is not usually considered to be a crime,

it is often discussed in conjunction with homicide. For a review
of the literature on seasonal fluctuation in suicide, see Lester
(1972) . The Authority has analyzed Cook County suicides from
1963, and has found no evidence of seasonal fluctuation, either
in suicides with a gun or without a gun. F values range from 0.5
to 1. 8.

16
 Table 4
Bell-Canada p* of Stable Seasonality, Chicago Homicide 196~
Additive Multiplicative
Assumption Assumption
Total Homicide 4.25 4.13
Firearm 2.86 2.75
Knife 2.17 2.07
Blunt Instrument 3.15 **
Assault 5.38 5.20
Robbery 2.07 **
Family 1. 23 1. 26
Acquaintance 2.36 2.20
Stranger 2.58 **
Black victim 3.05 3.05
White victim 1. 86 2.12
Male victim 5.34 5.00
Pemale victim .39 .39
Offender age 15-19 1.43
Offender age 20-24 1.13 **
1. 50
Offender other ages 4.15 4.10
Single offender 5.05 4.94
Multiple offenders 1. 83 1. 59
Inside a residence 1. 28 .90
Inside a nonres. 3.03 3.24
Out of doors, vehicle 18.60 16.56

'The Bell-Canada is a quick screener for seasonality, simi-

lar to the Census X-II but allowing no user options. The Bell-
Canada F value can be interpreted as the Census X-II F (see page
5, ab~¥e).
Because the series contains one or more zero values, a
multiplicatiVe adjustment is not applicable.

In addition to the types of homicide pr~sented in Table 4,

we analyzed more detailed Chicago homicide types, such as multi-
ple offender assault homicide with a gun. We also conducted more
oetailed analyses, using the Census X-II, on any series having a
Bell-Canada F value of 2.41 or over. After analyzing more than a
hundred Chicago series, we finally found one that showed some
seasonal fluctuation. This may be one of the few cases in Which
the maxim, lithe exception proves the rule, II is true. Of the

17
dozens of series, the only type of homicide that fluctuates
with the seasons is homicide oecurring out of doors or in a
vehicle. This type of homicide tends to occur half as often in
January and almost twice as often in August as in an average
month.
Twenty years of Canadian homicide data were categorized into
components comparable to those used in thG Chicago analysis: age,
sex, and ethnicity of victim; weapon; victim-offender relation-
ship; and precipitating event (assault versus robbery). In addi-
. tion, we analyzed each province separately. (For details, seA
Block,et al. 1983.) Consistently, each Canadian component showed
a lower stable seasonality F value than the corresponding Chicago
series (see Table 5). This d~fference was not due to more obser-
vations in the Chicago data. The Canadian series actually is
longer, and contains about the same number of homicides per month
as the Chicago series.

None of the California homicide series, nor the 20 component

California series, was seasonal (however, note that these series
are short). Neither the New York City series nor the Boston
series was seasonal.
The only exception to the general finding of. very low F
values for homicide is United States homicide from 1970 through
1982. There are several possible reasons for this. First, the
relatively high F value may reflect the high number of cases,
which ranges from 1,087 to 1,866 per month. Second, it may
reflect some artifact in the data. These data were collected by
the FBI from reports from local jurisdictions, and the number of
jurisdictions increases over time (see note 4, above). Third, it
is possible that homicide fluctuates seasonally in some U. S.
localities that we have not analyzed, Southern states for ex-
ample. If so, seasonal fluctuation in these areas might override
the lack of seasonal fluctuation in Illinois, New York City,
Boston, and California to produce seasonality in the national
aggregate.

Even though the United States Index murder series is sig-

nificantly seasonal according to the Plewes criteria (page 5,
above), the degree of seasonal fluctuation is very small. An in-
dicator of this is the "final seasonal factors,1I a result of the
Census X-11 analysis method. A seasonal factor can be inter-
preted as a monthly weight. If it is more than 1.00, then the
month tends to be high; if it is less than 1.00, then the month
tends to be low. For example, the January seasoral factors for
Chicago homicides occurring outside or in a vehicle (see Table 4)
are all less than 0.50 over the 17 years, and the August seasonal
factors all approach 2.00. Twice as many outdoor homicides tend
to occur in August, and half as many in January, as in a typical
month.

18
 Table 5
Census X-II Results in Selected Homicide Series
Additive Multiplicative
AssumDtion Assumption
Stable % Contribution Stable % Contribution
F Season Irreg. F Season. Irreg.

Total Homicide
United States 1970-82 28.9 70.1% 28.2% 31. 0 71. 5% 27.1%
Canada 1961-80 1.4 13.7 86.0 1.4 16.7 83.0
Illinois 1972-81 * 2.8 29.0 70.4 2.8 29.4 70.0
California 1976-80 4.3 44.0 53,8 7.8 49.2 48.7
Ontario 1961-80 0.8 18.7 81. 0 0.8 16.6 83.2
Chicago 1965-81 5.6 30.5 67.2 5.3 27.1 70.7
Othe~ Illinois 1973-82 0.5 9.7 90.0 0.5 5.4 94.5
New York City 1973-§2 7.2 45.2 54.5 7.1 41. 0 58.7
Los Angeles 1976-79 2.7 39.1 60.4 2.4 35.8 63.7
Boston 1966-75 2.8 26.8 72.4 2.7 27.7 71.5
Homicide with a Gun
Canada 1961-80 2.1 19.0% 80.6% 2.3 19.1% 80.7%
Illinois 1973-1982 * 2.0 26.9 72.5 2.1 9.0 90.3
California 1976-1979 1.5 35.2 61. 3 1.6 9.6 89.9
Ontario 1961-80 1.8 22.0 77.6 ** ** **
Chicago 1965-81 3.8 29.3 69.7 3.6 25.2 74.0
Other Illinois 1973*82 0.5 12.4 87.4 0.5 5.0 94.8
Los Angeles 1976-79 1.4 41.8 57.7 1.3 31.4 67.7
Assault Homicide
Canada 1961-80 1.9 22.0% 77.8% 1.7 26.6% 73.3
Illinois 1973-1982* 3.5 23.8 75.8 3.5 10.5 89.4
California 1976-79 3.0 29.8 69.3 3.0 25.8 73.4
Chicago 1965-81 6.2 2l. 6 77.6 5.7 22.2 77.2
Other Illinois 1973-82 1.1 17.5 82.1 1.0 10.8 89.0
Robbery Homicide
Canada 1961-80 1.9 25.1% 74.5% ** ** **
Illinois 1973-82 • 1.3 19.8 77.3 ** ** **
California 1976-79 0.5 9.0 90.4 0.5 9.6 89.8
Chicago 1965··81 2.4 9.1 80.5 2.4 19.1 80.5
Other Illinois 1973-82 0.7 13.6 86.0 ** ** «. *

*Note that this series contains fewer than seven years. See
II Leng~~
of Ser'les, above.
II

Because the series contains one or more zero values, a

multiplicative adjustment is not applicable.

