1 Occupation and Suicide* Steven Stack, Wayne State University Objective. Research on occupation and suicide has neglected multivariate models. It is not clear, for example, if persons in alleged high-risk occupations have high suicide risk because of occupational stress associated with the occupation or because of the demographic composition of the people in the occupation. The present study explores the relationship between occupation and suicide for 32 occupational groups. Methods. Data are from the national mortality file tapes, which cover 21 states. They refer to 9,499 suicides and 134,386 deaths from all other causes in Results. Bivariate logistic regression models find a total of 15 occupations with either significantly higher (e.g., dentists, artists, machinists, auto mechanics, and carpenters) or lower (e.g., clerks, elementary school teachers, cooks) risk than the rest of the working-age population. Multivariate models that remove the demographic covariates of occupation find only eight occupations with greater or lower than expected risk of death by suicide. Conclusion. The results underscore the need for demographic controls in the assessment of occupational risk of suicide. They are consistent with a previous study based on data from England. The findings provide the first systematic evidence on the problem for the United States. Research on occupations and suicide is often marked by inconsistent findings. The patterns that are reported in one study may not be reported in another study (Bedeian, 1982; Lester, 1992:108 12; Boxer, Burnett, and Swanson, 1995; see review in Stack, 2000a). For example, 11 of 18 studies of police suicide report a high rate, three report an average rate, and four conclude that police have a low rate (Stack and Kelley, 1994). Research on physician suicide is also marked by similarly inconsistent findings (Bedeian, 1982; Stack, 1998, 2000a). Methodological problems often contribute to such discrepant findings. These problems include failure to control for the covariates of occupational status, such as demographic covariates; failure to compare a suicide rate with an appropriate reference group; use of different time periods; use of small numbers of suicides, which can bias rates either upward or downward; and *Direct all correspondence to Steven Stack, Department of Criminal Justice, Wayne State University, Detroit, MI Requests for the data used in this article should be directed to either the U.S. Public Health Service or the Inter-University Consortium for Political and Social Research (ICPSR), University of Michigan, Ann Arbor. The latter agency provided the data to the author under existing ICPSR contractual arrangements. An earlier version of this article was presented at the annual meetings of the American Sociological Association, August 11 15, 2000, Washington, D.C. SOCIAL SCIENCE QUARTERLY, Volume 82, Number 2, June by the Southwestern Social Science Association
2 Occupation and Suicide 385 difficulties in classifying many occupations into clearly defined categories (Boxer, Burnett, and Swanson, 1995; Wasserman, 1992; Stack, 1998; Stack and Kelley, 1994). The present study focuses on one shortcoming of the past work: the relative lack of multivariate analysis (Boxer, Burnett, and Swanson, 1995). In particular, the present investigation employs a logistic regression model of occupation and suicide with controls introduced for basic demographic correlates of occupation: gender, race, age, and marital status. In this fashion, the risk of suicide in an occupation can be calculated independent of the demographic covariates of the occupation. In addition, the present study employs national data from the U.S. Public Health Service in order to generate a large sample of suicides and more reliable estimates of occupational suicide risk. Previous Work on Occupation and Suicide The link between occupation and suicide is not well understood. Suicide rates are not available for most occupations. Most research on occupation and suicide has focused either on a single occupation or a small set of occupations (see Boxer, Burnett, and Swanson, 1995). Of the 76 investigations on occupation and suicide reviewed by Boxer, Burnett, and Swanson (1995), nearly all covered a single occupation or a single industry. Often suicide is just one of many causes of death reported on in the typical descriptive epidemiological survey of an industry (e.g., chemical