- Email: [email protected]
Violent deaths in the United States, 1900–1975
SOL-. SC;. Med. Vol. 16, pp. 1929 to 1938. 1982 Printed in Great Britain
VIOLENT
0277-9536/X2;‘?21929-10603.00/O Pergamon Press Ltd
DEATHS
RELATIONSHIPS
IN THE 1900-1975
UNITED
STATES,
BETWEEN SUICIDE, HOMICIDE ACCIDENTAL DEATHS
PAUL C. HOLINGER’*
and
AND
ELAINE H. KLEMEN~
‘Institute for Psychosomatic and Psychiatric Research and Training, Michael Reese Hospital and Medical Center. Chicago. IL 60616 and ‘Community Mental Health Center, Northwestern Memorial Hospital. Chicago. IL. U.S.A. Abstract-Time trends are presented for suicide. homicide and accident mortality rates in the United States. 1900-1975. These data suggest that national mortality rates for suicide, homicide and motorvehicle accidents tend to be parallel over time. Non-motor-vehicle accidents, while showing some Ructuations similar lo those of other forms of violent deaths. manifest a more general decrease throughout the century. In addition. suicide rates tend to be significantly correlated over time with homicide. motorvehicle and non-motor-vehicle accident death rates for most race and sex combinations. These results need to be viewed with caution due to the methodologic problems inherent in using national mortality data. However. the findings raise serious questions about the traditional understanding of the relationships between the various forms of violent death in populations; in particular, the popular view that suicide and homicide are inversely related in populations is questioned. The explanatory usefulness of conceptualizing violent deaths as reflecting self-destructive tendencies is discussed.
INTRODUCTION
Controversy exists regarding the nature of the relationships between suicide, homicide and accidental deaths on both individual and population bases. Furthermore, it is not firmly established whether or not homicide and accidental deaths. along with suicide, reflect self-destructive tendencies. Many studies have addressed these issues from varying perspectives and using several different kinds of data. e.g. psychoanalytic, sociologic and epidemiologic. The present study utilizes population data (national mortality data from the United States. 1900-1975) to examine these questions. The purpose of this study is two-fold. First, the time patterns of suicide, homicide, motorvehicle accidents and non-motor-vehicle accidents will be compared in order to better understand the relationships of these forms of death in large populations. Second. these data will be used to address the question as to whether or not homicides and accidental deaths. along with suicide, may reflect self-destrucwork was. in part. done at and supported by the Harvard University School of Public Health, Boston, Massachusetts and, was also supported by the Judith Baskin Offer Research Fund. The authors gratefully acknowledge the assistance of Leo Levy (Univeriity of Illinois School of Public Health. Chicago. IL). Jane Murphy (Harvard Medical School. Boston. MA) and Daniel Offer (Michael Reese Hospitai and Medical Center. Chicago. IL) in the preparation of this report. The authors also gratefully acknowledge the assistance of Eric Ostrov (Michael Reese Hospital and Medical Center. Chicago. IL) in the statistical analysis. *Send reorint requests to: Dr Paul C. Holinger Psychosomatic-and Psychiatric Institute. Michael Reese Hospital and Medical Center. 2959 South Cottage Grove. Chicago. IL 60616. U.S.A.
This
1929
tive tendencies on a population level: the strategy employed is to compare the epidemiologic patterns of overt self-destructiveness (i.e. suicide) with homicides and accidental deaths. Theoretically, this work necessitates some bridging of intrapsychic disciplines involving individuals. e.g. clinical psychiatry and psychoanalysis (from which the hypotheses are derived), and fields which deal with populations, e.g. sociology and epidemiology (in which the hypotheses are tested). Such an attempt must be made with great caution, however. due to the serious conceptual problems involved in trying to understand the data and findings of one scientific discipline by using models derived from data of a very different scientific discipline. Gedo and Goldberg [I] have cogently discussed this potential for error in applying metapsychological propositions about intrapsychic phenomena to social psychology and vice versa. Why have suicides, homicides, and accidental deaths been studied together in the scientific literature’! The term ‘violent deaths’ has been used to include this combination of suicide. homicide and accidents (e.g. Weiss [2]), and Wehsrer’s [S] defines the term ‘a violent death’ to mean “caused by force: not natural”. Thus. the first reason probably related to the sense that these causes of mortality are often considered to be ‘unnatural’ as compared with. e.g. cancer or heart disease: violent deaths do not seem to involve the deterioration of the body’s internal organs. but rather seem somehow imposed from without. Second. clinical evidence suggests that homicides and accidents. as well as suicides. may be Selfinflicted: the victim to some degree provokes his or her own death by ‘being in the wrong place at the wrong time’. The-idea of risk-taking is relevant here. mith people who take more risks being in greater danger of dying by accident or homicide (homicide
P.ALL C. HOLISGER and ELAISE H. KLESIES
19Y.l
referring to those who are killed. not the killers). Suicide, homicide and accidents. in this framewcrk. are seen as related in that all ma) represent some expression of self-inflicted mortality. with suicide being most overt and homicide and accidents being more subtle manifestations of self-destructive tendencies and risk-taking. This paper is focused in two specific ways. First, the ph,=nOmenolOgiC rather than motivational aspects Of self-destructive behaviors will be examined: the intrapsychic models regarding the motivations underlying such self-destructiveness (e.g. masochism, narcissism and grandiosity, etc.) will be examined in future communications. Second, this paper is focused on self-destructive rather than aggressive behaviors, i.e. it is victim-oriented rather than perpetratororiented; however, this decision to study human behavior from a perspective of self-destructiveness in no way implies any intention to deny the importance of aggression nor imply a lessening of the responsibility of the perpetrator of the crime or accident. Clinically, the victim-oriented approach is optimistic in that it implies that people can, by recognizing and controlling unconscious self-destructive tendencies (whatever the origins), take greater control over their lives and more actively avoid misfortunes which have previously been perceived as unavoidable externallyoccurring events. LITERATL RE
There is. of course. a massive literature dealing with suicide. homicide. and accidents from both intrapsychic and population perspectives. This literature review will be limited to those studies which address the two issues under consideration: (1) the relationships between suicide, homicide and accidents; and (2) the question as to whether or not suicides, homicides and accidents may represent self-destructive tendencies. There are many excellent studies which separately discuss suicide [4-81, homicide [9-l I], or accidents [ 12-161 from either intrapsychic or sociologic perspectives or both. as well as many epidemiologic studies which deal primarily with the frequency and distribution of violent deaths [Z, I I, 17, 18-j. However, few studies explicitly examined the relationships among these causes of mortality on a population level. Perhaps the best known work which investigated these relationships was Henry and Short’s 1954 Slricidr curd Homicide [19]. Utilizing data which included suicide, homicides and aggravated assaults for specific localities over time spans of a relatively few years. the authors found that homicide and aggravated assault correlated positively with the business cycle, in contrast to suicide. They stated that “while suicide increases during (economic) depression, homicide decreases. While suicide drops during prosperity homicide increases” [ 191. Porterfield’s [ZO] lessknown but important work involved studies of metropolitan areas and found that their combined suicidehomicide rates correlated positively with their traffic fatality rates. i.e. he demonstrated that cities with a high suicide-homicide rate tended to have high traffic fatality rates. Porterfield suggested that persons involved in suicide. homicide, and traffic fatalities
were similar in having “little regard for the& own lives or the lives of others. or both”. Although Porterfield did not separate the suicide. homicide and motorvehicle accident rates. his assumptions. type of data and conclusions clearly differed from those of Henry and Short. Brenner’s [21] conclusions also seem to differ from Henry and Short’s findings. Brenner showed that indicators of economic instability and insecurity, such as unemployment and unusually rapid economic growth rates. were associated over time with higher mortality rates. Specifically. data for the entire United States showed that indices of acute pathological disturbances, including suicide and homicide. rose within a year of increased unemployment rates. Klebba [22] described trends for both suicide and homicide in the United States and her data also seemed to support a positive rather than inverse relationship between suicide and homicide over time. There is a large body of literature which addresses the question as to whether accidents and homicides. in addition to suicide, represent self-destructive tendencies. Regarding accidents. Freud [ 121 emphasized the role of psychic determinism in human behavior and, using clinical data. attempted to demonstrate that accidents were unconsciously precipitated by the victim. Menninger [4] and Farberow’s [23] work has been particularly convincing in clinically documenting the varied and ubiquitous nature of self-destructive tendencies underlying accidents and apparently externally-precipitated misfortunes. Several studies using small group data have been conducted to address the issue of accident-proneness. i.e. that certain people are more liable to have accidents. Greenwood and Woods [24] wrote an historically important paper in 1919. reporting that in a British munitions factory a relatively small number of workers had most of the accidents. The authors cautiously suggested that some people were inherently more likely to have accidents. Tabachnick [ 131 has carefully summarized the studies exploring this hypothesis and the controversies surrounding it. Some data suggest that deaths due to accidents. including motor-vehicle. may be due to accident-proneness and risk-taking that reflect depression and suicidal tendencies [15. 