19
 In contrast, the seasonal factors for United States Index
murder (multiplicative assumption) are close to 1.00 for all
months. For 1982, they are the following:

January .99 May .99 September 1. 03

February .90 June 1.00 October .99
March .9"( July 1. 08 November .98
April .92 August 1.07 December 1. 07

Seven to 8 percent more Index murders tend to occur in July,

August, and December than in the ~ypical month. About 10 percent
fewer tend to occur in February. Otherwise, no month tends to
be particularly high or low. However, because this weak pattern
is consistent throughout the 13 years of the series, and because
the number of cases is so high, the F value is relatively high.

In summary, it is possible that certain types of homicide,

such as those occurring out of doors or in certain parts of the
country, vary with the season of the year. However, for most
types of homicide in most places, seasonal fluctuation, if it ex-
ists, may be too weak to affect practical administrative or
policy decisions.

Forcible Rape

Forcible rape, like homicide, is seasonal in the United

States as a whole ('.fable 6). However, it is not seasonal, by the
Plewes criteria, in New York City or in any of the Illinois
jurisdictions we have analyzed, including three categories of
counties: Cook County (which contains Chicago), small-city coun-
ties (nonmetropolitan counties with a city of 25,000 to 49,999
population), and rural counties (all other nonmetropolitan
counties) .

For other analyses of forcible rape, see, in the Review

Bibliography below, Deutsch (1978), Edgerton et al. (1978), FBI
(1981), Lamp (1983), Marshall (1977), Michael and Zumpe (1983).
In the II Survey of the States," see California, Delaware,
Kentucky, Maine, and North Carolina.

8This is not surprising, given that February has about 7

percent fewer days than 30-day months and 10 percent fewer than
31-day months.

20
 'l'ab Ie 6
Census X-11 Results in Selected Forcible Rape Series
Additive Multiplicative
Assumption Assumption
Stable % Contribution stablel% Contribution
F Season. Irreg. F Season. Irreg.

U.S. 1970-82 110.8* 86.6% 12.1% 283.2 86.3% 12.6%

Illinois 1972-81 10.0 20.5 79.0 10.0 17.0 82.5
Cook County 1972-79 3.8 27.6 71. 9 4.4 30.9 68.6
Chicag,:> 1972-81 2.6 17.2 82.4 3.1 15.5 84.2
Small-city Coun. 72-79 1.7 27.6 72.1 **
11.2
** **
Rural Coun. 1972-79 12.7 36.2 63.4 25.4 74.3
New York City 1973-82 10.9 40.3 59.4 10.8 28.1 71. 8

*Moving seasonality present at the 1 percent level.

**Because the series contains one or more zero value, a mUl-
tiplicative adjustment is not applicable.

RobberY.,
Robbery is a violent Index crime that also involves the
taking of property. We have suggested above (IiDoes Crime Occur
Seasonally I or Is It Reported Seasonally? II) that the seasonal
fluctuation found in some robbery series may reflect a tendency
for less serious robberies (attempts, for example) to become
known to the police more often in the summer months. The
analyses of Index robbery conducted by the Authority do not offer
much illimunination of this question (Tables 7 and 8).
We would expect that robberies known to the police, espe-
cially less serious robberies occurring in cities, to fluctuate
seasonally. The number of Index robberies reported to the FBI in
the United States as a whole does fluctuate seasonally (Table 8),
as do Index robberies in New York City. In New York City, August
and December have more robberies reported to the police than
other months. In the United States as a whole, December is 16
percent to 18 percent higher, and April and May are 10 percent to
12 percent lower than the average month. If we consider armed
robbery to be relatively serious, we would expect that it would
not fluct~ate seasonally and that is what the lO-year Boston
I

data indicate.
Robbery victimizations (occurring to noncommercial victims
aged 12 and older) do not fluctuate seasonally (Table 8).
Neither the weapon nor the relationship of the offender to the
victim makes a difference in the seasonal fluctuation of
robberies in the National Crime Survey.
21
 Table 7
Bell-Canada F of Stable Seasonality, Index Robbery
Selected Illinois Cities, 1972 to 1981
Population Stable Seasonality F
Additive Multiplicative
Assumption Assumption

Chicago 3,005,072 6.4 6.6

Springfield 99,098 4.2 4.6
Joliet 78,165 1.2 1.1
Champaign 57,176 1.1 1.0
East st. Louis 54,966 1.1 1.2
Quincy 42,048 1.9 *

*Because there is one or more zero value in the series, a

multiplicative adjustment is not applicable.

Table 8
Census X-11 Results in Selected Robbery Series
Additive Multiplicative
Assumption Assumption
Stable % Contribution Stable % Contribution
F Season. Irreg. F Season. Irreg.

U.S. 1970-82 146.7 87.5% 10.4% 159.0 86.5% 11.4%

Nat. Crime Survey 1973-79
Total Robbery 4.6 34.0 63.7 4.5 27.3 70.4
Robbery with a gun 4.3 38.4 61. 2 4.6 30.8 68.9
Rob. without a gun 2.3 24.8 72.9 2.3 23.4 74.7
Rob. by a stranger 4.8 39.7 58.5 4.5 34.0 64.1
Rob. by acquaintance 0.4 31. 6 68.0 0.5 25.8 73.8
Illinois 1972-81 10.9 32.3 65.0 10.9 28.9 69.0
Chicago 1972-81 6.4 26.3 71. 6 6.6 21. 7 76.8
New York City 1973-82 36.1. 74.0 23.2 35·1 72.7 24.3
Boston 1966-75, armed 6.0 36.8 60.0 5.6 41. 2 55.4

*Moving seasonality present at the 1 percent level.

22
 On the other hand 1 none of the Illinois series we have
analyzed is seasonal by the Flewes criteria, whether the series
includes data from the state as a whole, Chicago (Table 8), or
various large and small jurisdictions (Table 7).
For other analyses of robbery, see, in the Review
Bibliography below, Block (1979), Deutsch (1978), Deutsch and Alt
(1977), Hay and McCleary (1979), FBI (1981), Ku and Smith
(1977,1978), Marshall (1977), Michael and Zumpe (1983) 1US/BJS
(1980). In the "survey of the states," see California, Delaware,
Kentucky, Maine, and North Carolina.

Aggravated Assault
Assault is repeatedly used as an example in "Issues in the
Seasonality of Crime," above. We suggest there that assault vic-
timization may not vary seasonally, but that less serious assault
may become known to the police more frequently in the sununer
months because it tends to be more public. The findings for as-
sault in the Authority's analysis (Tables 9 and 10) do not,
generally, conflict with this interpretu~ion. However, neither
do they give it strong support.
Among the Illinois cities we have analyzed for the 1972 to
1981 period (Table 9), assault fluctuates ~·dth the seasons in
Chicago, but not in the smaller cities. If we analyze 15 years
of data for Chicago, the presence of seasonal fluctuation is even
clearer (Table 10). For Cook County and for Illinois as a whole,
the irregular contribution is too high to make any definitive
statement about seasonality.
Boston data for firearm assaults over 10 years show no
seasonal fluctuation at all. New York City Index assault data,
which include aggravated assault by any weapon, have a high
stable seasonality F value, but, like the Illinois data, also
have an irregular that is too high for a definitive statement.
There seems to be no question, however, of seasonal fluctua-
tion in assault victimization, whatever the weapon or the
relationship of the offender to the victim (Table 10). For all
components of assault, F values are low and the irregular con-
tribution is high.
POl' other analyses of assault, see, in the Review
Bibliography below, Deutsch (1978), Deutsch and Alt (1977), Hay
and McCleary (1979), FBI (1981), Marshall (1977) 1 Michael and
Zumpe (1983), Pittman (196 1-1), US/BJS (1980). In the II Survey of
the States, see California, Delaware, Kent'lcky, Maine, and North
II

Carolina.