workers). The tendency to study the suicide risk of a single occupation has continued (e.g., Stack and Kelley, 1994; Stack, 1996b). Lampert, Bourque, and Kraus (1984) dealt only with 11 general categories (e.g., professional technical workers). Milham (1983) gives information on 22 specific occupations (e.g., sheepherder, hairdresser) for the state of Washington. Lalli and Turner (1968) provide national data from 1950 but for only six general occupational categories for males ages Labovitz and Hagedorn (1971) present data on suicide rates for 36 selected specific occupations for males in Many specific occupations are found in only one of the articles. This makes it difficult to assess the reliability of the estimate. Finally, many of the studies do not address female suicide risk. Most of the work is restricted to men. Since a substantial proportion of the labor force is female, this may bias the estimates of suicide risk in some occupations. For a further description of the few published studies that survey suicide rates in three or more occupations, see Appendix 1. Even crude suicide rates remain undocumented for literally hundreds of specific occupations, especially some occupations that are relatively uncommon. The rationale for differences in suicide by occupation remains unexplored in any rigorous sense. Detailed explanations of suicide risk are available for only a few occupations, including physicians, dentists, and artists (Boxer, Burnett, and Swanson, 1995; Stack, 1996a, 1996b). The
3 386 Social Science Quarterly available research has typically not sought to weigh the relative importance of certain correlates of occupational status thought to contribute to suicide risk. These include demographic, stress, opportunity, and psychiatric morbidity correlates. Occupation and Suicide: A Model The relationship between occupation and suicide can be best conceptualized in terms of a multicausal heuristic model (Stack, 1996a; Wasserman, 1992). 1 This model comprises four major components or factors contributing to occupational suicide risk: demographics, internal occupational stress, preexisting psychiatric morbidity, and differential opportunities for suicide. These explanations are not mutually exclusive. For example, an occupation may attract persons with high psychiatric morbidity and also be populated by demographic groups at high risk (e.g., men, whites). Because of data limitations, the present study can test only the demographic dimensions of the full model. Given this limitation, the latter three aspects of the full model will receive only limited attention in this section. Demographics Occupations vary in their demographic characteristics, such as age, gender, and race. Suicide rates tend to increase with age for men and peak for women at around age 50. Whites have a suicide rate twice that of African Americans. Men have a suicide rate four times that of women (Stack, 1982, 2000a; Lester, 1992). Further, occupational groups vary in the extent to which their members are divorced, single, and widowed, all risk factors for suicide (Stack, 1982; Lester, 1992). Research in England has demonstrated the importance of controlling for the demographic covariates of occupation. Charlton (1995) analyzed 13,117 suicides and 252,833 natural deaths in England and Wales. For men, only five of the ten occupations with the highest suicide rates had significantly higher suicide rates after demographic controls were included in the analysis. For women, only three of the occupations (veterinarian, doctor, and nurse) had significantly higher suicide rates after demographic controls were incorporated. 1 Generally speaking, there is an inverse relationship between occupational prestige and the risk of suicide (e.g., Lampert, Bourque, and Kraus, 1984; Liu and Waterbor, 1994; for a review see Boxer, Burnett, and Swanson, 1995). For example, in California, Lampert, Bourque, and Kraus (1984) found a five-to-one ratio in the suicide rate of farm laborers (135.1 per 100,000) versus that of managerial professional workers (26.5 per 100,000). However, there is considerable variation within broad occupational groups. For example, at high prestige levels we have professors and electrical engineers with relatively low suicide rates, on the one hand, and physicians and dentists with relatively high suicide rates, on the other hand (Labovitz and Hagedorn, 1971; Stack, 1996a; Wasserman, 1992).