161; other investigations do not tend to support the hypothesis that suicidal tendencies were predominant among those involved in accidents [ 13, 14,251. To summarize. both clinical data and studies of small groups seem to suggest that the concepts of accident-proneness and excessive risk-taking have therapeutic and predictive value and that accidents and accidental deaths may be psychologically determined by underlying self-destructive tendencies: however. testing the hypothesis that accidents reflect unconscious self-destructive tendencies is extremely difficult and fraught with methodologic problems and several studies do not confirm the hypothesis. The most important work relating homicide to selfdestructive tendencies has been done by Wolfgang [9]. Studying the records of several hundred persons murdered in Philadelphia, Wolfgang found that more than 25”d of those people could be said to have provoked their own deaths. He spoke of this group as having committed “suicide by means of victim-precipitated homicide”. Wolfgang’s study supported clini-
Violent deaths in the United States, 1900-1975 cal notions that some patients could bring trouble upon themselves by not taking care of themselves and by being “in the wrong place at the wrong time”. METHODOLOGIC
CONSIDERATIONS
Serious methodologic problems exist in using national mortality data to study suicide, homicide, and accident trends in populations. These difficulties include underreporting, overreporting, misclassification and changes in classification over time. These issues have been discussed in detail elsewhere [26-341 and the changes in federal classification over time have been outlined by government reports. There has been little change over the century in federal classification for suicide. homicide and motor-vehicle accidents; accidents excluding motor-vehicle accidents have shown a slight tendency toward a decrease in the number of deaths listed as non-motor-vehicle accidents over the sixth (1949-1957). seventh (1958-1967) and eighth (1968 to date) Revisions of the International Classification of Diseases*. The data Rresented here include deaths that occurred in all U.S. states which participated in federal registration of deaths (death registration states), 19OGl932 and deaths which occurred in all U.S. states 1933-1975 (with Alaska added in 1959 and Hawaii in 1960). Data in this paper will be presented in two time periods. 1900-1975 and 1933-1975, inasmuch as (1) it was only in 1933 that data for the U.S.A. included the entire population of the U.S.A.; and (2) motor-vehicle accident deaths slowly increased in prominence from the early 1900s as the use of automobiles became widespread and only in the late 1920s and early 1930s did motor-vehicle accident death rates attain the general levels maintained until the present. RESULTS
Figure 1 presents the age-adjusted mortality rates for suicide. homicide. motor-vehicle accidents, nonand total accidents, accidents motor-vehicle 190&1975, United States. Of particular interest in
*Comparability ratios for the sixth. seventh and eighth Revisions were. respectively. for suicide 1.00. 1.03 and 0.9472: for homicide 1.00. 1.00 and 0.9969; for motorvehicle accidents 1.00. 1.00 and 0.9921; for accidents excluding motor-vehicle accidents 0.93.0.95 and 0.9250; and for accidents (total) 0.95. 0.97 and 0.9570. Comparability ratios were not calculated by the government prior to the sixth Revision because of the nature of the available data and the difficulty quantifying the classification changes. The comparability ratios and discussions of these changes can be found in References [26-341. Comparability ratios are calculations of the extent to which mortality rates for a cause of death may differ from one Revision to the next solely on the basis of reclassification of deaths. A comparability ratio of 1.00 indicates. usually. that the number of deaths attributed to a cause of mortality in the later Revision is essentially the same as for the earlier Revision: a ratio of less than 1.00 indicates that fewer dea’ths were attributed to the later Revision. while a ratio of greater than 1.00 indicates that more deaths were attributed to the later Revision due to classification changes.
1931
examining the relationship between the various forms of violent deaths are: the increases in suicide, homitide, motor-vehicle accidents and non-motor-vehicle accidents during the economic depression in the earlv 1930s; the decreases in suicide, homicide and motorvehicle accidents during World War II in the early 1940s; the slight increases in suicide, homicide and motor-vehicle accidents in the late 1940s and early 1950s; and the increases in suicide, homicide and motor-vehicle accidents throughout the 1960s. Nonmotor-vehicle accidents have tended to decrease throughout the century (with some of this decrease over the past 30 years being artifact due to classification changes-see Methodology); they show the increase during the first decade of the 1900s and the decreases during World War I which characterize suicide, as well as the increase during the early 1930s that characterizes all forms of violent deaths. but they increase in the early 1940s unlike the other causes of mortality. The relationship between suicide and homicide can most clearly be seen in Fig. 2, which presents their age-adjusted time trends 1900-1975 in greater detail. Before 1933, the major difference is that the suicide rate is biphasic (peaking around 1910 and 1930 and decreases during World War I which characterize suicide, as well as the increase during the early 1930s 1933, the rates are rather parallel: peaks in the early 1930s decreases in the early 1940s. slight increases in the late 1940s and early 1950s and larger increases in the 1960s and 1970s. Figures 3-6 present the age-adjusted violent death trends by race and sex for 1914-1975 (data for both race and sex are only available from 19 14 on). The rather parallel nature of the fluctuations in mortality rates is especially noticeable for suicide, homicide and motor-vehicle accidents. Correlations are also presented for the age-adjusted mortality time trends and they are separated into the years 1900 (or 1914 if race and sex are used) to 1975 (Table I) and 1933-1975 (Table 2). Of particular concern in addressing the second purpose of this paper (re: the self-destructive nature of violent deaths) are the correlation coefficients of suicide with the other forms of violent death. For example, for 1900-1975. suicide, when correlated with homicide, motor-vehicle accidents, non-motor-vehicle accidents and total accidents, showed coefficients of +0.22 (P < 0.05). -0.46 (P < 0.001). +0.63 (P < 0.001) and 0.71 (P < 0.001) respectively. For 1933-1975, suicide correlated +0.58 (P < 0.001). +0.53 (P < 0.001). +0.54 (P < 0.001) and +0.66 (P < 0.001) with homicide. motor-vehicle accidents, non-motor-vehicle accidents and total accidents respectively. The inverse correlation between suicide and motor-vehicle accidents from 1900-1975 appears to result from the continuous increase in motor-vehicle deaths from 1900 to 1975 when cars were just beginning to appear in’the population and assume prominence as a cause of mortality. Once the use of automobiles had become widespread. the time trends showed a significantly positive correlation between suicide and motor-vehicle accidents. 1933-1975. When the data are broken down into race and sex differences (19141975 and 1933-1975). the correlation patterns between suicide and the other forms of violent death are maintained with the follow-
P.ACL
1932
C.
and
HOLISGER violent
EL.AISE
H.
KLEMES
deaths Sunde Homude Motor-vahlcle Non-motor Totol
accidents vsn~cle occldents
Year
Fig. I. Age-adjusted
violent death rates by type of mortality. United States, 1900-1975. Sources of data: Reports 43, (for 1900-1953); Grove R. D. and Hetzel A. M. Vitul Sratisrics Rates in the United Stutes: 1940-1960. U.S. Government Printing Office, 1968 (for 19541960): Vird Srutistics in rhr United Stutes. Morraliry, 1961-1975 (for age-adjusted motor-vehicle accident and non-motorvehicle accident mortality rates, 1961-1975); Vito/ urtd ffedth Sturisrics. Series 20. Number 16. U.S. Government Printing Office, 1974 [for age-adjusted accident (total), suicide, and homicide mortality rates. 1970-19751.
Viral Stabstics-Speciul
20
r
I
1900
1910
1
I
,920
1930
homicide
rates,
8
I
1940
1950
W
Sumde
t.
Homictde
I
1960
I
,
1970 1975
YWr Fig. 2. Age-adjusted
suicide
and
United Fig. I.
States,
1900-1975.
Sources
of data:
same
as
Violent
deaths
in the United
States.
190~1975
1933
White male
Suude Homwde Motor-vehicle accidents bg;motor-vehxle occldents
Fig. 3. Age-adjusted
violent
death
rates. white males, United Fig. 1.
States.
1914-1975.
Sources
of data:
same as
White female o-o s-e
n-n
Suicide Homicide Motor-vehicle accidents Non-motor- vehicle accidents Total
~--1
i\
e-e
Year
Fig. 4. Age-adjusted
violent
death
rates. white females. United as Fig. I.
States.
1914-1975.
Sources
of data:
same
PALL C. HOLI~GER
1934
and ELAISE H. KLESIEY
NonwhIte male
e-9 w
Sulclde Hornlade Motor-vehicle
n-¤Non-motor-vehicle .-. Total
Fig.
Age-adjured
violent
death
rates,
nonwhite males. United same as Fig. I.
accidents accidents
States.
of data
1914-1975.
Nonwhite female -
Fig. 6. Age-adjusted
violent
death
rates.
nonwhite females. United same as Fig. I.
Suade Homude Motor-vehtcle occldents Non-motor-vehicle accidents Total
States.
1914-1975.
Sources
of data:
Violent deaths in the United States. 19Wl975
1935
Table 1. Correlation coefficients for violent death rates (age-adjusted). United States, 1900-1975 Total
violent
H 0.228
S
S H MVA NMVA TA
Violent S
Violent S
S H MVA NMVA TA
190&l
MVA - 0.4?* 0.39***
deuths.
H 0.72***
S H MVA NMVA TA
deaths,
white female.