23

L
 Table 9
Bell-Canada F of Stable Seasonality, Index Aggravated Assault
Selected Illinois Cities, 1972 to 1981
Population Stable Seasonality F
Additive Multiplicative
Assumption Assumption

Chicago 3,005,072 32.3 28.9

Springfield 99,098 6.0 7.0
Joliet 78,165 2.0 1.5
Evanston 73,278 4.6 5.6
Champaign 57,176 4.8 4.8
East St. Louis 54,966 1-1, 6 *
.:+
Schaumburg 52.083 0.7
Quincy 1-12,048 1.2 *
Carbondale 26,144 0.9 *
*Because there is one or more zero value in the series, a
multiplicative adjustment is not applicable.

'rable 10
Census X-11 Results in Selected Assault Series
Additive Multiplicative
Assumption Assumption
Stable % Contribution Stable % Contribution
F Season. Irreg. F Season. Irreg.

U.S. 1970-82 94.7* 83.2% 15.2% 237.8 83.9% 14.6%

Nat. Crime Survey 73-79
Total Agg. Assault 7.5 31. 0 68.4 7.4 29.9 69.5
Assault with a gun 1.8 22.8 76.8 1.7 16.3 83.4
Ass. without a gun 9.4 47.3 52.0 9.4 45.8 53.6
Ass. by a stranger 4.8 18.8 80.4 4.6 20.6 78.7
Ass. by acquaintance 6.2 29.2 70.4 6.4 34.9 64.6
Illinois 1972--81 45.7 59.6 38.5 42.9 58.9 39.5
Cook County 1972-82 46.5 52.4 45.7 44.6 52.6 45.8
Chicago 1967-81
Total agg. assault 1 0 9.1* 70.0 27.3 104.5* 70.3 27.4
Ass. with a gun 60.9 60.0 35.4 54.2 57.4 38.7
Ass. without a gun 89.2* 65.6 31. 5 9 0 .6* 65.0 32.4
New York City 1973-82 58.1 62.6 37.0 60.6 59.4 40.1
Boston 1966-75, gun 2.8 24.3 74.8 2.9 12.8 86.6

*Moving seasonality present at the 1 percent level.

24
 Is Property Crime Seasonal?
Burglary
In the jurisdictions we have analyzed, burglary generally
has a low F of stable seasonality (though not usually as low as
homicide) and a high percent contribution of the irregular com-
ponent (Tables 11 and 12). However, it is strongly seasonal in
the United states as a whole and in New York City.
The contrast between New York City and Chicago is striking.
Chicago X-11 results contain no hint of seasonality, while the
same measures indicate consistent (but weak) seasonality in New
York.9 This trend is the same for robbery (Table 8, above), and
for motor vehicle theft (Table 15, below).10 A possible explana-
tion, and one that could be tested, is that certain types of
burglary are more apt to become known to the police in New York
City but not in Chicago during some seasons, either because they
usually become known or because they usually do not become known
to the police in Chicago, regardless of the season.
For other analyses of burglary, see, in the Review Biblio-
graphy below, Block (1979), Deutsch (1978), FBI (1981), Ku and
Smith (1977,1978), Marshall (1977), Schneider and Sumi (1977),
US/BJS (1980). In the "Survey of the States," see California,
Delaware, Kentucky, Maine, and North Carolina.
Larceny/Theft
Of all the crime types we have analyzed over the years,
Index larceny/theft seems to have the most consistently seasonal
pattern. The F of stable seasonality is high not only in Chicago
(Table 13), but also in smaller cities (although the irregular
contributions are often high). The F value is extremely high in
the United States as a whole and in Illinois as a whole. In the
United States, Illinois and New York City, seasonality con-
tributes 80 percent to 90 percent of the month-to-month variation
in the number of offenses known to the police. Cool{ County and
those Illinois nonmetropolitan counties containing a city of
25,000 to ij9,999 population also show seasonal fluctuation in the
number of larceny/thefts known to the police.

gIn New York City, July and August have 6 percent to 10 per-
cent more burglaries than the typical month.
10It is difficult to compare the seasonality of assault in
the two cities (Table 10), because the Chicago series is much
longer, and the New York series contains moving seasonality in
both the additive and multiplicative adjustment. Moving
seasonality indicates that more detailed analysis is necessary.
For larceny/theft (Table 1ij), even though New York City is
seasonal and Chicago is not by the Plewes criteria, the contrast
is not as striking as for burglary or motor vehicle theft.

25
 Table 11
Bell-Canada F of Stable Seasonality, Index Burglary
Selected Illinois Cities, 1972 to 1981
Population Stable Seasonality F
Additive Multiplicative
Assumption Assumption

Chicago 3,005,072 5.4 5.2

Springfield 99,098 1.3 0.9
Joliet 78,165 6.2 6.5
Evanston 73,278 3.6 3.4
Oak Lawn
Champaign
60,358
57,176
3.2
7.3
*
7.2
East st. Louis
Schaumburg
54,966
52.083
3.1
1.4
*
1.2
Quincy 42,048 2.2 2.2
Carbondale 26,144 1.3 1.3

*Because there is one or more zero value in the series, a

multiplicative adjustment is not applicable.

Table 12
Census X-II Results in Selected Burglary Series
Additive Multiplicative
Assumption Assumption
Stable % Contribution Stable % Contribution
F Season. Irreg. F Season. Irreg.

United States 1970-82 99.6* 89.8% 8.1% 130.6 90.1% 7.8%

Illinois 1972-81 19.3 55.9 41.7 19.0 53.6 44.2
New York City 1973-82 46.6 78.0 18.4 42.7 78.5 18.5
Chicago 1972-81 5.4 32.0 65.7 5.2 28.8 69.4
Champaign Co. 1972-79 5.1 40.2 57.9 1.5 43.0 55.0
Kankakee Co. 1972-79 4.4 30.3 66.1t 4.1 23.8 72.9
Macon Co. 1972-79 2.4 16.7 80.9 2.4 19.6 78.0

*Moving seasonality present at the 1 percent level.