4 Occupation and Suicide 387 Some research on individual occupations in the United States has also demonstrated the need for demographic controls. The suicide rate of laborers is significantly greater than that of the general population. However, when controls for gender, marital status, and other covariates of laborer status are employed, the relative risk of suicide is the same as that for the working-age population. Laborers evidently have a high suicide rate mainly because they tend to be men and are more likely to be divorced and single than the rest of the working-age population (Stack, 1995). The same pattern held in a study of carpenters (Stack, 1999a). One of the most studied occupations in the suicide literature is policing. The time frame for such studies ranges from the 1930s through the 1990s (Stack and Kelley, 1994). Police are often considered to have a high suicide rate, with most studies reporting a rate higher than the general population (Stack and Kelley, 1994). The internal occupational stress factors behind these allegedly high rates include shift work, the public s antipolice sentiments, failure of the courts to punish offenders arrested by the police, and constant danger on the job. For example, a feature article on police suicide in The New York Times reported that the police were at high risk of suicide (James, 1993). However, the rate was 21 suicides per 100,000 officers. This is actually below the national rate for males of working age. The suicide rate for the general population is not an appropriate reference group, since most police are men, and men have a suicide rate four times that of women. In like manner, Stack and Kelley (1994) report a suicide rate of 25.6 per 100,000 for police, which is, indeed, twice the national average of 12 per 100,000. However, the suicide rate for age- and gender-matched controls (men 15 64) is 23.8 per 100,000. Police have a suicide rate that is only slightly greater than men of the same age. The difference is not statistically significant (Stack and Kelley, 1994). Burnett, Boxer, and Swanson (1992), using data from 26 states, also report only a slightly elevated suicide rate for police. Suicide risk in an occupation may vary over time as demographic groups with low suicide risk (e.g., women and minorities) enter or leave the occupation (Lampert, Bourque, and Kraus, 1984). As such, suicide rates will vary over time for an occupation with changes in its demographics. Reports based at one point in time may find different results from reports from decades ago. However, the association between occupation and suicide withstands controls for demographic factors. Dentists tend to be white males and have a suicide rate that is often several times or more greater than that of the general population. However, is this due to their being dentists or due to their being white males, or both? One study controlled the effects of race, gender, and other covariates of dentistry. It determined that dentists were 6.64 times more likely than the working-age population to die of suicide. In the case of dentists, even with the covariates of dentistry controlled, occupational stress evidently is associated with a very high risk of suicide (Stack, 1996a). The
5 388 Social Science Quarterly same is true for artists (Stack, 1996b): Artist status remained a significant predictor of suicide risk, even after demographic controls were incorporated into the analysis. However, the general extent to which demographic controls are associated with nonsignificant findings between an occupational status and suicide potential is unclear. Internal Occupational Stress Stress associated with the nature of an occupation may contribute to suicide risk. Features of work such as client dependence, status integration, and social isolation may increase suicide risk (Bedeian, 1982; Wasserman, 1992). Client dependence is the degree to which people in an occupation are dependent on clients for their livelihood. Occupations such as sole business operator and physician would be expected to have more suicide risk than machinists and mail carriers, who are not directly dependent on clients for the source of their livelihood. In a pilot study of 36 occupations, Labovitz and Hagedorn (1971) found a mean suicide rate of 40.5 per 100,000 for persons in client-dependent occupations versus a mean of 25.9 per 100,000 for persons in non-client-dependent occupations. Gibbs and Martin s (1964) theory of status integration provides a framework for predicting occupational stress. Persons in statistically infrequent role sets (e.g., female doctor) should have higher suicide rates than their counterparts (e.g., female schoolteacher). Such role sets are presumably stressful, since people have tended to avoid them. Although little systematic work has been done on occupational stress and suicide, a series of studies have found that persons in statistically infrequent occupation-based role sets (e.g., female chemist, female soldier) do indeed have higher than average risk of suicide (e.g., Bedeian, 1982; Seiden and Gleiser, 1990; Stack, 1995; for a partial exception see Alston, 1986). Psychiatric Morbidity Occupations may vary in the extent to which they recruit persons at psychiatric risk of suicide. That is, if an occupation attracts persons with personality traits that put them at risk of suicide, such occupations may be marked by high suicide rates. These suicidal characteristics are not necessarily due to occupational stress. They may exist before entry into the job (Bedeian, 1982; Wasserman, 1992). For example, it has often been argued, although never rigorously demonstrated, that the high suicide rate often reported for psychiatrists is due to preemployment psychiatric morbidity. That is, highly educated persons with depressive disorders may tend to select this occupation (Wasserman, 1992). The same argument has been made for other occupations, such as artist (Andreason, 1987; Stack, 1996b).