MVA 0.001 o.ost
deaths, nonwhite
Violent
1914-I
female,
TA 0.71*** 0.17t -0.55*** 0.93***
S
TA 0.55*** 0.16t -0.15t 0.94***
S
975
NMVA 0.49*** 0.12t -0.47***
deaths,
H 0.67***
Violent
white mole. 1914-1975
MVA -0.08t - 0.26*
deaths, nonwhite
H 0.28*
MVA 0.01 i 0.27’
NMVA 0.56*** 0.49*** -0.66***
male, 1914-I
NMVA -0.02t 0.071 -0.75***
TA 0.66*** 0.49*** -0.33’8 0.92***
975
TA -0.02-l 0.38*** -0.16t 0.76***
1914-1975
NMVA 0.04t - 0.06t -0.75***
MVA 0.05t 0.29*
H 0.33**
975
NMVA 0.63*** 0.38*** -O.Sl***
Abbreviations. S = suicide; H = homicide; TA = total accidents. *P < 0.05; **P < 0.01; ***P < 0.001; tNS. Sources of data: same as Fig. 1.
TA 0.10t 0.08t - 0.43*** 0.91***
MVA = motor-vehicle
accidents;
NMVA = non-motor-vehicle
accidents;
Table 2. Correlation coefficients for violent death rates (age-adjusted), United States, 1933-1975 Total S
S
violent
H 0.58***
H MVA NMVA ylmt
d;ths.
0.68***
S
H MVA NMVA TA Vi;/ent
S
H MVA NMVA TA
deaths,
H 0.53***
deaths.
f933-1975
MVA 0.53*** 0.43***
white female,
MVA 0.56*** 0.55***
nonwhite
NMVA 0.54*** -o&It 0.04t
Violent
TA 0.66*** 0.08t 0.32; 0.96***
S
TA 0.42** -0.16t 0.02t 0.96***
S
1933-1975
NMVA 0.27’ - 0.29* - 0.22t
H 0.58***
VioFt
0.58***
deaths,
white
MVA 0.52*** 0.32*
deaths, nonwhite
MVA 0.42** 0.39**
male, 1933-I
975
NMVA 0.49*** 0.05t 0.15t
TA 0.59*** 0.13t 0.42** 0.96***
male, 1933-I
975
NMVA -0.51*** 0.22t -o&It
TA -0.23t 0.40** 0.48*** 0.85***
female, 1933-1975
MVA 0.61*** 0.15t
NMVA -0.44+** o.OOt -0.42**
TA -0.18t 0.07t o.OOt 0.90***
Abbreviations: same as Table 1. *P < 0.05; **P < 0.01; ***P < 0.001: INS. Source of data: same as Fig. I.
ing major exceptions: for all groups there is no significant inverse correlation between suicide and motorvehicle accidents 1914-1975. probably because much of the initial increase in MVA mortality is excluded: and the nonwhite male and nonwhite female suicide rates. for 1914-1975. sho\v no significant correlation with non-motor-vehicle accidents and. 1933-1975. show a significant inverse correlation with nonmotor-vehicle accidents.
In examining the data by race and sex variables, the following relative comparisons should be noted: male rates for all forms of violent deaths are higher than females; for whites. non-motor-vehicle accidents show the highest mortality rates. followed by motor-vehicle accidents. suicide and homicide, respectively; however, for nonwhites, non-motor vehicle accidents are followed by homicide, motor-vehicle accidents and suicide rates respectively. Age patterns have been
POLL
1936
C. HOLISGER
and ELAN
presented elsewhere [35] and \vill be mentioned onl! briefly: suicide rates tend to increase bvith age for males ~thile female rates are highest between the ages of Jj-64 years: homicide rates are highest among young adults for males and among infants and !oung adults for females: the highest rates of motor-vehicle accident deaths for both sexes are found in teenagers and young adults and the oldest age groups: and the oldest and boungest (infancy) ages are at greatest risk of dying by non-motor-vehicle accidents for both males and females.
.
DISCUSSION
Time trends leave been presented for suicide, homicide and accident mortality rates in the United States. The focus has been on the years 1933-1975 inasmuch as the data include the entire United States population only during those years and at approximately the same time motor-vehicles reached levels they were to maintain as a prominent cause of mortality. These data suggest that mortality rates for suicide, homicide. motor-vehicle accidents and to a lesser extent nonmotor-vehicle accidents tend to be parallel over time. In particular. homicide. motor-vehicle and nonmotor-vehicle accident death rates tend to be significantly correlated with suicide rates over time for most race and sex combinations. At least two specific methodologic artifacts might cause the reported mortality rates for suicide. homicide and accidents to simultaneously increase or decrease. First, federal classification changes must be considered as a possible cause of parallel fluctuations. For example: a person shoots himself (or is shot). does not die immediately. and then dies 2 weeks later of pneumonia, which indirectly results from the shooting. A classification change which shifted the cause of such deaths from pneumonia to suicide (or homicide) would lead to simultaneous increases in the suicide and homicide rates. However, the comparability ratios indicate that no such simultaneous classification shifts have occurred large enough to account for the parallel fluctuations. A second possible artifact which might cause simultaneous fluctuations in the various forms of violent deaths involves increases or decreases in the number of reported deaths based on local or national budget, or political issues. For example. if budgeting were low for health officials. more deaths of ‘undetermined etiology’ might be reported with a general decrease in the specilic forms of violent deaths. Although in this instance the actual trend is opposite (i.e. in times of great economic depression the violent death rates have increased). the example illustrates the potential kind of problem which could lead to simultaneous fluctuations of violent deaths. Due to methodologic problems inherent in using national data, these results must be viewed with caution. However. the data seem to raise serious questions about the traditional understanding of the relationship between the various forms of violent death in populations. particularly with respect to suicide and homicide. The classic view that suicide and homicide rates are inversely related in populations (e.g. Henry and Short [ 193) is not supported by these data:
H. KLFME\
rather. the data seem to support the opposite trndenq. that suicide and homicide rates fluctuate in parallel over time. Brenner‘s [?I] and Klehb:~‘s [X] \\ork also support the parallel Huctuation 01 ,uicidc and homicide rates. One must ash ho\\ the traditianal view that there is an inverse relationship bct\\ccn \LIItide and homicide came to ha\ e such \\ ide acceptance in light of the contradictor! e\ idsnce presented in this paper and elsewhere [?I. 221. First. and perhap mo\t importantly. the data used ti! Hcnr! and Short [ 191 were very ‘different than those utilized in the present article and by Brenner [?I] and Klebba [?I. Hem-b and Short’s [ 191 data were suicide and homicide rates for specific locations (e.g. cities) and for relatively short time spans. crimes against propert! and people. social status. strength of the relational s!strm and measures of the business cycle. The data presented in this paper as well as in those of Brenner [21] and Klebba [Z?Z] differ in two important ways. in that the! are national mortality data (i.e. not a sample but an entire population) and they focus on longer time trends. Such differences in data could account for the contradictory findings. Second, Henry and Short [ 191 conceptualized the suicide-homicide problem from the perspective of aggression: the subtitle of their book is “Some Economic. Sociological and Psychological Aspects of Aggression”. Freud’s intrapsychic understanding of aggression may have influenced Henry and Short’s hypothesis that suicide and homicide were inversely related. Freud [36] suggested that. as a result of frustration, aggression could be directed either externally (possibly resulting in homicide) or turned back on the self (possibly resulting in suicide). Consistent with this. Henry and Short state that suicide and homicide “are alike in the sense that they are both aggressive reactions to frustrations.. [but] become sharply differentiated aggressive acts when we consider the object of the aggression rather than its source”. However. Henry and Short may have a misapplied Freud’s intrapsychic model as they made the shift to population levels. The transition from individuals (from which the hypothesis derives) to populations is complicated by the often-overlooked to those fact that homicide L 1 refer specifically killed (the objects of the aggression). not to the killers (those who when frustrated direct their aggression outward). In other words, homicide data say nothing. strictly speaking, about the aggressors and the nature or amount of aggression. It seems unwarranted to assume anything about the aggression in a population or characteristics of killers based on data which describe only characteristics of people who are killed. Data on suicide and homicide in populations might best be conceptualized as similar in that both measure the object of the aggression. not the source. Having addressed the first purpose of this stud! (the comparison of the time trends and relationships between the various forms of violent deaths in the population) the second question must be dealt with: whether or not homicides :tnd accidents. in addition to suicide. can be said to retlect self-destructive tendencies. The initial strategy was to compare suicide rates (overt self-destructiveness) with homicide and accident rates. Examination of the hypothesis that violent deaths reflect self-destructive tendencies is made especially ditficult due to the different levels of