26

.
~.
 Table 13
Bell-Canada F of Stable Seasonality, Index Larceny/Theft
Selected Illinois Cities, 1972 to 1981
City P02ulation Stable Seasonalit;'i F
Additive Multiplicative
Assumption Assumption

Chicago 3,005,072 39.9 36.9

Springfield 99,098 18.6 15.4
Joliet 78,165 10.9 12.1
Evanston 73,278 15·5 18.6
Oak Lawn 60,358 2.2 1.9
Champaign 57,176 7.4 6.2
East st. Louis 54,966 10.8 9.3
Schaumburg 52.083 7.9 8.0
Quincy 42,048 8.5 9.1
Carbondale 26,144 2.4 2.6

Table 14
Census X-II Results in Selected Larceny/Theft Series
Additive Multiplicative
Assum2tion Assum2tion
Stable % Contribution Stable % Contribution
F Season. Irreg. F Season. Irreg.

U.S. 1970-82 267.1 91.5% 6.7% 291.0 91. 0% 7.0%

Illinois 1972-81 95.8 80.7 18.2 98.2 77.2 21. 4
New York City 1973-82 59.1 80.2 17.1 65.3 79.5 18.0
Chicago 1972-81 39.9 60.1 38.3 36.9 54.4 44.1
Cook County 1972-79 75. 1 * 79.3 19.4 71.1 74.9 23.4
Small-city co. 72-79 46.5* 70.9 25.8 56.6 67-.3 29.2
Rural counties 72-79 23.9 53.6 43.2 29.3 50·5 46.9
Quincy 1972-81 8.6 32.3 65.4 9.0 32.3 64.8

*Moving seasonality present at the 1 percent level.

27
 Such a strong seasonal pattern can be useful in practical
si tuations . In our experience, a good prediction of larceny /
theft can often be made by knowing the seasonal pattern and lit-
tle else. ll
For other analyses of larceny/theft, see, in the Review
Bibliography below, Deutsch (1978), FBI (1981), Lamp (1983),
US/BJS (1980). In the "Survey of the States, II see California,
Delaware, Kentucky, Maine, and North Carolina.

Motor Vehicle Theft

Index motor vehicle theft is seasonal in the United States

as a whole and New York City, but not in any of the other juris-
dictions we analyzed. In New York City and in the nation as a
whole, the number of motor vehicle thefts known to the police
tends to be high from July through October. As with burglary,
the contrast between the lack of seasonal variation in Chicago
and the strong seasonal variation in New York City is striking.
As we suggested for burglary, this may be due to differences be-
tween the cities in reporting practices for certain types of
motor vehicle theft.

For other analyses of motor vehicle theft, see, in the

Review Bibliography below, Deutsch (1978), FBI (1981), and US/BJS
(1980). In the IISurvey of the States, II see California, Delaware,
Iowa, Kentucky, Maine, and North Carolina.

Table 15

Census X-II Results in Selected Motor Vehicle Theft Series

Additive Multiplicative
Assumption Assumption

stable % Contribution Stable % Contribution

F Season. Irreg. F Season. Irreg.

U.S. 1970-82 167.9 90.0% 8.9% 172.5 90.4% 8.5%

Illinois 1972-81 13.0* 39.1 59.6 12. 'r * 38.1 60.5
New York City 73-82 46.0 74.8 23.0 39.2 72.6 25.1
Chicago 1972-81 5.0 17.9 81. 8 5.0 17.2 82.5
Small~city co. 72-79 6.9 41. 4 55.8 7.1 38.9 58.6
Rural co. 1972-81 12.8 38.0 59.3 10.7 30.7 68.6
Quincy 1972-81 4.0 21. 5 78.0 4.5 9.3 90.4

11See the forthcoming Authority report, liThe Predictability

of Crime. II

28
 Summary
As we stated in the Introduction to this report, the answer
to the question, "Is crime seasonal?" is not nearly as straight-
forward as is usually believed. However, with the accummulated
knowledge from the seasonal analyses summarized in this section,
and with knowledge derived from other studies of seasonality (see
the Review Bibliography), perhaps we can, at least, define the
question a little more carefully.
Any discussion of seasonal fluctuation in crime must be
carefully qualified by several considerations: 1) the place
(rural versus urban, Southern versus Northern); 2) the conceptual
and operational definitions of the crime (victimizations versus
crimes known to the police, attempts versus completed crimes); 3)
circumstances relating to public or private crime (weapon, place
of occurrence, injury, victim-offender relationship, property
loss); and 4) numerical aspects of the series that would increase
the likelihood of significant results (longer series and higher
observations). Before anyone can make a definite statement as to
the seasonality of crime, it is necessary to design a well-
controlled study taking all these factor~ into account.
A working hypothesis for such a study, taking the totality
of our knowledge of the seasonality of crime into account, might
be the following: The direct effect of the seasons on the number
of crimes occurring is small or nonexistent for most types of
crime. Although this small effect may be unimportant for practi-
cal administrative or policy decisions, it may be important in a
detailed explanatory or predictive analysis. However, the
seasons are related to the likelihood that less serious crimes
will become publicly known, in particular, known to the police.
The analysis of seasonal fluctuation in crime may have
broader utility than providing a yes-or-no answer to the substan-
tive question, "Is crime seasonal? II Several of the above
analyses (seasonal fluctuation in New York City versus Chicago,
and in assault versus assault homicide) suggest that seasonal
fluctuation or lack of seasonal fluctuation in a data set may
provide a clue as to the way in which the crime under discussion
was defined, and the process by which the criminal justice agency
collected and maintained records of the crime's occurrence. We
suspect that seasonal analysis may be even more useful as an in-
dicator of data definition and data quality than it is as an in-
dicator of seasonal fluctuation in crime.