6 Occupation and Suicide 389 Opportunity Factors Occupations vary according to the opportunities available for access to lethal means of suicide. The availability of lethal drugs in the medical profession (physicians, pharmacists, dentists, nurses) has been linked to corresponding high suicide risk (Boxer, Burnett, and Swanson, 1995; Burnett, Boxer, and Swanson, 1992; Peipins, Burnett, and Alterman, 1997; Stack, 1996a; Wasserman, 1992). However, opportunity factors do not apparently explain the low to average risk of suicide among some occupations with high access to firearms (e.g., military and public safety workers, respectively) (e.g., Burnett, Boxer, and Swanson, 1992; Burnley, 1995; Helmkamp, 1996; Stack and Kelley, 1994). The current investigation cannot address three aspects of the model of suicide: occupational stress, preexisting psychiatric morbidity, and opportunity. Any association found between occupation and suicide, which is independent of demographic controls, could be due to one or more of these unmeasured conditions. This limitation is found in essentially all of the previous work on occupation and suicide. Methodology Data on suicide are from the 21 states that report occupational data in the 1990 national mortality detail file (U.S. Public Health Service, 1994): Colorado, Georgia, Idaho, Indiana, Kansas, Kentucky, Maine, Nevada, New Hampshire, New Jersey, New Mexico, North Carolina, Ohio, Oklahoma, Rhode Island, South Carolina, Utah, Vermont, Washington, West Virginia, and Wisconsin. The study is limited to the working-age population, ages The dependent variable is a binary variable where 1 = death from suicide and 0 = death from all other causes. The sample consisted of 9,499 suicides and 137,687 deaths from other causes. Some caution needs to be exercised in interpreting the results of the present investigation, since it is based on official suicide statistics. Some underreporting of suicide may exist due to such factors as the degree of professionalism among local coroners (e.g., Kushner, 1993). However, even with controls introduced for measures of the degree of professionalism in the certification of the cause of death, research has shown that the relationship between core sociological variables and suicide remains essentially the same (Pescosolido and Mendelsohn, 1986). The measurement errors in the official statistics are not large enough to preclude meaningful analysis. Since the dependent variable is a dichotomous variable, logistic regression techniques are appropriate (DeMaris, 1992). The logistic regression equations were estimated using the logistic regression utility programs in SAS (SAS, 1990). SAS uses the iteratively reweighted least squares algorithum to compute the maximum likelihood estimators of the regression parameters. The expected value of the hessian matrix from the last iteration is used to
7 390 Social Science Quarterly generate the estimated covariance matrix of the maximum likelihood estimates. Odds ratios are computed by taking the antilogarithm of the unstandardized logistic regression coefficients (SAS, 1990: ). Thirty-two bivariate logistic regressions are estimated, one for each occupational group. These are followed by 32 multivariate logistic regressions, which determine if the relationship remains after demographic controls are incorporated into the analysis. Occupational groups were defined using standard federal occupational codes (U.S. Public Health Service, 1994). Occupations were principally selected on the basis of having at least 50 suicides. In some instances, occupations were combined in order to have enough suicides. For example, managers and executives were combined in this fashion. Further research might introduce additional occupational groups combining specific occupations into larger categories. The present study presents data mostly on specific occupations that are well known. It includes representative occupations from six major occupational headings: (1) Managerial-Professional: executives/managers, doctors, accountants, engineers, nurses, and artists; (2) Clerical: bookkeepers, clerks, and postal workers; (3) Service: police, private security, and bartenders; (4) Agricultural and Extractive: farmers, farm workers, and miners; (5) Skilled Manual: machinists, auto mechanics, and electricians; (6) Semiskilled and Unskilled Manual: welders, laborers, and truck drivers. A total of 32 occupational groups are analyzed. These do not constitute the universe of all occupational groups. They do constitute a set from the occupations that are available. Controls are introduced for a series of sociodemographic variables that often covary with occupational status. Gender is coded as a binary variable with 1 = male, 0 = female. To the extent that an occupation is male dominated, it will have an elevated suicide rate. Men have a suicide rate four times greater than women (Stack, 1982, 2000a; Lester, 1992). Race is coded as white = 1, all others = 0. To the extent that an occupation attracts whites, it will have a potentially elevated