Violent deaths in the United States, 1900-1975 abstraction involved. The findings that suicide, homicide and motor-vehicle accident rates tend to be parallel over time and that homicides, motor-vehicle accidents and non-motor-vehicle accidents correlate overt self-destructiveness-i.e. significantly with suicide--can perhaps at best be said to suggest that self-destructiveness is an important variable which is associated with various forms of violent deaths on a population level. Within this hypothesis. the differences between non-motor-vehicle accidents and the other forms of violent death would be explained as follows: (1) non-motor-vehicle accidents have enough major fluctuations to correlate positively with suicide (the increase around 1910: the decreases in World War I: the increases in the early 1930s; and the recent ‘leveling’ off. which is in reality an increase which manifests as a ‘leveling off because of the comparability ratio): and (2) the tendency of non-motor-vehicle accidents and motor-vehicle accidents toward an inverse relationship (with the exceptions of the marked fluctuations both have with suicide) would suggest that for persons who were at risk of dying an accidental death throughout the century, all that may have changed was the form of the accidental death. with a relatively increased number of motor-vehicle deaths occurring as motor-vehicles became more prominent. It should be noted that a number of intrapsychic explanations underlie the hypothesis that conscious and unconscious self-destructive tendencies may account for the correlation between suicide, homicide and accidents on a population level. For example. on an intrapsychic level of abstraction. grandiosity, needs for admiration and appreciation, masochism. inwardly-directly aggression, and rebelliousness and rule breaking may all result in any form of violent death and be the intrapsychic etiologic factor leading to the deaths and the parellelism seen on the population level. This hypothesis re: self-destructiveness seems to be supported by Brenner’s [?I. 373 work on the relationship between national mortality rates and economic cycles. Brenner [?I. 373 demonstrated that indicators of economic instability and insecurity. such as unemployment. were associated over time with higher mortalit! rates. His explanation for this association was that the Iach of economic security is stressful: social and farnil! structures break down and habits that are harmful to health are adopted. The effect may manifest acuteI\ as a psychopathological event (e.g. suicide and homicide). or. after a time lag of a few years. as a chronic disease (e.g. cancer or heart disease). Brenner’s model suggest ;I reason why suicide. homicide and accident rates ;tre parallel over time: economic qcles. In addition. his explanation of why the econom! is related to mortalit! rates (i.e. immoderate and unstable life habits) seems consistent with the idea that to some extent self-destructive tendencies may underlie all forms of violent deaths. Portertield [20] found a positive correlation of suicide-homicide. other crimes and accident death rates and he tended toward an explanation of aggression in motor-vehicle accidents and homicide data rather than underlying self-destructiveness: “There is no reason to doubt that aggressive. hazardous driving is likeI\ ti, be characteristic of persons similar to those ixho have suicidal or homicidal or both tenden-
1937
cies-and aice-wrsa”. However, in an interesting footnote. Porterfield addressed the issue of self-destructiveness in violent deaths quite clearly: “What would lead the investigator to begin with the prediction that ‘the same conditions that generate the suicide prone and the homicide prone’ would ‘also generate the accident prone’. Such a rationale would have to make room for a milieu which produces a posiriue correlation between different types of behavior and, at the same time, the ‘different sorts of persons’ who ‘react respectively in one way or the other’ to yield these positively correlated ‘different’ types of responses”. The data presented in this paper seem to support this kind of rationale: there are positive correlations between suicide. homicide and accidents but the specific type of violent death involved and the degree of risk is based on the age, race and sex of the individual. For example, suicide and homicide rates are positively correlated over time, but older white males are at greatest risk of dying by suicide and younger nonwhite males are at greatest risk of dying by homicide. Further research is necessary on both population as well as intrapsychic levels in order to enhance the understanding of the phenomenology of and motivations underlying violent deaths. REFERENCES
I. Gedo J. and Goldberg A. Models oj’thr Mired. University of Chicago Press. Chicago. IL. 1973. 2. Weiss N. S. Recent trends in violent deaths among young adults in the United States. Am J. Epidm. 103. 416422. 1976. 3. Webster’s New Colleyiate Dicrionary. p, 1306. G. & C. Merriam, Springfield. MA, 1973. 4. Menninger K. A. Mart Ayairtst Hintself Harcourt & Brace, New York. 1938. 5. Durkheim E. Suicide: A Study in Socioloy) (translated by Spaulding and Simpson). The Free Press, New York. 1951. Warren M. On suicide. J. Am. Psychout~alyt. Ass. 24, 199~234. 1976. Dublin L. Suicide: A Sociological and Statisricul Study. Ronald Press. New York. 1963. Kramer M.. Pollack E. S.. Redick R. W. and Locke B. Z. Mental DisorderslSuicide. Harvard University Press, Cambridge. MA. 1972. 9. Wolfgang M. E. Parterm irl Criminal Homicide. University of Pennsylvania Press. Philadelphia. PA. 1968. IO. Klebba A. J. Homicide trends in the United States. 1900-1974. P&l. H/r/i Rep. 90. 195-204. 1975. 11 Constantine J. P.. Kuller L. H.. Perper J. A. and Cypess R. H. An epidemiologic study of homicides in Allegheny County. Pennsylvania. Arm J. Epiderw 106. 314-324. 1977. I’ Freud S. T/u, P.s!~c~hopcrr/to/o~~~. of Ewr~da!~ L(fi. The Hogarth Press. London. 1960 [1901]. N. Accidenr or Suicide? Desrrwrion h.~, 13. Tabachnick Ammohik. Charles C. Thomas. Springfield. IL. 1973. I4 Schmidt C. W.. Sharer J. W.. Zoltowitz H. I. cr ~1. Suicide b! vehicular crash. 4r~. J. Psyhitrr. 134. 175-178.
1977.
I5 Seizer M. L. and Paine suicide. and unconscious 119. 237-240.
C. E. Automobile accidents. motivation. .4m J. P.t~~hrrl.
1962.
16 Kaplan H. B. and Pokorny A. D. Self-derogation and suicide--II. Suicidal responses. self-derogation and accidents. So<,.Sci. :\letl. 10. 119~231. 1976. 17 Lester D. Temporal variation in suicide and homicide. 4~1. J. Epitlrrrt. 109. 517-520.
1979.
1938
PAUL C. HOLINGER and ELAISE H. KLEWES
18. ~ogot E.. Fabsitz R. and Feinleib M. Daily variation in USA mortality. A/II. J. Epirlent. 103. 198-21 I. 1976. 19. Henrv A. F. and Short J. F. Suicide trt7tl Horuicitlr. The Free Press. Glencoe. IL. 1954. 20. Porterfield A. L. Traffic fatalities. suicide. and homicide .4rn. Socirrl. Rev. 25. 897-901. 1960. and the national economy. 21. Brenner M. H. Mortality Lmcet 568-573. 1979. of trends for suicide and 22. Klebba A. J. Comparison homicide in the United States. 1900-1976. In Violertce md rhe Violmr Idicidrrul (Edited by Hays J. R.). SP Medical & Scientific Books. New York. 1979. N. The Many Faces of Suicide. McGrdw23. Farberow Hill. New York, 1979. M. and Woods H. M. The incidence of 24. Greenwood industrial accidents upon individuals with special reference to multiple accidents. In Accidenf Research (Edited by Hadden. Suchman and Klein). Harper & Row, New York, 1964. 25. Van Buren G. H. Some things you can’t prove by mortality statistics. Vital Statistics-Special Reports 12, No. 13, 1940. 26. Dunn H. L. Shackley W. Comparison of cause-ofdeath assignments by the 1929 and .I938 revisions of the International List; Deaths in the United States. 1940. Vita/ Statistics-Specid Reports 12, No. 14. 1944. ratios 27. Faust M. M. and Dolman A. B. Comparability based on mortality statistics for the fifth and sixth revisions: United States, 1950. Vitul Staristics-Specinl Reports 51, No. 3, 1963. of mor28. Faust M. M. and Dolman A. B. Comparability tality statistics for the fifth and sixth revisions: United
States. 1950. l’ircrl Starisrics--Sprc,itrl Rrporrs 51. No. 2. 1963. 29. Faust M. M. and Dolman A. B. Comparability of mortality statistics for the sixth and seventh revisions: United States. 1958. b’irtrl Srtrri.sric,.s-Sprcic// Repom 51. No. 4. 1965. of terms and comparability of titles 30. Classification through five revisions of the International List of Causes of Death. Virtrl Sr~rrl.sric,s-SprcirII Reports 19. No. 13. 1944. of terms and comparability of titles 31. Classification through five revisions of the Internation List of Causes of Death. Vital Starisrics-Special Reporrs 19. No. 13. 1944.
32. Klebba A. J. and Dolman A. B. Comparability of mortality statistics for the seventh and eighth revisions of the international Classification of Diseases. United States. Vitul trr7d Hetrlrh Sttrristic.s. Series 2. Number 66. U.S. Government Printing Office. Washington. DC. 1975. 33. Holinger P. C. Violent deaths among the young: Recent trends in suicide. homicide. and accidents. A/n. _I. Psychiat. 136, 1144-l 147. 1979. 34. Death rates by age. race, and sex: United States. 1900-1953. Vif~tl Srurisrics-Speck/ Reporrs. 43. U.S. Government Printing Office. Washington. DC. 1956. 35. Holinger P. C. Violent deaths as a leadine . cause of mortality: An epidemiologic study of suicide. homicide and accidents. Aat. J. Ps!~chiut. 137. 472-476. 1980. 36. Freud S. Morrrrti!u/ trr7cl .Mehcholitr. The Hogarth Press. London. 1957 [l915]. 37. Brenner M. H. Economic changes and heart disease mortality. A/II. J. pub/. H/r/7 61. 606-61 I. 1971.