29
30

~~~--- ~-
 REVIEW BIBLIOGRAPHY

This is a comprehensive listing of analyses of seasonal

fluctuation in crime, published and unpublished, concentrating on
research from the 1960' s to the present. (For earlier research,
see Wolfgang's review, 1966.) Only analyses of crime incidents,
reported or unreported to the police, are included in the review;
analyses of arrests, court appearances or judgments, prison
population, or other cl'iminal justice data series are not in-
cluded here. The review also includes a summary of the findings
of a previously unpublished "Survey of the States," conducted in
1979, in which government officials from each state were asked to
describe existing analyses of seasonal fluctuation of crime in
their State.
The review does not include those unpublished analyses that
the Authority has conducted over the years, and that have been
covered earlier in this report: homicide in California, Canada,
the United States, and Illinois; Index crime in the United States
and New York City i certain crimes in Boston; National Crime
Survey robbery and assault estimates fo~ the United States; and
Index crime in Illinois and cities within Illinois.
The studies have not been chosen for their quality, and are
not necessarily methodologically sound. They have been chosen
only' because each addresses the question of seasonality in crime.
The results given for each study represent the findings of its
author or authors, not the opinion of the author of this review
or the Authority.
For each study, the review specifies title and author, type
of crime analyzed, time period and geographical area, method and
criteria used I as well as the findings. The included studies
vat'y greatly in method, cl~imes analyzed and theit' definitions,
and even the author's definition of seasonal fluctuation. It is
incumbent upon the reader, therefore, to take characteristics of
each study into account when making a decision as to its con-
tribution to our knowledge on the subject.
The reader should be especially cautious regarding the
length of the time series analyzed in each study. As we discuss
above (see "Length of Series "), the results of any analysis of
fewer than seven years should be read with skepticism. Many of
the studies in this section utilize fewer than seven years of
data, and some base their conclusions on one year only.
'1'he reader may not be familiar with some of the terms used
in the Review Bibliography under the categories "method, II "cri-
teria, II 8t1d II resul ts . I' These terms are explained in the report,
HO\1_ to Handle Seasonal.l~ (Block, 1983), which also contains, for
those who want more detail, an annotated bibliography of the sea-
sonal analysis literature. For the purpose of understanding the
present report, however, the following may be helpful:

31
Preceding page blank
ARIMA model A stochastic model of the form (p,d,q)(sp,sd,sq)
where p = degree of autoregressive process, d =
degrees of differencing, q = degree of moving
average process, and sp, sd, and sq indicate
seasonal autoregressive, seasonal differencing,
and seasonal moving average process, respec-
tively. If the second term is not in the model,
it is not a seasonal model.
Autocorrelation Describes the relationship of observations with-
in a series. In a seasonal series, observations
12 months apart are correlated. In a successful
model, the residuals will not be autocorrelated.
Bell-Canada A short version of the Census X-ll, used as a
screener.
Census X-ll A method of analyzing seasonality, developed by
the U.S. Census, and widely used from the 1950's
to the present. It divides a series into three
components: seasonal, trend/cycle and irregular.
Contribution of A statistic generated by the X-11; the percent
the Irregular contribution of the irregular component to
month-to-month variation in the series. Contri-
butions of the three components add to 100
percent. See page 5, above.
Plewes rule-of- Criteria for the presence of seasonal fluctua-
thumb tion in a series. If the irregular contributes
30 percent or more of the total month-to-month
variation, the decision should be "no stable
seasonality," regardless of the F value. If the
percent contribution is 25 to 29 the F value
needs to be at least 15, and if the percent con-
tribution is 15 to 24 the F value needs to be at
least 2.41 for the series to be considered
seasonal.
Stable Season- A statistic generated by the X-11: the ratio be-
ality F tween the seasonal and irregular components. See
page 5, above.
Stochastic model An approach to seasonal analysis that emphasizes
forecasts and the relationship of each observa-
tion to previous observations. The ilBox/
Jenkins II and "ARIMAII methods involve stochastic
models.

32
Banks, Jerry and David Vatz
1976 Sinusoidal pattern analysis jn criminal incidence. Crimin-
olo~ 14(2,August):241-250.

CRIME: "Serious offenses committed by juveniles." (No

greater detail provided.)
TIME, PLACE: Fulton County, Georgia. January 1972 to March 1974
(two and one-half years).
METHOD: Multiple stepwise regression; trend and sine-cosine
components.
CRITERIA: Regression components must exceed two times their
standard error, and if one member of a sine-cosine
pair is significant, the other is significant.
RESULTS: Two peaks and two valleys occur each year. June and
December are low.

Block, Carolyn Rebecca

1979 Descriptive Time Series Analysi8 for Criminal Justice Deci-
sion Makers: Local Illinois Robbery and Burglary. Chicago:
Statistical Analysis Center, Illinois Law Enforcement Com-
mission.
CRIME: Index Burglary, Index Robbery.
1!1IME, PLACE: 77 Illinois jurisdictions (counties, cities with
more than 25,000 population, planning regions)
1972-1977. Note: series are six years long.
METHOD: Bell-Canada screener; Census X-11.
CRITERIA: Plewes rule-of-thumb for stable seasonality F and
percent contribution of irregular.
RESULTS: No seasonality in any series.

Block, Carolyn Rebecca and Richard L. Block

1980 Patterns of Change in Chicago Homicide: The Twenties, The
Sixties, and the Seventies. Statistical Analysis Center,
Illinois Criminal Justice Information Authority.
CRIME: Eighteen types of homicide (homicides attributed to
white, blacl{ or Latin offenders, to young or older
offenders, to male or females; homicides of victims
by age, sex, and race categories; homicide with a
gun or another weapon; homicides precipitated by a
fight or robbery; and combinations of these).
'rIME PLACE:
I Chicago, 1965-1976.
METHOD: Bell-Canada screener, Census X11, stochastic model.
CRITERIA: Plewes rule-of-thumb for stable seasonality F and
percent contribution of the irregular; simplest
model with no significant autocorrelations in
residuals.
REr:3UL'fS : No seasonality in any series.

33
Block, Carolyn Rebecca, Craig McKie, and Louise S. Miller
1983 Patterns of change over time in Canadian and United states
Homicide. Policy Perspectives 3(2):121-180.
CRIME: Thirteen types of homicide (gun vs. not gun; offen-
der age 15-24 vs. other; fight vs. robbery; family
vs. acquaintance vs. stranger; "native" vs. other