suicide rate, since whites have higher suicide rates than blacks (Stack, 1982, 2000a; Lester, 1992). Marital status generally has a strong relationship to suicide risk (Stack, 2000b). Marital status is coded as a binary variable with 1 = married and 0 = all others. Persons in lower socioeconomic status occupations tend to have higher rates of divorce. As such, it is unclear if any elevated risk of suicide in a manual occupation is due to occupational stress or marital stress (e.g., Stack, 1995). Age is coded in years. Suicide rates generally increase with age. To the extent that an occupation is composed of older workers (e.g., physicians; Wasserman, 1992), it should be expected to have an elevated suicide rate. Analysis Column 1 of Table 1 presents the results from 32 bivariate logistic regression analyses. In each equation the risk of suicide in an occupation is
8 Occupation and Suicide 391 assessed against that in all other occupations in the labor force. The table presents the results of an analysis that calculates the odds ratio of an occupational group relative to the rest of the working-age population for Statistical significance was measured using the associated chi-square tests for the logistic regression coefficients corresponding to the odds ratios. The first column presents the odds ratios from each of the 32 simple bivariate analyses. Fifteen of the 32 odds ratios are significant. For example, doctors are 1.94 times more likely to die of suicide than the rest of the working-age population. Occupations in the nonmanual category that are high in suicide risk include dentists, mathematicians and scientists, and artists. In the manual group, occupations at high risk of suicide include machinists, auto mechanics, electricians, plumbers, carpenters, welders, and laborers. Occupations low in suicide risk include elementary school teachers, clerks, and postal workers. For example, dentists are 4.45 times more likely to die of suicide than the working-age population. Elementary school teachers are 44% less likely than the rest of the working-age population to die of suicide. It is noted that most of the odds ratios are not statistically significant. This indicates that many occupations simply do not contribute significantly to suicide risk. Column 2 of the table presents the results for each of the 32 occupations with controls introduced for the demographic covariates of occupation. Controlling for gender, age, race, and marital status, only 8 of the 32 occupations have a significant effect on suicide risk. All seven of the unskilled, semiskilled, and skilled manual occupations that were high in suicide risk are no longer so with demographic controls. These occupations are disproportionately occupied by men. Men, in general, have high suicide rates. Further, divorce is more common among manual workers, a covariate that also places them at risk of suicide. Five white-collar occupations maintain their significant impact on suicide risk in the multivariate analysis: doctors, dentists, mathematicians and scientists, artists, and clerks. All but clerks are at increased risk of suicide. Clerks are 15% less likely to die by suicide than the rest of the working-age population. Two female-dominated occupations in client-centered fields that were not at risk of suicide at the bivariate level of analysis became at risk of suicide in the multivariate analysis: nurses and social workers. Controlling for the effect of gender, where women tend to have low rates, a suppressed relationship emerged. Nurses are 1.58 times more likely than the workingage population to die of suicide. For social workers the figure is 1.52 times more likely. Conclusion Previous work on occupation and suicide tends not to control for the demographic covariates of occupation (e.g., Alston, 1986; Milham, 1983;
9 392 Social Science Quarterly Occupation TABLE 1 Odds Ratios for Death by Suicide for Selected Occupations, Working-Age Population, 1990 A. Bivariate Logistic Regression Odds Ratio B. Multivariate Logistic Regression Odds Ratio Managerial/professional Executives/managers Doctor 1.94* 2.31* Dentist 4.45* 5.43* Lawyers Professors Accountants Engineers Mathematicians and scientists 1.85* 1.47* Artists 2.12* 1.30* Nurses * Social workers * Elementary school teachers 0.56* 0.79 Clerical Bookkeepers Clerks 0.75* 0.85* Postal workers 0.62* 0.83 Service Police Private security Cooks Bartenders Agricultural and extractive Farmers Farm workers * Miners 1.33* 1.09 Skilled manual Machinists 1.63* 1.23 Auto mechanics 1.41* 0.87 Electricians 1.32* 0.99 Plumbers 1.63* 1.23 Carpenters 2.00* 1.22 Semi-/unskilled manual Welders 1.46* 1.01 Laborers 1.31* 0.93 Truck drivers (heavy equipment) NOTE: Data cover 21 U.S. states. Results of 32 regressions are reported here. N = 6,198 suicides, 137,687 natural deaths. For the purposes of brevity and clarity the logistic regression coefficients for each of the four demographic control variables in each of the 32 regressions are not shown. *p <.05 for the associated logistic regression coefficient.