VIOLENT
0277-9536/X2;‘?21929-10603.00/O Pergamon Press Ltd
DEATHS
RELATIONSHIPS
IN THE 1900-1975
UNITED
STATES,
BETWEEN SUICIDE, HOMICIDE ACCIDENTAL DEATHS
PAUL C. HOLINGER’*
and
AND
ELAINE H. KLEMEN~
‘Institute for Psychosomatic and Psychiatric Research and Training, Michael Reese Hospital and Medical Center. Chicago. IL 60616 and ‘Community Mental Health Center, Northwestern Memorial Hospital. Chicago. IL. U.S.A. Abstract-Time trends are presented for suicide. homicide and accident mortality rates in the United States. 1900-1975. These data suggest that national mortality rates for suicide, homicide and motorvehicle accidents tend to be parallel over time. Non-motor-vehicle accidents, while showing some Ructuations similar lo those of other forms of violent deaths. manifest a more general decrease throughout the century. In addition. suicide rates tend to be significantly correlated over time with homicide. motorvehicle and non-motor-vehicle accident death rates for most race and sex combinations. These results need to be viewed with caution due to the methodologic problems inherent in using national mortality data. However. the findings raise serious questions about the traditional understanding of the relationships between the various forms of violent death in populations; in particular, the popular view that suicide and homicide are inversely related in populations is questioned. The explanatory usefulness of conceptualizing violent deaths as reflecting self-destructive tendencies is discussed.
INTRODUCTION
Controversy exists regarding the nature of the relationships between suicide, homicide and accidental deaths on both individual and population bases. Furthermore, it is not firmly established whether or not homicide and accidental deaths. along with suicide, reflect self-destructive tendencies. Many studies have addressed these issues from varying perspectives and using several different kinds of data. e.g. psychoanalytic, sociologic and epidemiologic. The present study utilizes population data (national mortality data from the United States. 1900-1975) to examine these questions. The purpose of this study is two-fold. First, the time patterns of suicide, homicide, motorvehicle accidents and non-motor-vehicle accidents will be compared in order to better understand the relationships of these forms of death in large populations. Second. these data will be used to address the question as to whether or not homicides and accidental deaths. along with suicide, may reflect self-destrucwork was. in part. done at and supported by the Harvard University School of Public Health, Boston, Massachusetts and, was also supported by the Judith Baskin Offer Research Fund. The authors gratefully acknowledge the assistance of Leo Levy (Univeriity of Illinois School of Public Health. Chicago. IL). Jane Murphy (Harvard Medical School. Boston. MA) and Daniel Offer (Michael Reese Hospitai and Medical Center. Chicago. IL) in the preparation of this report. The authors also gratefully acknowledge the assistance of Eric Ostrov (Michael Reese Hospital and Medical Center. Chicago. IL) in the statistical analysis. *Send reorint requests to: Dr Paul C. Holinger Psychosomatic-and Psychiatric Institute. Michael Reese Hospital and Medical Center. 2959 South Cottage Grove. Chicago. IL 60616. U.S.A.
This
1929
tive tendencies on a population level: the strategy employed is to compare the epidemiologic patterns of overt self-destructiveness (i.e. suicide) with homicides and accidental deaths. Theoretically, this work necessitates some bridging of intrapsychic disciplines involving individuals. e.g. clinical psychiatry and psychoanalysis (from which the hypotheses are derived), and fields which deal with populations, e.g. sociology and epidemiology (in which the hypotheses are tested). Such an attempt must be made with great caution, however. due to the serious conceptual problems involved in trying to understand the data and findings of one scientific discipline by using models derived from data of a very different scientific discipline. Gedo and Goldberg [I] have cogently discussed this potential for error in applying metapsychological propositions about intrapsychic phenomena to social psychology and vice versa. Why have suicides, homicides, and accidental deaths been studied together in the scientific literature’! The term ‘violent deaths’ has been used to include this combination of suicide. homicide and accidents (e.g. Weiss [2]), and Wehsrer’s [S] defines the term ‘a violent death’ to mean “caused by force: not natural”. Thus. the first reason probably related to the sense that these causes of mortality are often considered to be ‘unnatural’ as compared with. e.g. cancer or heart disease: violent deaths do not seem to involve the deterioration of the body’s internal organs. but rather seem somehow imposed from without. Second. clinical evidence suggests that homicides and accidents. as well as suicides. may be Selfinflicted: the victim to some degree provokes his or her own death by ‘being in the wrong place at the wrong time’. The-idea of risk-taking is relevant here. mith people who take more risks being in greater danger of dying by accident or homicide (homicide
P.ALL C. HOLISGER and ELAISE H. KLESIES
19Y.l
referring to those who are killed. not the killers). Suicide, homicide and accidents. in this framewcrk. are seen as related in that all ma) represent some expression of self-inflicted mortality. with suicide being most overt and homicide and accidents being more subtle manifestations of self-destructive tendencies and risk-taking. This paper is focused in two specific ways. First, the ph,=nOmenolOgiC rather than motivational aspects Of self-destructive behaviors will be examined: the intrapsychic models regarding the motivations underlying such self-destructiveness (e.g. masochism, narcissism and grandiosity, etc.) will be examined in future communications. Second, this paper is focused on self-destructive rather than aggressive behaviors, i.e. it is victim-oriented rather than perpetratororiented; however, this decision to study human behavior from a perspective of self-destructiveness in no way implies any intention to deny the importance of aggression nor imply a lessening of the responsibility of the perpetrator of the crime or accident. Clinically, the victim-oriented approach is optimistic in that it implies that people can, by recognizing and controlling unconscious self-destructive tendencies (whatever the origins), take greater control over their lives and more actively avoid misfortunes which have previously been perceived as unavoidable externallyoccurring events. LITERATL RE
There is. of course. a massive literature dealing with suicide. homicide. and accidents from both intrapsychic and population perspectives. This literature review will be limited to those studies which address the two issues under consideration: (1) the relationships between suicide, homicide and accidents; and (2) the question as to whether or not suicides, homicides and accidents may represent self-destructive tendencies. There are many excellent studies which separately discuss suicide [4-81, homicide [9-l I], or accidents [ 12-161 from either intrapsychic or sociologic perspectives or both. as well as many epidemiologic studies which deal primarily with the frequency and distribution of violent deaths [Z, I I, 17, 18-j. However, few studies explicitly examined the relationships among these causes of mortality on a population level. Perhaps the best known work which investigated these relationships was Henry and Short’s 1954 Slricidr curd Homicide [19]. Utilizing data which included suicide, homicides and aggravated assaults for specific localities over time spans of a relatively few years. the authors found that homicide and aggravated assault correlated positively with the business cycle, in contrast to suicide. They stated that “while suicide increases during (economic) depression, homicide decreases. While suicide drops during prosperity homicide increases” [ 191. Porterfield’s [ZO] lessknown but important work involved studies of metropolitan areas and found that their combined suicidehomicide rates correlated positively with their traffic fatality rates. i.e. he demonstrated that cities with a high suicide-homicide rate tended to have high traffic fatality rates. Porterfield suggested that persons involved in suicide. homicide, and traffic fatalities
were similar in having “little regard for the& own lives or the lives of others. or both”. Although Porterfield did not separate the suicide. homicide and motorvehicle accident rates. his assumptions. type of data and conclusions clearly differed from those of Henry and Short. Brenner’s [21] conclusions also seem to differ from Henry and Short’s findings. Brenner showed that indicators of economic instability and insecurity, such as unemployment and unusually rapid economic growth rates. were associated over time with higher mortality rates. Specifically. data for the entire United States showed that indices of acute pathological disturbances, including suicide and homicide. rose within a year of increased unemployment rates. Klebba [22] described trends for both suicide and homicide in the United States and her data also seemed to support a positive rather than inverse relationship between suicide and homicide over time. There is a large body of literature which addresses the question as to whether accidents and homicides. in addition to suicide, represent self-destructive tendencies. Regarding accidents. Freud [ 121 emphasized the role of psychic determinism in human behavior and, using clinical data. attempted to demonstrate that accidents were unconsciously precipitated by the victim. Menninger [4] and Farberow’s [23] work has been particularly convincing in clinically documenting the varied and ubiquitous nature of self-destructive tendencies underlying accidents and apparently externally-precipitated misfortunes. Several studies using small group data have been conducted to address the issue of accident-proneness. i.e. that certain people are more liable to have accidents. Greenwood and Woods [24] wrote an historically important paper in 1919. reporting that in a British munitions factory a relatively small number of workers had most of the accidents. The authors cautiously suggested that some people were inherently more likely to have accidents. Tabachnick [ 131 has carefully summarized the studies exploring this hypothesis and the controversies surrounding it. Some data suggest that deaths due to accidents. including motor-vehicle. may be due to accident-proneness and risk-taking that reflect depression and suicidal tendencies [15. 161; other investigations do not tend to support the hypothesis that suicidal tendencies were predominant among those involved in accidents [ 13, 14,251. To summarize. both clinical data and studies of small groups seem to suggest that the concepts of accident-proneness and excessive risk-taking have therapeutic and predictive value and that accidents and accidental deaths may be psychologically determined by underlying self-destructive tendencies: however. testing the hypothesis that accidents reflect unconscious self-destructive tendencies is extremely difficult and fraught with methodologic problems and several studies do not confirm the hypothesis. The most important work relating homicide to selfdestructive tendencies has been done by Wolfgang [9]. Studying the records of several hundred persons murdered in Philadelphia, Wolfgang found that more than 25”d of those people could be said to have provoked their own deaths. He spoke of this group as having committed “suicide by means of victim-precipitated homicide”. Wolfgang’s study supported clini-