for Canada; black vs. other for Chicago; female vs.
male victim).
TIME, PLACE: Chicago, 1965-1981; Canada 1961-1980, also each
Canadian province for same time period.
METHOD: Bell-Canada screener, Census X-11.
CRITERIA: Plewes rule-of-thumb for F of stable seasonality and
percent contribution of irregular.
RESULTS: No seasonality in any series. Highest F of Canadian
series, 2.18. Highest F of Chicago series, 4.23
(irregular contribution 60 percent).

Block, Carolyn Rebecca, Louise S. Miller, Richard Block, Douglas

Hudson
1981 Explaining patterns of change over time in Chicago homi-
cides with a gun. Manuscript. Statistical Analysis Center,
Illinois Law Enforcement Commission.
CRIME: Homicide with a firearm (same types as above), hand-
gun, rirIe, and shotgun registrations, gun and not-
gun suicides, gun and not-gun aggravated assaults;
also number of general assistance (GA) and aid to
dependent children (ADC) cases.
TIME, PLACE: Homicide, Chicago 1965-1978; registrations, Chicago
1969-1980; suicide, Cook County 1963-1979; assault,
Chicago, 1965-1978; GA and ADC, Cook County 1965-79.
METHOD: Bell-Canada screener, Census X-I1.
CRITERIA: Plewes rule-of-thumb for stable seasonality F and
percent contribution of irregular.
RESULTS: No homicide series was seasonal. Rifle and shotgun
but not handgun registrations seasonal. Assault
seasonal. Neither gun nor not-gun suicide seasonal.
ADC seasonal after 1975, not before. GA not
seasonal.

Deutsch, Stuart Jay

1978 Stochastic models of crime rates. International Journal of
Comparative and Applied Criminal Justice 2(2):127-151.
CRIME: Seven Index crimes: homicide, forcible rape, rob-
bery, aggravated assault, burglary, larceny/theft,
and motor vehicle theft.
TIME, PLACE: Ten cities: Jan. 1966 to July 1975 for st. Louis,
Portland, Los Angeles, Kansas City, Atlanta, Cleve-
land, Boston and Denver. Jan. 1970 to July 1975
(five and one-half years) for Dallas and Cincinnati.

34
MErrHOD: Stochastic modelling.
CRITERIA: Simplest model with no significant autocorrelations
in residuals.
RESULTS: Homicide and rape not seasonal in any city. ARIMA
model for all other crimes in all cities was
(0,1,1)(0,1,1).

Deutsch, stuart Jay and Francis B. Alt

1977 The effect of Massachusetts' gun control law on gun-related
crime in the city of Boston. Evaluation Quarterly 1(4,
November):543-568.
Deutsch, Stuart Jay and Lu Ann Sims
1978 An identification algorithm for dynamic intervention mo-
deling with application to gun control. Series no. J-78-
29, Georgia Institute of Technology, Atlanta 30332. Mimeo-
graphed.
Hay, Richard A., Jr. and Richard McCleary
1979 Box-Tiao time geries models for impact assessment: A com-
ment on the recent work of Deutsch and Alt. Evaluation
Quarterly 3(2,May):277-314.
Deutsch, stuart Jay
1979 Lies, damn l.i.es and statistics: A rejoinder to the comment
by Hay and McCleary. Evaluation Quarterly 3(2, MaY):315-
328.
CRIME: Homicide, assault with a gun, armed robbery.
TIME, PLACE: Boston, Jan. 1966 to Oct. 1975 for Deutsch and Alt
and Hay and McCleary; Jan. 1966 to Sept. 1977 for
Deutsch and Sims.
METHOD: Stochastic modelling.
CRITERIA: Simplest model with no significant autocorrelation
in residuals.
RESULTS: All analysts agree that homicide is not seasonal.
ARIMA models for assault and robbery found by
Deutsch and Alt and Deutsch and Sims:(0,1,1)(0,1,1)
Hay and McCleary, using the same method and data,
find a (0,1,1)(0,0,1) model for assault, and
log (0,1,1)(0,0,1) for robbery.

Edgerton, Julie, Linda Phelps,Karen Boley-Chang,Constance Osgood

1978 Ecology of Rape, Kansas City Metropolitan Area: Summary Re-
port of the Rape Data Bank. Institute for Community Stud-
ies, University of Missou)'i, Kansas City. Report prepared
for the Metropolitan Organization to Counter Sexual
Assault.
CRIME: Repes known to the police.
TIME, PLACE: Kansas City, Missouri, Kansas City, Kansas, and
Independence, Missouri, 1971 and 1975 (two years).
MErrIWD: Inspection of monthly data.
CRltrERIA: Not specified.
RESULTS: "No definite seasonal pattern. II

35
Federal Bureau of Investigation
1981 Crime in the United states: Uniform Crime Reports 1980.
U.S. Department of Justice, Washington, D.C. 20525.
CRIME: Index offense rates, agencies reporting all 12
months of each year.
TIME, PLACE: 1971-1980, United States.
METHOD: Quarterly data, expressed as a relative crime rate
(proportion of the first quarter, 1971). Ratio-to-
moving average for each quarter, averaged over 10
years into a "seasonal index. II
CRITERIA: Not specified.
RESULTS: Seasonal indices (see method), by quarter, are:
Q1 Q2 Q3 Q4
Murder: 95.0 95.4 104.2 105.5
Forcible Rape: 85.9 101.5 116.8 95.8
Robbery: 97.5 89.7 101.0 111.7
Agg. Assault: 88.6 103.4 110.8 97.2
Burglary: 96.1 94.6 104.8 104.4
Larceny/Theft: 88.5 102.5 109.1 99.9
Motor Veh. Theft: 91.7 98.0 106.6 103.8

Ku, Richard and Bradford Smith

1977 First Year Evaluation of the Illinois Urban High Crime Re-
duction Program: Final Report. Manuscript. Abt Associates,
Inc., Cambridge, Massachusetts.
1978 Second Year Evaluation of the Illinois Urban High Crime Re-
duction Program: Final report. Manuscript. Abt Associates,
Inc., Cambridge, Massachusetts.
CRIME: Residential burglaries and robberies known to the
police.
TIME, PLACE: Jan. 1972 to June 1978 in three Illinois cities:
Peoria, Champaign, and Joliet.
METHOD: Ratio-to-moving-average.
CRITERIA: Not specified.
RESULTS: Not specified, but subsequent analysis uses season-
ally adjusted data.

Lamp, Rainer
1983 Jahreszeit und Kriminalitat. (Time of year and criminality)
Paper presented at the International Congress on Criminol-
ogy, Vienna. Max-Planck-Institut, Freiburg.
CRIME: Total crime, theft, rape, murder.
TIME, PLACE: Federal Republic of Germany, 1971-1980.
METHOD: Census X-11, stochastic modelling.
CRITERIA: Percent contributions of three components, accuracy
of 1981 forecasts, simplest model with no signifi-
cant autocorrelations in residuals.
RESULTS: Both methods agree: Murder not seasonal, but theft,
rape, and total crime are seasonal (strong moving
average effect).
36
 -
.
....

Marshall, Clifford W.
1977a Application of Time Series Methodology to Crime Analysis.
The Polytechnic, Institute, 33 Jay St '. Brooklyn, 11201. Law
Enforcement Assistance Administration Grant #76-TA-99-0028
1977b The State Space Forecasting Technique Applied to Reported
Crime Data. Supplement to 1977a, above.
CRIME: Robbery, aggravated assault, burglary, rape.
TIME, PLACE: "District 1" of Cincinnati, 1968-1974.
ME'rHOD: Census X-II.
CRITERIA: Stable seasonality F of 2.41, "reasonable" graphs
of three components.
RESULTS: Robbery seasonal with December consistently high.
Aggravated assault seasonal (F = 2.78). Burglary
and rape show no seasonal effects.

Michael, Richard P. and Doris Zumpe

1981 Sexual violence in the United States and the role of sea-
son. The American Journal of Psychiatry 140(7,July):
883-886.
CRIME: Rates of forcible rape, aggravated assault, robbery,
murder.
TIME, PLACE: 1975-1979 (five years): Alabama, Arizona, Georgia,