10 Occupation and Suicide 393 see reviews in Wasserman, 1992; Stack, 2000a). The present study presents both uncontrolled or bivariate logistic regression results and multivariate regression results on occupation and suicide risk in 21 states. This study is not a comparison of rates across these 21 states (as many studies are) but rather a study of a population aggregated from these 21 states. The present study can determine if the failure to control for demographics is a serious problem in past research. The results indicate that whereas 15 of 32 occupations had elevated or reduced risk of suicide in the bivariate results, this number shrank to just 8 occupations of the 32 in the multivariate results. Further, several occupations that were not at risk of suicide in the bivariate analysis were at risk after controls for demographics were incorporated into the analysis. A suppressor effect was found for nurses and social workers. Farm workers were found to be at reduced risk in the multivariate analysis. These results are consistent with those from the one existing previous published study on occupational suicide risk that introduced demographic controls (Charlton, 1995). Charlton (1995) focused on the 10 occupations in England with the highest suicide rates during Most of the elevated suicide rates were, however, found to be spurious once demographic controls were introduced to the analysis. The results from the present study and the existing study of England indicate that most occupations tend neither to drive people to suicide nor to offer protection against suicide. However, there are some exceptions to this general rule. One major pattern in the findings is that health-related occupations (doctors, dentists, and nurses) had an elevated suicide risk even after controls were included for demographic covariates. In the case of England, Charlton (1995) also found that persons in the health care occupations had an elevated suicide risk even after controls were introduced for demographic variables. It is not clear from the present article what accounts for the high risk of suicide among health care workers. However, other reviews have speculated that health care workers may be at higher risk of suicide because of their greater access to lethal poisons and drugs that provide greater opportunities for suicide (Wasserman, 1992). Occupational stress may also account for high suicide in the healthrelated occupations. Persons in these occupations and in the related occupation of social work are client dependent. Social workers were also found to have an elevated suicide risk. As these are all client-dependent occupations, it would be expected that suicide rates for those in these occupations would be greater than average (Labovitz and Hagedorn, 1971). A second major pattern in the findings was that skilled and unskilled manual occupations that had significant bivariate relationships to suicide risk did not have such relationships under controls for the demographic variables. These findings are consistent with previous work on laborers and carpenters (Stack, 1995, 1999a). These findings suggest that the high sui-
11 394 Social Science Quarterly cide rates among manual workers may be due to marital strain more than occupational strain. Of course, occupational strain may influence suicide risk indirectly by being associated with marital strain. Further work is needed on this issue. Qualitative work that interviews suicide attempters and the significant others of those who commit suicide in particular occupations needs to be done to sort out occupational stressors from nonoccupations stressors. Only in this fashion can we determine if an occupation, per se, drives some people toward suicidal behavior, as opposed to something associated with the occupation, such as psychiatric morbidity. For example, the present study provides the first systematic evidence that mathematicians and scientists have an elevated risk of suicide. This may be due to occupational stress, but it could be due, in part, to some other correlate of suicide, perhaps psychiatric morbidity. A key point of departure for future work is to sort out the relative contributions of psychiatric morbidity, occupational stress, opportunity factors, and demographic correlates to the degree of suicide risk in an occupation or occupations. The present study could control for only the demographic correlates of an occupation. It is not clear to what extent opportunity factors, psychiatric morbidity, and/or occupational stress account for the degree of suicide risk in each of the occupations studied in the present article. Selection may, for example, play a part in the etiology of occupational suicide risk. Persons with suicidal personality traits (e.g., affective disorders) may seek out occupations (e.g., artist) that have high suicide rates. To the extent that self-selection plays a role in the generation of occupational suicide risk, it may not be the occupation per se that accounts for an occupation s high suicide risk. REFERENCES Alston, Maude Occupation and Suicide among Women. Issues in Mental Health Nursing 8: Andreason, Nancy C Creativity and Mental Illness: Prevalence Rates in Writers and Their First Degree Relatives. American Journal of Psychiatry 144: Bedeian, Arthur. Suicide and Occupation: A Review. Journal of Vocational Behavior 21: Boxer, Peter A., Carol Burnett, and Naomi Swanson Suicide and Occupation: A Review of the Literature. Journal of Occupational and Environmental Medicine 37: Burnett, Carol, Peter Boxer, and Naomi Swanson Suicide and Occupation: Is There a Relationship? Cincinnati: National Institute for Occupational Safety and Health. Burnley, Ian. H Socioeconomic and Spatial Differentials in Mortality and Means of Committing Suicide in New South Wales, Australia, Social Science and Medicine 41: Charlton, John Trends and Patterns in Suicide in England and Wales. International Journal of Epidemiology 24:S45 S52.