Violent deaths in the United States, 1900-1975 cal notions that some patients could bring trouble upon themselves by not taking care of themselves and by being “in the wrong place at the wrong time”. METHODOLOGIC
CONSIDERATIONS
Serious methodologic problems exist in using national mortality data to study suicide, homicide, and accident trends in populations. These difficulties include underreporting, overreporting, misclassification and changes in classification over time. These issues have been discussed in detail elsewhere [26-341 and the changes in federal classification over time have been outlined by government reports. There has been little change over the century in federal classification for suicide. homicide and motor-vehicle accidents; accidents excluding motor-vehicle accidents have shown a slight tendency toward a decrease in the number of deaths listed as non-motor-vehicle accidents over the sixth (1949-1957). seventh (1958-1967) and eighth (1968 to date) Revisions of the International Classification of Diseases*. The data Rresented here include deaths that occurred in all U.S. states which participated in federal registration of deaths (death registration states), 19OGl932 and deaths which occurred in all U.S. states 1933-1975 (with Alaska added in 1959 and Hawaii in 1960). Data in this paper will be presented in two time periods. 1900-1975 and 1933-1975, inasmuch as (1) it was only in 1933 that data for the U.S.A. included the entire population of the U.S.A.; and (2) motor-vehicle accident deaths slowly increased in prominence from the early 1900s as the use of automobiles became widespread and only in the late 1920s and early 1930s did motor-vehicle accident death rates attain the general levels maintained until the present. RESULTS
Figure 1 presents the age-adjusted mortality rates for suicide. homicide. motor-vehicle accidents, nonand total accidents, accidents motor-vehicle 190&1975, United States. Of particular interest in
*Comparability ratios for the sixth. seventh and eighth Revisions were. respectively. for suicide 1.00. 1.03 and 0.9472: for homicide 1.00. 1.00 and 0.9969; for motorvehicle accidents 1.00. 1.00 and 0.9921; for accidents excluding motor-vehicle accidents 0.93.0.95 and 0.9250; and for accidents (total) 0.95. 0.97 and 0.9570. Comparability ratios were not calculated by the government prior to the sixth Revision because of the nature of the available data and the difficulty quantifying the classification changes. The comparability ratios and discussions of these changes can be found in References [26-341. Comparability ratios are calculations of the extent to which mortality rates for a cause of death may differ from one Revision to the next solely on the basis of reclassification of deaths. A comparability ratio of 1.00 indicates. usually. that the number of deaths attributed to a cause of mortality in the later Revision is essentially the same as for the earlier Revision: a ratio of less than 1.00 indicates that fewer dea’ths were attributed to the later Revision. while a ratio of greater than 1.00 indicates that more deaths were attributed to the later Revision due to classification changes.
1931
examining the relationship between the various forms of violent deaths are: the increases in suicide, homitide, motor-vehicle accidents and non-motor-vehicle accidents during the economic depression in the earlv 1930s; the decreases in suicide, homicide and motorvehicle accidents during World War II in the early 1940s; the slight increases in suicide, homicide and motor-vehicle accidents in the late 1940s and early 1950s; and the increases in suicide, homicide and motor-vehicle accidents throughout the 1960s. Nonmotor-vehicle accidents have tended to decrease throughout the century (with some of this decrease over the past 30 years being artifact due to classification changes-see Methodology); they show the increase during the first decade of the 1900s and the decreases during World War I which characterize suicide, as well as the increase during the early 1930s that characterizes all forms of violent deaths. but they increase in the early 1940s unlike the other causes of mortality. The relationship between suicide and homicide can most clearly be seen in Fig. 2, which presents their age-adjusted time trends 1900-1975 in greater detail. Before 1933, the major difference is that the suicide rate is biphasic (peaking around 1910 and 1930 and decreases during World War I which characterize suicide, as well as the increase during the early 1930s 1933, the rates are rather parallel: peaks in the early 1930s decreases in the early 1940s. slight increases in the late 1940s and early 1950s and larger increases in the 1960s and 1970s. Figures 3-6 present the age-adjusted violent death trends by race and sex for 1914-1975 (data for both race and sex are only available from 19 14 on). The rather parallel nature of the fluctuations in mortality rates is especially noticeable for suicide, homicide and motor-vehicle accidents. Correlations are also presented for the age-adjusted mortality time trends and they are separated into the years 1900 (or 1914 if race and sex are used) to 1975 (Table I) and 1933-1975 (Table 2). Of particular concern in addressing the second purpose of this paper (re: the self-destructive nature of violent deaths) are the correlation coefficients of suicide with the other forms of violent death. For example, for 1900-1975. suicide, when correlated with homicide, motor-vehicle accidents, non-motor-vehicle accidents and total accidents, showed coefficients of +0.22 (P < 0.05). -0.46 (P < 0.001). +0.63 (P < 0.001) and 0.71 (P < 0.001) respectively. For 1933-1975, suicide correlated +0.58 (P < 0.001). +0.53 (P < 0.001). +0.54 (P < 0.001) and +0.66 (P < 0.001) with homicide. motor-vehicle accidents, non-motor-vehicle accidents and total accidents respectively. The inverse correlation between suicide and motor-vehicle accidents from 1900-1975 appears to result from the continuous increase in motor-vehicle deaths from 1900 to 1975 when cars were just beginning to appear in’the population and assume prominence as a cause of mortality. Once the use of automobiles had become widespread. the time trends showed a significantly positive correlation between suicide and motor-vehicle accidents. 1933-1975. When the data are broken down into race and sex differences (19141975 and 1933-1975). the correlation patterns between suicide and the other forms of violent death are maintained with the follow-
P.ACL
1932
C.
and
HOLISGER violent
EL.AISE
H.
KLEMES
deaths Sunde Homude Motor-vahlcle Non-motor Totol
accidents vsn~cle occldents
Year
Fig. I. Age-adjusted
violent death rates by type of mortality. United States, 1900-1975. Sources of data: Reports 43, (for 1900-1953); Grove R. D. and Hetzel A. M. Vitul Sratisrics Rates in the United Stutes: 1940-1960. U.S. Government Printing Office, 1968 (for 19541960): Vird Srutistics in rhr United Stutes. Morraliry, 1961-1975 (for age-adjusted motor-vehicle accident and non-motorvehicle accident mortality rates, 1961-1975); Vito/ urtd ffedth Sturisrics. Series 20. Number 16. U.S. Government Printing Office, 1974 [for age-adjusted accident (total), suicide, and homicide mortality rates. 1970-19751.
Viral Stabstics-Speciul
20
r
I
1900
1910
1
I
,920
1930
homicide
rates,
8
I
1940
1950
W
Sumde
t.
Homictde
I
1960
I
,
1970 1975
YWr Fig. 2. Age-adjusted
suicide
and
United Fig. I.
States,
1900-1975.
Sources
of data:
same
as
Violent
deaths
in the United
States.
190~1975
1933
White male
Suude Homwde Motor-vehicle accidents bg;motor-vehxle occldents
Fig. 3. Age-adjusted
violent
death
rates. white males, United Fig. 1.
States.
1914-1975.
Sources
of data:
same as
White female o-o s-e
n-n
Suicide Homicide Motor-vehicle accidents Non-motor- vehicle accidents Total
~--1
i\
e-e
Year
Fig. 4. Age-adjusted
violent
death
rates. white females. United as Fig. I.
States.
1914-1975.
Sources
of data:
same
PALL C. HOLI~GER
1934
and ELAISE H. KLESIEY
NonwhIte male
e-9 w
Sulclde Hornlade Motor-vehicle
n-¤Non-motor-vehicle .-. Total
Fig.
Age-adjured
violent
death
rates,
nonwhite males. United same as Fig. I.
accidents accidents
States.
of data
1914-1975.
Nonwhite female -
Fig. 6. Age-adjusted
violent
death
rates.
nonwhite females. United same as Fig. I.
Suade Homude Motor-vehtcle occldents Non-motor-vehicle accidents Total
States.
1914-1975.
Sources
of data:
Violent deaths in the United States. 19Wl975
1935
Table 1. Correlation coefficients for violent death rates (age-adjusted). United States, 1900-1975 Total
violent
H 0.228
S
S H MVA NMVA TA
Violent S
Violent S
S H MVA NMVA TA
190&l
MVA - 0.4?* 0.39***
deuths.
H 0.72***
S H MVA NMVA TA
deaths,
white female.