Honolulu, Illinois, Los Angeles, Maine, New Mexico,
North Carolina, Oregon, Puerto Rico, San Francisco,
South.Carolina, Tennessee, Texas, Utah.
METHOD: Harmonic analysis of crime rates and monthly mean
temperature, separate analyses for each crime and
place. Locations ranked by latitude.
CRITERIA: Significance of relationships between crime, lati-
tude, and temperature, Spearman rank correlation.
RESULTS: Assault seasonal in 12 locations, rape in 14, rob-
bery in five, murder in one. Latitude had no ef-
fect, but temperature was significant for assault
and rape.

Munford, Robert S., Ross S. Kazer, Roger A. Feldman, and Robert

R. Stivers
1976 Homicide trends in Atlanta. Criminology 14(2,August):213-
232.
CRIME: Criminal homicide, as determined by investigating
authQritie~. Justifiables not included.
TIME, PLACE: Residents of Atlanta killed in Atlanta or in DeKalb
County, I961-1962, 1971-1972 (four years.)
METHOD: Crosstabulations.
CRITERIA: Chi square; p< .01.
RESULTS: IINo seasonal trend in the occurrence of homicide was
observed in either time period. II

37
 Pittman, David J. and William Handy
1964 Patterns in criminal aggravated assault. Journal of Crimi-
nal Law, Criminology, and Police Science 55:462-470.
CRIME: Aggravated assaults known to police.
TIME, PLACE: St. Louis, 1961. Note: one year only.
METHOD: Crosstabulations.
CRITERIA: Chi square.
RESULTS: No seasonal pattern in zero-level tables, no rela-
tion between indoor-outdoor location and season.

Schneider, Anne L. and David Sumi

1977 Patterns of Forgetting and Telescoping in LEAA Survey Vic-
timization Data. Institute of Policy Analysis, 777 High
Street, Suite 222, Eugene, Oregon 97401.
CRIME: Burglary victimization, National Crime Survey.
TIME, PLACE: Eighteen cities (all available cities), 10 months
of 1972, 1973, or 197j~ (one year per city).
METHOD: "Seasona1 Distinctiveness" (range of mean monthly
temperature for year of survey), recall pt. tterns
(telescoping and forgetting), and season in which
survey was fielded.
CRITERIA: Correlations, inspection of graphs.
RESULTS: Seasons influence memory. More crimes are reported
to victim survey interviewers in the summer. This
is especially true of crimes that the victim had
not reported to the police.

Stein, Donald P., Jay-Louise Crawshaw and Algrid R. Barskis

1967 Computer-Aided Crime Prediction in a Metropolitan Area.
Technical Reports 1-202 and 1-202-A, The Franklin Institute
Research Laboratories, Philadelphia.
CRIME: Part I Index offenses, 5 percent sample.
TIME, PLACE: Philadelphia SMSA, 1966. Note: one year only.
METHOD: Multiple regression. Other variables: weather,
time of day, day of week, phase of the moon.
CRITERIA: Probability that a certain type of crime would
occur, given that some crime did occur.
RESULTS: Criteria render results irrelevant.

United States, Bureau of Justice Statistics

1980 Crime and Seasonality. National Crime Survey Report SD-
NCS-N-15,NCJ-64818. Report written by Richard W. Dodge and
Harold R. Lentzner, Crime Statistics Analysis staff, Center
for Demographic Studies, U.S. Bureau of the Census.

Ir 38
CRIME: Household larceny (total, and more and less than
$50), personal larceny without contact (total, and
more and less than $50), residential burglary (to-
tal, forcible entry, and unlawful entry), motor ve-
hicle theft, assault (total, aggravated and simple),
nonconunercial robbery. Crimes occurring in IIseries li
(recurring offenses) not included. National Crime
Survey incident-level data.
TIME, PLACE: United states, 1973-1977 (five years), persons aged
12 and older.
METHOD: Census X-II.
CRITERIA: Stable seasonality F of 10 or more. F 2.34 - 9.99
considered Ilmerely indicative of seasonality. II Per-
cent contribution of seasonal component, and sea-
sonal factors.
RESULTS: Household larceny F=51.75 (less than $50, F=22.61;
more than $50, F=27.65); personal larceny F=25.08
less than $50, F=43.50; more than $50, F=15.36);
residential burglary F=22.98 (forcible entry
F=7.04; unlawful entry F=35.96);
motor vehicle theft F=7.09; assault F=7.52
(aggravated F=5.80; simple F~4.79); robbery F=2.15.

Warren, Charles W., Jack C. Smith and Carl W. Tyler

1981 Seasonal variation in suicide and homicide: A question of
consistency. Unpublished manuscript. Public Health Ser-
vice, U.S. Centers for Disease Control, Atlanta, 30333.
CRIME: Homicides, except for deaths resulting from legal
intervention, and suicides of people over age 14,
National Vital Statistics Mortality files.
TIME, PLACE: United States, 1969-1978.
METHOD: Periodic regression analysis (PRA). See Bliss
(1958, 19'70).
CRITERIA: Goodness of fit of PRA model.
RESULTS: Both homicide and suicide have significant monthly
patterns within each year. For homicide, however,
the pattern differs from year to year. December,
for example I is a Ilpeak month ll in some years and a
Iitrough month in other years. Suicide is consis-
ll

tently high in the spring and low in the winter.

Wolfgang, Marvin E.
1966 Patterns in Criminal Homicide. New York: John Wiley & Sons
CRIME: Homicide, by race and sex of victim. Number of
multiple suspects arrested for criminal homicide.
TIME, PLACE: Philadelphia, 1948-1952 (five years); United States,
1930 and 1950.
METHOD: Proportion of total homicides, aggregated over the
five years, occurring in each month.

39
CRITERIA: Inspection of table.
RESULTS: Rejected "hypothesis (of) a relationship between
monthly or seasonal changes and variations in crimi-
nal homicide. II For the United States, 1930 and 1950
September is a high month. Arrests of multiple
suspects show greater seasonal fluctuation.

40
 Survey of the States, 197~

In late 1979, the Illinois Law Enforcement Commission

(predecessor to the Authority) surveyed each state, asking the
director of each criminal justice planning agency the following:
Has anyone analyzed reported Index crimes in your
state to determine whether they increase and decrease
with the seasons? If so, we would appreciate a copy
of the publication, or a summary of the findings if
they have not been published.
We received responses from 17 states and the District of
Columbia. Six of these states said that they had not analyzed
seasonality. The findings of the other respondents appear below.

Arizona
CRIME: Property crimes (burglary, larceny/theft, and motor
vehicle theft).
TIME, PLACE: Arizona, by county, 1975 to 1977 (three years).
METHOD: Mean across three years of number occurring in each
month. Northern counties and Maricopa and Pima
Counties analyzed separately.
CRITERIA: Inspection of graph.
RESULTS: Number of property offenses varies with the tourist
season (northern counties high in summer months,
Maricopa and Pima high in winter months).
California
CRIME: Each Index crime (willful homicide, forcible rape,
robbery, aggravated assault, burglary, theft, motor
vehicle theft).
TIME, PLACE: California, 1974-1978 (five years).
METHOD: Graphs of number of offenses known to police each
month over the five-year period.
CRITERIA: Inspection.
RESULTS: Homicide: decreases each February, many fluctuations