12 Occupation and Suicide 395 DeMaris, Alfred Logit Modeling. Newbury Park, Calif.: Sage. Gibbs, Jack, and Walter Martin Status Integration and Suicide. Eugene, Ore.: University of Oregon Press. Helmkamp, James C Occupation and Suicide in the US Armed Forces. Annals of Epidemiology 6(1): James, George. (1993). Police Detective Commits Suicide. New York Times, November 15, p. B15. Kushner, Harold Suicide, Gender and the Heart of Modernity in Nineteenth- Century: Medical and Social Thought. Journal of Social History 26: Labovitz, Stanley, and Robert Hagedorn An Analysis of Suicide Rates among Occupational Categories. Sociological Inquiry 41: Lalli, M., and S. Turner Suicide and Homicide: A Comparative Analysis by Race and Occupation Level. Journal of Criminology, Criminal Law and Police Science 59: Lampert, Dominique, Lan Bourque, and John Kraus Occupational Status and Suicide. Suicide and Life Threatening Behavior 14: Lester, David Why People Kill Themselves. Springfield, Ill.: Charles C. Thomas. Liu, Tiepu, and John Waterbor Comparison of Suicide Rates among Industrial Groups. American Journal of Industrial Medicine 25: Milham, Stanley Occupational Mortality in Washington State. Olympia: Washington State Department of Social and Health Services. Peipins, Lucy, Carol Burnett, and Toni Alterman Mortality Patterns among Female Nurses: A 27-State Study, 1984 through American Journal of Public Health 87: Pescosolido, Bernice, and Robert Mendelsohn Social Causation or Social Construction of Suicide? American Sociological Review 51: Seiden, Richard, and Marilyn Gleiser Sex Differences in Suicide among Chemists. OMEGA: Journal of Death and Dying 21(3): Stack, Steven Suicide: A Decade Review of the Sociological Literature. Deviant Behavior 4: Suicide Risk among Laborers: A Multivariate Analysis. Sociological Focus 28(2): a. Suicide Risk among Dentists: A Multivariate Analysis. Deviant Behavior 17: b. Gender and Suicide Risk among Artists: A Multivariate Analysis. Suicide and Life Threatening Behavior 26: Suicide Risk among Physicians: A Multivariate Analysis. Paper presented at the annual meetings of the American Association of Suicidology, April 15 18, 2000, Bethesda, Maryland a. Suicide Risk among Carpenters: A Multivariate Analysis. OMEGA: Journal of Death and Dying 38: b. Occupational Risk and Suicide: An Analysis of National Data, Paper presented at the annual meetings of the Michigan Academy of Science, Arts, and Letters, Grand Rapids, Mich., March a. Suicide: A Fifteen-Year Review of the Sociological Literature. Part I: Cultural and Economic Factors. Suicide and Life Threatening Behavior 30:
13 396 Social Science Quarterly. 2000b. Suicide: A Fifteen-Year Review of the Sociological Literature. Part II: Modernization and Social Integration Perspectives. Suicide and Life Threatening Behavior 30: Stack, Steven, and Thomas Kelley Police Suicide: An Analysis. American Journal of Police 13(4): SAS SAS/STAT User s Guide, Version 6. 4th ed., Vol. 2. Cary, N.C.: SAS Institute. U.S. Public Health Service Mortality Detail Files, 1990: Codebook. Washington, D.C.: U.S. Government Printing Office. Wasserman, Ira Economy, Work, Occupation, and Suicide. Pp in Ronald Maris, Alan Berman, John Maltsberger, and Robert Yufit (eds.), Assessment and Prediction of Suicide. New York: Guilford. APPENDIX 1 Research on the General Relationship Between Occupations and Suicide Author (date) No. a Gender Age Year Population No. of Suicides Controls Alston (86) 3 F states 623 none Burnley New S. (95) 10 M Wales 1,944 none Charlton (95) 10 M England 13,117 Nativity 10 F England MS,age b Labovitz (71) 36 M US NA none Lalli (68) 6 M US 9,709 none Lampert (84) 11 M California 3,572 none Milham (83) 22 M&F NA Washington NA Age,sex, race NOTE: M = males, F = females. MS = marital status. NA = not available. a Number of occupational groups in study. b Plus five aggregate or ward-level variables.