MVA 0.001 o.ost
deaths, nonwhite
Violent
1914-I
female,
TA 0.71*** 0.17t -0.55*** 0.93***
S
TA 0.55*** 0.16t -0.15t 0.94***
S
975
NMVA 0.49*** 0.12t -0.47***
deaths,
H 0.67***
Violent
white mole. 1914-1975
MVA -0.08t - 0.26*
deaths, nonwhite
H 0.28*
MVA 0.01 i 0.27’
NMVA 0.56*** 0.49*** -0.66***
male, 1914-I
NMVA -0.02t 0.071 -0.75***
TA 0.66*** 0.49*** -0.33’8 0.92***
975
TA -0.02-l 0.38*** -0.16t 0.76***
1914-1975
NMVA 0.04t - 0.06t -0.75***
MVA 0.05t 0.29*
H 0.33**
975
NMVA 0.63*** 0.38*** -O.Sl***
Abbreviations. S = suicide; H = homicide; TA = total accidents. *P < 0.05; **P < 0.01; ***P < 0.001; tNS. Sources of data: same as Fig. 1.
TA 0.10t 0.08t - 0.43*** 0.91***
MVA = motor-vehicle
accidents;
NMVA = non-motor-vehicle
accidents;
Table 2. Correlation coefficients for violent death rates (age-adjusted), United States, 1933-1975 Total S
S
violent
H 0.58***
H MVA NMVA ylmt
d;ths.
0.68***
S
H MVA NMVA TA Vi;/ent
S
H MVA NMVA TA
deaths,
H 0.53***
deaths.
f933-1975
MVA 0.53*** 0.43***
white female,
MVA 0.56*** 0.55***
nonwhite
NMVA 0.54*** -o&It 0.04t
Violent
TA 0.66*** 0.08t 0.32; 0.96***
S
TA 0.42** -0.16t 0.02t 0.96***
S
1933-1975
NMVA 0.27’ - 0.29* - 0.22t
H 0.58***
VioFt
0.58***
deaths,
white
MVA 0.52*** 0.32*
deaths, nonwhite
MVA 0.42** 0.39**
male, 1933-I
975
NMVA 0.49*** 0.05t 0.15t
TA 0.59*** 0.13t 0.42** 0.96***
male, 1933-I
975
NMVA -0.51*** 0.22t -o&It
TA -0.23t 0.40** 0.48*** 0.85***
female, 1933-1975
MVA 0.61*** 0.15t
NMVA -0.44+** o.OOt -0.42**
TA -0.18t 0.07t o.OOt 0.90***
Abbreviations: same as Table 1. *P < 0.05; **P < 0.01; ***P < 0.001: INS. Source of data: same as Fig. I.
ing major exceptions: for all groups there is no significant inverse correlation between suicide and motorvehicle accidents 1914-1975. probably because much of the initial increase in MVA mortality is excluded: and the nonwhite male and nonwhite female suicide rates. for 1914-1975. sho\v no significant correlation with non-motor-vehicle accidents and. 1933-1975. show a significant inverse correlation with nonmotor-vehicle accidents.
In examining the data by race and sex variables, the following relative comparisons should be noted: male rates for all forms of violent deaths are higher than females; for whites. non-motor-vehicle accidents show the highest mortality rates. followed by motor-vehicle accidents. suicide and homicide, respectively; however, for nonwhites, non-motor vehicle accidents are followed by homicide, motor-vehicle accidents and suicide rates respectively. Age patterns have been
POLL
1936
C. HOLISGER
and ELAN
presented elsewhere [35] and \vill be mentioned onl! briefly: suicide rates tend to increase bvith age for males ~thile female rates are highest between the ages of Jj-64 years: homicide rates are highest among young adults for males and among infants and !oung adults for females: the highest rates of motor-vehicle accident deaths for both sexes are found in teenagers and young adults and the oldest age groups: and the oldest and boungest (infancy) ages are at greatest risk of dying by non-motor-vehicle accidents for both males and females.
.
DISCUSSION
Time trends leave been presented for suicide, homicide and accident mortality rates in the United States. The focus has been on the years 1933-1975 inasmuch as the data include the entire United States population only during those years and at approximately the same time motor-vehicles reached levels they were to maintain as a prominent cause of mortality. These data suggest that mortality rates for suicide, homicide. motor-vehicle accidents and to a lesser extent nonmotor-vehicle accidents tend to be parallel over time. In particular. homicide. motor-vehicle and nonmotor-vehicle accident death rates tend to be significantly correlated with suicide rates over time for most race and sex combinations. At least two specific methodologic artifacts might cause the reported mortality rates for suicide. homicide and accidents to simultaneously increase or decrease. First, federal classification changes must be considered as a possible cause of parallel fluctuations. For example: a person shoots himself (or is shot). does not die immediately. and then dies 2 weeks later of pneumonia, which indirectly results from the shooting. A classification change which shifted the cause of such deaths from pneumonia to suicide (or homicide) would lead to simultaneous increases in the suicide and homicide rates. However, the comparability ratios indicate that no such simultaneous classification shifts have occurred large enough to account for the parallel fluctuations. A second possible artifact which might cause simultaneous fluctuations in the various forms of violent deaths involves increases or decreases in the number of reported deaths based on local or national budget, or political issues. For example. if budgeting were low for health officials. more deaths of ‘undetermined etiology’ might be reported with a general decrease in the specilic forms of violent deaths. Although in this instance the actual trend is opposite (i.e. in times of great economic depression the violent death rates have increased). the example illustrates the potential kind of problem which could lead to simultaneous fluctuations of violent deaths. Due to methodologic problems inherent in using national data, these results must be viewed with caution. However. the data seem to raise serious questions about the traditional understanding of the relationship between the various forms of violent death in populations. particularly with respect to suicide and homicide. The classic view that suicide and homicide rates are inversely related in populations (e.g. Henry and Short [ 193) is not supported by these data:
H. KLFME\
rather. the data seem to support the opposite trndenq. that suicide and homicide rates fluctuate in parallel over time. Brenner‘s [?I] and Klehb:~‘s [X] \\ork also support the parallel Huctuation 01 ,uicidc and homicide rates. One must ash ho\\ the traditianal view that there is an inverse relationship bct\\ccn \LIItide and homicide came to ha\ e such \\ ide acceptance in light of the contradictor! e\ idsnce presented in this paper and elsewhere [?I. 221. First. and perhap mo\t importantly. the data used ti! Hcnr! and Short [ 191 were very ‘different than those utilized in the present article and by Brenner [?I] and Klebba [?I. Hem-b and Short’s [ 191 data were suicide and homicide rates for specific locations (e.g. cities) and for relatively short time spans. crimes against propert! and people. social status. strength of the relational s!strm and measures of the business cycle. The data presented in this paper as well as in those of Brenner [21] and Klebba [Z?Z] differ in two important ways. in that the! are national mortality data (i.e. not a sample but an entire population) and they focus on longer time trends. Such differences in data could account for the contradictory findings. Second, Henry and Short [ 191 conceptualized the suicide-homicide problem from the perspective of aggression: the subtitle of their book is “Some Economic. Sociological and Psychological Aspects of Aggression”. Freud’s intrapsychic understanding of aggression may have influenced Henry and Short’s hypothesis that suicide and homicide were inversely related. Freud [36] suggested that. as a result of frustration, aggression could be directed either externally (possibly resulting in homicide) or turned back on the self (possibly resulting in suicide). Consistent with this. Henry and Short state that suicide and homicide “are alike in the sense that they are both aggressive reactions to frustrations.. [but] become sharply differentiated aggressive acts when we consider the object of the aggression rather than its source”. However. Henry and Short may have a misapplied Freud’s intrapsychic model as they made the shift to population levels. The transition from individuals (from which the hypothesis derives) to populations is complicated by the often-overlooked to those fact that homicide L 1 refer specifically killed (the objects of the aggression). not to the killers (those who when frustrated direct their aggression outward). In other words, homicide data say nothing. strictly speaking, about the aggressors and the nature or amount of aggression. It seems unwarranted to assume anything about the aggression in a population or characteristics of killers based on data which describe only characteristics of people who are killed. Data on suicide and homicide in populations might best be conceptualized as similar in that both measure the object of the aggression. not the source. Having addressed the first purpose of this stud! (the comparison of the time trends and relationships between the various forms of violent deaths in the population) the second question must be dealt with: whether or not homicides :tnd accidents. in addition to suicide. can be said to retlect self-destructive tendencies. The initial strategy was to compare suicide rates (overt self-destructiveness) with homicide and accident rates. Examination of the hypothesis that violent deaths reflect self-destructive tendencies is made especially ditficult due to the different levels of