from month to month. Rape: high July-October, low
January-February. Robbery: high December, low May-
June. Assault: peak in summer or fall. Burglary:
high in fall and winter, lowest in June. Theft:
September low in all years, but pattern inconsistent
for other months. Motor vehicle theft: no consis-
tent pattern for all years.

41
Delaware
CRIME: Each Index crime.
TIME, PLACE: Total Delaware, and Sussex County, 1976-1978 (three
years) 0

METHOD: Ratio-to-moving-average.
CRITERIA: Not given.
RESULTS: Homicide, burglary, and motor vehicle theft not sea-
sonal. Rape, robbery, assault, and larceny seasonal.
Sussex County, which has a lot of tourism, has large
seasonal fluctuations.
Iowa
CRIME: Motor vehicle theft, by type of vehicle.
TIME, PLACE: Iowa, 1975-1978 (four years).
METHOD: Graph of each type of motor vehicle theft.
CRITERIA: Inspection of graph.
RESULTS: Decrease of total motor vehicle theft in winter, due
entirely to a decrease in theft of motorcycles.
Kentucky
CRIME: Each Index crime.
TIME, PLACE: Total state, 1973-1978 (six years). For homicide,
1978 only.
METHOD: Comparison of high and low months.
CRITERIA: Inspection.
RESULTS: 1978 homicide high in late spring and summer. Rape
consistently high in summer, low in January. Rob-
bery highest in final quarter of year. Assault high
May to September, low January, February. Breaking
and entering slightly more frequent in second half
of year. Larceny consistently high June to August.
Auto theft high July to August, low January-
February.
Maine
CRIME: Each Index crime.
TIME, PLACE: Maine, 1975-1978 (four years).
METHOD: Graphs of number of offenses known to police per
month.
CRITERIA: Inspection.
RESULTS: Homicide, rape, robbery not seasonal. Assault high
May-October. Burglary high October-November. Lar-
ceny/theft strongly seasonal, high June-October.
Motor vehicle theft high July-October. Seasonality
may be an artifact of population fluctuation. Total
Index crime rates, counting residents plus tran-
sients in the denominator, is not seasonal.

42
New Jersey
CRIME: Violent Index crime (homicide, rape, robbery,
assault); nonviolent Index crime (burglary, larceny/
theft, motor vehicle theft)
TIME, PLACE: New Jersey, 1975-1977 (three years).
METHOD: Graphs of crimes knOffi1 to the police.
CRITERIA: Inspection.
RESULTS: Violent crime not seasonal. Nonviolent crime high
June-September, low January-February.
North Carolina
CRIME: Each Index crime.
TIME, PLACE: North Carolina, 1977-1978 (two years).
ME'l'HOD: Graphs of crimes known to the police.
CRITERIA: Inspection.
RESULTS: Murder and motor vehicle theft are not seasonal.
Rape and aggravated assault: high in summer.
Robbery, burglary high in December; larceny/theft
high in August.
Virginia
CRIME: Personal Index crime (homicide, rape, assault);
property Index crime (robbery, burglary, larceny/
theft, motor vehicle theft).
TIME, PLACE: Virginia, 1975-1978 (four years).
METHOD: Graphs of crimes known to the police.
CRITERIA: Inspection.
RESULTS: Personal crimes high June-August; property crimes
high July-August.
Washington
CRIME: Homicide.
TIME, PLACE: Washington, 1958-1962 (five years).
METHOD: Graph.
CRITERIA: Inspection.
RESUL'rS: No seasonal pattern.
WaShington, D.C.
CRIME: Violent and property Index crime.
TIME, PLACE: District of Columbia, 1977 and 1978 (two years).
METHOD: Graphs.
CRITERIA: Inspection.
RESULTS: Violent crime not seasonal. Property crime peaks in
August.

43
 BIBLIOGRAPHY

Bliss, C. I.
1958 Periodic regression in biology and climatology. Connecti-
cut Agricultural Experimental Station Bulletin 615:1-56.
1970 Statistics in Biology. Vol. II. New York:McGraw-Hill.
Block, Carolyn Rebecca
1983 How to Handle Seasonality: Introduction to the Detection
and Analysis of Seasonal Fluctuation in Criminal Justice
Time Series. Statistical Analysis Center, Illinois
Criminal Justice Information Authority.
Block, Carolyn Rebecca and Louise S. Miller
1982 Manual for the Pattern Description of Time Series. Chicago:
Statistical Analysis Center, Illinois Criminal Justice
Information Authority.
Block, Carolyn Rebecca, Craig McKie, and Louise S. Miller
1983 Patterns of change over time in Canadian and United States
Homicide. Policy Perspectives 3(2):121-180.
Block, Carolyn Rebecca and Richard L. Block
198~ Crime definition, crime measurement, and victim surveys.
Journal of Social Issues ~O(Spring).
Block, Richard L.
1977 Violent Crime. Lexington, Massachusetts: D,C. Heath & Co.
Brearley, Harrington Cooper
1932 Homicide in the United States. Chapel Hill: The University
of North Carolina Press.
Coldren, James R., Jr.
1980 Aggregation problems in the analysis of Illinois state-
wide data. Statistical Analysis Center, Illinois Criminal
Justice Information Authority.
Kallek, Shirley
1978 An overview of the objectives and framework of seasonal ad-
justment. Pp. 3-25 in Zellner (ed) Seasonal Analysis of
Economic Time Series. Proceedings of the Conference on the
Seasonal Analysis of Economic Time Series, September 9-10,
1976, U.S. Department of Commerce, Bureau of the Census,
Economic Research Report ER-1.
LEAA (Law Enforcement Assistance Administration)
1972 San Jose methods test of known crime victims. Statistics
Division Technical Report # 1, Washington, D.C.
Lester, David
1972 Why People Kill Themselves. Springfield, Illinois: Charles
Thomas.
~~
 McCain, Leslie J. and Richard McCleary
1979 The statistical analysis of the simple interrupted time-
series quasi-experiments. Pp. 233~293 in Quasi-experimen-
tation: Design and Analysis Issues for Field Sett~ngs, by
Thomas D. Cook and Donald T. Campbell. Chicago: Rand
McNally.
. McCleary, Richard, and Richard A. Hay, Jr .
1980 &~plied Time Series Analysis for the Social Sciences.
Beverly Hills:Sage Publications.
Miller, Louise S. and Carolyn Rebecca Block
1983 Illinois Murder Victim Data, 1973 to 1981: Guide to Qua-
lity, Availability and Interpretation. Illinois Criminal
Justice Information Authority.
Parker, Robert Nash and M. Dwayne Smith
1979 Deterrence, poverty, and type of homicide. American Journal
of Sociology 85 (3) : 6J)~-624 (special volume on Gun Control,
Philip J. Cook, ed.).
Pierce, Glenn L. and William J. Bowers
1981 The Bartley-Fox gun law's short-term impact on crime in
Boston. The Annals of the American Academy of Political
and Social Science 455(May):120-137.
President's Commission on Law Enforcement and the Administration
of Justice
1967 The Challenge of Crime in a Free Society. U.S. Government
Printing Office.
Quetelet, Adolphe
1842 A Treatise on Man and the Development of his Faculties.
English translation 1968. New York: Burt Franklin.
Sutherland, Edwin H. and Donald R. Cressey
1978 ~riminology. Philadelphia: Lippincott.
Sylvester, Sawyer F.
1982 Adolphe Quetelet: At the beginning. Federal Probation 46
(December,4):14-19.
Wolfgang, Marvin E.
1966 Patterns in Criminal HOlnicide. New York: John Wiley & Sons
Zimring, Franklin E.
1979 Determinants of the death rate from robbery: A Detroit time
study. Pp. 31-50 in Harold M. Rose (ed.) Lethal Aspects of
Urban Violence. Lexington, Massachusetts: Lexington Books.

45