Violent deaths in the United States, 1900-1975 abstraction involved. The findings that suicide, homicide and motor-vehicle accident rates tend to be parallel over time and that homicides, motor-vehicle accidents and non-motor-vehicle accidents correlate overt self-destructiveness-i.e. significantly with suicide--can perhaps at best be said to suggest that self-destructiveness is an important variable which is associated with various forms of violent deaths on a population level. Within this hypothesis. the differences between non-motor-vehicle accidents and the other forms of violent death would be explained as follows: (1) non-motor-vehicle accidents have enough major fluctuations to correlate positively with suicide (the increase around 1910: the decreases in World War I: the increases in the early 1930s; and the recent ‘leveling’ off. which is in reality an increase which manifests as a ‘leveling off because of the comparability ratio): and (2) the tendency of non-motor-vehicle accidents and motor-vehicle accidents toward an inverse relationship (with the exceptions of the marked fluctuations both have with suicide) would suggest that for persons who were at risk of dying an accidental death throughout the century, all that may have changed was the form of the accidental death. with a relatively increased number of motor-vehicle deaths occurring as motor-vehicles became more prominent. It should be noted that a number of intrapsychic explanations underlie the hypothesis that conscious and unconscious self-destructive tendencies may account for the correlation between suicide, homicide and accidents on a population level. For example. on an intrapsychic level of abstraction. grandiosity, needs for admiration and appreciation, masochism. inwardly-directly aggression, and rebelliousness and rule breaking may all result in any form of violent death and be the intrapsychic etiologic factor leading to the deaths and the parellelism seen on the population level. This hypothesis re: self-destructiveness seems to be supported by Brenner’s [?I. 373 work on the relationship between national mortality rates and economic cycles. Brenner [?I. 373 demonstrated that indicators of economic instability and insecurity. such as unemployment. were associated over time with higher mortalit! rates. His explanation for this association was that the Iach of economic security is stressful: social and farnil! structures break down and habits that are harmful to health are adopted. The effect may manifest acuteI\ as a psychopathological event (e.g. suicide and homicide). or. after a time lag of a few years. as a chronic disease (e.g. cancer or heart disease). Brenner’s model suggest ;I reason why suicide. homicide and accident rates ;tre parallel over time: economic qcles. In addition. his explanation of why the econom! is related to mortalit! rates (i.e. immoderate and unstable life habits) seems consistent with the idea that to some extent self-destructive tendencies may underlie all forms of violent deaths. Portertield [20] found a positive correlation of suicide-homicide. other crimes and accident death rates and he tended toward an explanation of aggression in motor-vehicle accidents and homicide data rather than underlying self-destructiveness: “There is no reason to doubt that aggressive. hazardous driving is likeI\ ti, be characteristic of persons similar to those ixho have suicidal or homicidal or both tenden-
1937
cies-and aice-wrsa”. However, in an interesting footnote. Porterfield addressed the issue of self-destructiveness in violent deaths quite clearly: “What would lead the investigator to begin with the prediction that ‘the same conditions that generate the suicide prone and the homicide prone’ would ‘also generate the accident prone’. Such a rationale would have to make room for a milieu which produces a posiriue correlation between different types of behavior and, at the same time, the ‘different sorts of persons’ who ‘react respectively in one way or the other’ to yield these positively correlated ‘different’ types of responses”. The data presented in this paper seem to support this kind of rationale: there are positive correlations between suicide. homicide and accidents but the specific type of violent death involved and the degree of risk is based on the age, race and sex of the individual. For example, suicide and homicide rates are positively correlated over time, but older white males are at greatest risk of dying by suicide and younger nonwhite males are at greatest risk of dying by homicide. Further research is necessary on both population as well as intrapsychic levels in order to enhance the understanding of the phenomenology of and motivations underlying violent deaths. REFERENCES
I. Gedo J. and Goldberg A. Models oj’thr Mired. University of Chicago Press. Chicago. IL. 1973. 2. Weiss N. S. Recent trends in violent deaths among young adults in the United States. Am J. Epidm. 103. 416422. 1976. 3. Webster’s New Colleyiate Dicrionary. p, 1306. G. & C. Merriam, Springfield. MA, 1973. 4. Menninger K. A. Mart Ayairtst Hintself Harcourt & Brace, New York. 1938. 5. Durkheim E. Suicide: A Study in Socioloy) (translated by Spaulding and Simpson). The Free Press, New York. 1951. Warren M. On suicide. J. Am. Psychout~alyt. Ass. 24, 199~234. 1976. Dublin L. Suicide: A Sociological and Statisricul Study. Ronald Press. New York. 1963. Kramer M.. Pollack E. S.. Redick R. W. and Locke B. Z. Mental DisorderslSuicide. Harvard University Press, Cambridge. MA. 1972. 9. Wolfgang M. E. Parterm irl Criminal Homicide. University of Pennsylvania Press. Philadelphia. PA. 1968. IO. Klebba A. J. Homicide trends in the United States. 1900-1974. P&l. H/r/i Rep. 90. 195-204. 1975. 11 Constantine J. P.. Kuller L. H.. Perper J. A. and Cypess R. H. An epidemiologic study of homicides in Allegheny County. Pennsylvania. Arm J. Epiderw 106. 314-324. 1977. I’ Freud S. T/u, P.s!~c~hopcrr/to/o~~~. of Ewr~da!~ L(fi. The Hogarth Press. London. 1960 [1901]. N. Accidenr or Suicide? Desrrwrion h.~, 13. Tabachnick Ammohik. Charles C. Thomas. Springfield. IL. 1973. I4 Schmidt C. W.. Sharer J. W.. Zoltowitz H. I. cr ~1. Suicide b! vehicular crash. 4r~. J. Psyhitrr. 134. 175-178.
1977.
I5 Seizer M. L. and Paine suicide. and unconscious 119. 237-240.
C. E. Automobile accidents. motivation. .4m J. P.t~~hrrl.
1962.
16 Kaplan H. B. and Pokorny A. D. Self-derogation and suicide--II. Suicidal responses. self-derogation and accidents. So<,.Sci. :\letl. 10. 119~231. 1976. 17 Lester D. Temporal variation in suicide and homicide. 4~1. J. Epitlrrrt. 109. 517-520.
1979.
1938
PAUL C. HOLINGER and ELAISE H. KLEWES
18. ~ogot E.. Fabsitz R. and Feinleib M. Daily variation in USA mortality. A/II. J. Epirlent. 103. 198-21 I. 1976. 19. Henrv A. F. and Short J. F. Suicide trt7tl Horuicitlr. The Free Press. Glencoe. IL. 1954. 20. Porterfield A. L. Traffic fatalities. suicide. and homicide .4rn. Socirrl. Rev. 25. 897-901. 1960. and the national economy. 21. Brenner M. H. Mortality Lmcet 568-573. 1979. of trends for suicide and 22. Klebba A. J. Comparison homicide in the United States. 1900-1976. In Violertce md rhe Violmr Idicidrrul (Edited by Hays J. R.). SP Medical & Scientific Books. New York. 1979. N. The Many Faces of Suicide. McGrdw23. Farberow Hill. New York, 1979. M. and Woods H. M. The incidence of 24. Greenwood industrial accidents upon individuals with special reference to multiple accidents. In Accidenf Research (Edited by Hadden. Suchman and Klein). Harper & Row, New York, 1964. 25. Van Buren G. H. Some things you can’t prove by mortality statistics. Vital Statistics-Special Reports 12, No. 13, 1940. 26. Dunn H. L. Shackley W. Comparison of cause-ofdeath assignments by the 1929 and .I938 revisions of the International List; Deaths in the United States. 1940. Vita/ Statistics-Specid Reports 12, No. 14. 1944. ratios 27. Faust M. M. and Dolman A. B. Comparability based on mortality statistics for the fifth and sixth revisions: United States, 1950. Vitul Staristics-Specinl Reports 51, No. 3, 1963. of mor28. Faust M. M. and Dolman A. B. Comparability tality statistics for the fifth and sixth revisions: United
States. 1950. l’ircrl Starisrics--Sprc,itrl Rrporrs 51. No. 2. 1963. 29. Faust M. M. and Dolman A. B. Comparability of mortality statistics for the sixth and seventh revisions: United States. 1958. b’irtrl Srtrri.sric,.s-Sprcic// Repom 51. No. 4. 1965. of terms and comparability of titles 30. Classification through five revisions of the International List of Causes of Death. Virtrl Sr~rrl.sric,s-SprcirII Reports 19. No. 13. 1944. of terms and comparability of titles 31. Classification through five revisions of the Internation List of Causes of Death. Vital Starisrics-Special Reporrs 19. No. 13. 1944.
32. Klebba A. J. and Dolman A. B. Comparability of mortality statistics for the seventh and eighth revisions of the international Classification of Diseases. United States. Vitul trr7d Hetrlrh Sttrristic.s. Series 2. Number 66. U.S. Government Printing Office. Washington. DC. 1975. 33. Holinger P. C. Violent deaths among the young: Recent trends in suicide. homicide. and accidents. A/n. _I. Psychiat. 136, 1144-l 147. 1979. 34. Death rates by age. race, and sex: United States. 1900-1953. Vif~tl Srurisrics-Speck/ Reporrs. 43. U.S. Government Printing Office. Washington. DC. 1956. 35. Holinger P. C. Violent deaths as a leadine . cause of mortality: An epidemiologic study of suicide. homicide and accidents. Aat. J. Ps!~chiut. 137. 472-476. 1980. 36. Freud S. Morrrrti!u/ trr7cl .Mehcholitr. The Hogarth Press. London. 1957 [l915]. 37. Brenner M. H. Economic changes and heart disease mortality. A/II. J. pub/. H/r/7 61. 606-61 I. 1971.